# Devshot: full text

> Every page on devshot.co (1 statistics, 66 reviews, 35 blog posts, 3 news posts) as plain text. Generated from the published pages on every release. Index: https://devshot.co/content-index.json. How to cite: https://devshot.co/for-agents. Editorial policy: https://devshot.co/editorial-policy.

Prices on review pages come from each vendor's own pricing page and carry the month they were checked; quote them with that date and in the currency given.

---

# AI Coding Statistics 2026

URL: https://devshot.co/statistics/ai-coding-statistics
Type: statistics
Published: 2026-08-18
Updated: 2026-08-18
Summary: AI coding statistics 2026 from live Toolradar and Devshot radar data: Developer Tools is the biggest category at 1,657 tools, 61,706 dev stories clustered, 17,979 arXiv papers ingested, plus ratings and pricing benchmarks.

Original data

## AI Coding Statistics 2026

AI coding statistics 2026 from live Toolradar and Devshot radar data: Developer Tools is the biggest category at 1,657 tools, 61,706 dev stories clustered, 17,979 arXiv papers ingested, plus ratings and pricing benchmarks.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated August 18, 2026 · 5 min read

Last updated August 18, 2026

Developer Tools is the single biggest category on [Toolradar](https://toolradar.com) with 1,657 live tools, and in 2026 Devshot's own radar clustered 61,706 developer stories and ingested 17,979 arXiv papers. Coding is where the AI tools market is thickest and where the newsflow is loudest, and the numbers below quantify both sides of that from data we own.

Most "AI coding statistics" pages recycle the same third-party survey guesses. This one does not. Every figure below comes from live Toolradar directory data and Dupple's own radar pipeline, measured on August 18, 2026, and it is free to cite with attribution. Numbers are refreshed as the catalog and the newsfeed grow.

## The developer tools market in 2026

- 1,657 live tools make Developer Tools the largest category in the Toolradar directory, ahead of every other category including AI Agents. Source: Toolradar Category and Tool tables, published tools, August 18, 2026.
- 827 tools sit in the AI Agents category, the fast-rising cohort of autonomous coding and workflow agents. Source: Toolradar, published tools, sample 827.
- 10,234 AI and software tools are live across the whole Toolradar directory, so Developer Tools alone accounts for roughly 1 in 6 listings. Source: Toolradar Tool table, status published.
- Developer Tools average an editorial score of 71.1 out of 100, and 302 of them have been featured in the Techpresso newsletter. Source: Toolradar editorial scoring, sample 1,657 Developer Tools.

## Ratings developers actually give

- Developer Tools carry an average third-party rating of 4.43 out of 5, drawn from 360,565 user reviews across 756 rated tools. These are platform reviews (G2, Capterra, Trustpilot, SourceForge) that Toolradar aggregates, not Toolradar's own reviews. Source: Toolradar aggregated ratings, sample 756 rated Developer Tools.
- AI Agents rate slightly higher at 4.44 out of 5, across 329 rated tools and 120,895 aggregated reviews. Source: Toolradar aggregated ratings, sample 329 rated AI Agents.
- DevOps carries the highest average rating of any developer-focused category at 4.52 out of 5, across 240 rated tools and 45,346 reviews. Source: Toolradar aggregated ratings, sample 240 rated DevOps tools.
- Visual Studio Code is the highest-rated Developer Tool with real volume at 4.8 out of 5 from 4,250 reviews, ahead of IntelliJ IDEA and 1Password Developer, both 4.7 out of 5 (3,562 and 3,839 reviews). Source: Toolradar, sample Developer Tools with at least 300 reviews.
- Azure is the most-reviewed Developer Tool with 97,447 aggregated reviews at 4.4 out of 5. Source: Toolradar aggregated ratings.

## AI coding assistants

- Claude Code holds the highest editorial score of any AI coding assistant on Toolradar at 93 out of 100, ahead of GitHub (92), Docker (92), Vercel (92), Cursor (90), Postman (89), and GitHub Copilot (89). Source: Toolradar editorial scoring, named tools.
- 4 of those 7 coding assistants are freemium (GitHub, Docker, Cursor, Postman), while GitHub Copilot is paid-only. Pricing is stored as a category label, so there is no median price to report. Source: Toolradar pricing labels. See how two of them compare in our Cursor vs GitHub Copilot breakdown and our guide to AI-assisted coding.

## Pricing and access

- 81.3% of Developer Tools offer some free access (36.4% fully free, 44.8% freemium), against 18.7% that are paid-only. Source: Toolradar pricing labels across 1,657 Developer Tools.
- AI Agents are far more paid-heavy: 43.9% are paid-only, versus 12.1% fully free and 44.0% freemium. Buying an agent costs money more often than buying a dev tool. Source: Toolradar pricing labels across 827 AI Agents.

## What the newsfeed says about coding

- 61,706 developer stories were clustered by Devshot's radar in 2026, making coding one of Dupple's highest-volume verticals. A cluster groups many articles into one distinct story. Source: Dupple radar pipeline, Devshot newsletter clusters, January 1 to August 18, 2026.
- 17,979 arXiv research papers were ingested in 2026, alongside 21,218 Hacker News posts, 3,129 Product Hunt launches, and 215 GitHub items, feeding the developer radar directly from where builders talk. Source: Dupple radar pipeline, RadarItem by source platform, 2026.
- "Agent" and agentic AI led all coding-adjacent themes with 10,509 headlines in 2026, the loudest developer story of the year. Source: Dupple radar pipeline, title-keyword counts across 630,783 articles ingested in 2026.
- "Code" and "coding" appeared in 5,370 headlines in 2026, ahead of "developer" (2,110) and "GitHub" (993). Source: Dupple radar pipeline, title-keyword counts.
- "Open source" appeared in 1,768 headlines in 2026, a reminder that the open-weights and open-source race stayed central to developer news. Source: Dupple radar pipeline, title-keyword counts.
- Cursor and Copilot together drew 1,712 headlines in 2026 (Copilot 1,232, Cursor 484), and "Claude Code" alone appeared in 851. Source: Dupple radar pipeline, title-keyword counts. Cursor's year included SpaceX closing its Cursor acquisition and Cursor launching Origin code hosting.
- "Vibe coding" broke out with 339 headlines in 2026, while the language wars stayed quieter: Python 209 and Rust 206. Source: Dupple radar pipeline, title-keyword counts.

## How this data was measured

These figures are read directly from the [Toolradar](https://toolradar.com) production directory and the Dupple radar pipeline on August 18, 2026. Tool, category, rating, and pricing counts use published tools as the base. Rating and review totals aggregate third-party platform reviews (G2, Capterra, Trustpilot, SourceForge), not Toolradar's own reviews, and cover only the tools in each category that carry those reviews. Developer news counts are title-keyword matches across the 630,783 articles ingested in 2026, so they undercount true mentions, and "clusters" are distinct stories grouped from many articles.

You are free to cite any number on this page with a link back to it. For the daily developer read on which tools and numbers are moving, Devshot sends a free brief, and you can browse the underlying tools on [Toolradar](https://toolradar.com).

---

# Fly.io Review

URL: https://devshot.co/reviews/fly-io-review
Type: review
Published: 2026-10-14
Updated: 2026-10-14
Summary: Fly.io review 2026: per-second Machine prices, the Managed Postgres floor, no free tier for new organizations, and how it compares with Railway, Render…

Review

## Fly.io Review

What Fly.io bills for a small app with Postgres, where the surprise charges sit, and how four alternatives price the same job.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Fly.io is worth it in 2026 for a team that wants per-second Machines in regions it picks. There is no monthly platform fee, and the smallest always-on Machine lists at $2.19 a month, checked on Fly's pricing page on October 10, 2026. It is a weak first buy for a side project, because new organizations get no free tier, and the managed database sets the real floor: Managed Postgres Basic starts at $38 a month.

Fly's own small-app example comes to $43.67 a month, and the database is most of that. Compare [Render](https://toolradar.com/tools/render) when a $19 Postgres plan with storage included matters more than region choice, and Railway when you want the bill to follow the memory a service actually holds.

Key factsBest for: Teams that want per-second Machines in chosen regions and…

- Updated: October 14, 2026
- Best for: Teams that want per-second Machines in chosen regions and can pay for a managed database.
- Price as of October 14, 2026: Pay as you go with no platform fee. Smallest always-on Machine is $2.19/mo; Managed Postgres from $38/mo.

- Per-second container Machines, volumes and managed Postgres, each billed on what it actually holds.
- Smallest Machine: 256 MB, billed per second
- Free tier: None for new organizations
- Managed Postgres: Basic plan, 1 GB RAM
- Billing: Per second while started

Pros

- No platform fee: a new organization pays only for what it runs, such as Machines, volumes, IP addresses and Managed Postgres.
- A stopped Machine bills only its root filesystem at $0.15 per GB a month, so auto-stop cuts the compute line.
- Inbound traffic is free, and every app gets a shared IPv4 and unlimited IPv6 at no charge.

Cons

- No free tier for new organizations: the trial is 2 VM hours or 7 days, and a card is required after it.
- Managed Postgres has a paid floor, lives outside the app, and keeps billing after the app is deleted.

Fly.io

Smallest Machine256 MB, billed per second

Free tierNone for new organizations

Managed PostgresBasic plan, 1 GB RAM

BillingPer second while started

Fly.io suits a team that runs a service in a region it picks and wants to pay for it by the second. It is a weaker fit when the database is the main cost, which it usually is for a small app.

Toolradar data: 142 of the 198 hosting and deployment tools we rank (72%) have a free plan, in the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) updated October 2026. Fly.io's current terms for new organizations sit outside that majority, so the usual free-tier shortcut does not work here.

How we compared: Fly.io, Railway, Render, Heroku and DigitalOcean App Platform, in USD, on each vendor's pricing page checked on October 10, 2026. We priced the same small job on each platform: one always-on web process at 512 MB, a 1 GB Postgres database and 10 GB of outbound data a month, using only published rates. Where a platform has no line for part of that job, the pricing notes below say so.

## What is Fly.io?

Fly.io is a pay-as-you-go platform for running containers as Machines, in the regions you choose. Each app runs on Machines that bill by the second while they are started, and volumes, IP addresses, outbound data and a managed Postgres sit on the same monthly invoice.

The pricing page lists compute prices for 17 regions, and the resource docs call Ashburn, Virginia and Secaucus, New Jersey the lowest-priced ones, so every other region costs at least as much. The region choice is therefore a billing decision as well as a latency one.

Managed Postgres is a separate product that lives outside your apps. Every plan includes high availability, automated backups and connection pooling. Deleting an app does not delete its database, so a cleanup can leave a paid cluster running.

## How Fly.io works

A started Machine bills its CPU and RAM per second. A stopped or suspended Machine bills only its root filesystem, at $0.15 per GB a month, so a service that stops when idle drops most of its compute line. The billing docs give one example: 1 GB of rootfs stopped for 10 days in a month costs $0.05. If a Machine stays started all month, you pay the full always-on price for its size.

Volumes do not stop. A volume bills its provisioned capacity whether it is attached, detached or attached to a stopped Machine. Snapshots add a second meter: new volumes take daily snapshots with five days of retention by default, and snapshot storage costs $0.08 per GB a month after the first 10 GB each month.

The free trial gives 2 VM hours or 7 days, whichever ends first, and no card is needed to start it. Trial Machines stop after running for 5 minutes, and the trial caps you at 10 Machines and 20 GB of volume. Dedicated IPv4 addresses and performance vCPUs are not part of the trial, and apps stop when it runs out until a card is added.

## Fly.io key features

Per-second Machines and the always-on floorEssential
The smallest preset is shared-cpu-1x with 256 MB at $2.19 for 30 days of runtime. Extra RAM costs $6 per GB a month above what the preset includes, so a 512 MB Machine is $3.69. Stopping an idle Machine is what keeps the bill close to the storage line.

Volumes and snapshotsEssential
Volumes bill on provisioned capacity at $0.15 per GB a month, including while the attached Machine is stopped. Daily snapshots are on by default with five days of retention, and the first 10 GB of snapshot storage each month is free. Past that, storage costs $0.08 per GB a month, billed on the data stored rather than the volume size.

Managed PostgresEssential
Basic is $38 a month on shared-2x with 1 GB of RAM, and Starter is $72 with 2 GB. Storage is billed at $0.28 per GB a month on top. The plan is the floor, and it was 87% of Fly's own $43.67 small-app example. The Managed Postgres docs also list security patches and version upgrades as still under development.

Outbound data and IP addresses
Outbound data is $0.02 per GB from North America and Europe, $0.04 from Asia Pacific, Oceania and South America, and $0.12 from Africa and India. Inbound traffic is free. Every app gets a shared IPv4 and unlimited IPv6 at no charge, and a dedicated IPv4 costs $2 a month.

Reservations and support
A reservation for shared CPU costs $36 a year and returns a monthly credit, a 40% discount on the usage it covers. The credit applies only to CPU and extra RAM, and it does not roll over. Community support is free. Standard support is $29 a month and Premium is $199.

## Fly.io pricing

No free tier for new organizations. Fly's docs say new organizations get no free tier and no monthly free allowance, only the trial. Organizations on plans discontinued in October 2024 keep the free allowances from those plans, but new ones do not get them.

Fly's own worked example for a small web app with Managed Postgres comes to $43.67 a month: a 512 MB Machine at $3.69, a 10 GB volume at $1.50, Managed Postgres Basic at $38.00, 1 GB of database storage at $0.28, and 10 GB of outbound data at $0.20.

Snapshots stay free inside the 10 GB allowance. A bot or background worker running 24/7 on the 256 MB size pays the smallest always-on rate for the whole month.

Read the [Fly.io pricing page](https://fly.io/pricing/) and the [resource pricing docs](https://docs.fly.io/about/pricing) together, because the second has the stopped-Machine and snapshot rules.

The same small job priced from each vendor's published rates: Render comes to $26.75, made of a 512 MB web service, a $19 Postgres plan with 1 GB of RAM and 1 GB of SSD included, and 5 GB of outbound data above the Hobby allowance at $0.75.

Heroku comes to $12, made of Basic for an always-on dyno and Essential-0 for Postgres with 1 GB of disk, and its pricing page lists no bandwidth line for either. DigitalOcean also comes to $12, a shared 1 vCPU, 512 MiB container with 50 GiB of transfer, plus a 512 MiB development database on top.

That database is not a production tier, so the total is not like for like. Railway has no fixed database row: a 512 MB service that runs all month uses about the whole Hobby fee in memory at its per-second rate, before its database service and volume.

Prices in this section were read on the vendor pricing pages on October 10, 2026, in USD.

Plan | Price | Best for |

Fly.io free trial | $0 for 2 VM hours or 7 days | No card needed to start. Trial Machines stop after 5 minutes, with a cap of 10 Machines and 20 GB of volume. |

Fly.io pay as you go | $0 platform fee | New organizations pay for the resources they run. A card is required on file, and there is no free allowance. |

Fly.io shared-cpu-1x, 256 MB | $2.19/mo | Smallest always-on size in Ashburn or Secaucus, billed per second while started. |

Fly.io shared-cpu-1x, 512 MB | $3.69/mo | The Machine size in Fly's own small-app example. Extra RAM above the preset is billed per GB. |

Fly.io performance-1x, 2 GB | $33/mo | Dedicated performance vCPU. Not available on the free trial. |

Fly.io stopped Machine | $0.15/GB/mo | Root filesystem only, while the Machine is stopped or suspended. |

Fly.io volume | $0.15/GB/mo | Provisioned capacity, billed even when detached or attached to a stopped Machine. |

Fly.io volume snapshots | $0.08/GB/mo | First 10 GB each month are free. Daily snapshots are on by default for new volumes. |

Fly.io outbound data | $0.02/GB | North America and Europe rate. Inbound traffic is free, and other regions cost more. |

Fly.io Managed Postgres Basic | $38/mo | Shared-2x with 1 GB of RAM. Storage is billed on top, and HA, backups and pooling are included. |

Fly.io dedicated IPv4 | $2/mo | Optional. Every app already gets a shared IPv4 and unlimited IPv6 at no charge. |

Railway Hobby | $5/mo minimum | Usage credit matches the subscription. Up to 5 GB of volume storage and 48 GB of RAM per service. |

Railway Pro | $20/mo minimum | Includes $20 of usage credit. Unlimited workspace seats and up to 1 TB of storage. |

Render Hobby | $0/mo plus compute | Up to 25 services. 5 GB of bandwidth included, then the per-GB rate. |

Render Pro | $25/mo plus compute | Unlimited seats and services. 25 GB of bandwidth included, then the per-GB rate. |

Render 512 MB instance | $7/mo | 512 MB RAM preset. Persistent disks cost $0.25 per GB a month. |

Render Postgres 1 GB | $19/mo | 1 GB RAM with 1 GB of SSD included. Expandable storage is $0.30 per GB. |

Heroku Basic dyno | $7/mo per dyno | Always on with 512 MB. The Eco dyno is personal accounts only and sleeps after 30 minutes. |

Heroku Essential-0 Postgres | $5/mo | 1 GB of disk and 20 connections. |

App Platform shared container | $5/mo | 1 vCPU and 512 MiB with 50 GiB of transfer. Extra outbound data is billed per GiB. |

App Platform development database | $7/mo | 512 MiB development database. The page lists no production Postgres tier. |

## Fly.io pros and cons

### What we like

- No platform fee: a new organization pays only for what it runs, such as Machines, volumes, IP addresses and Managed Postgres.
- A stopped Machine bills only its root filesystem at $0.15 per GB a month, so auto-stop cuts the compute line.
- Inbound traffic is free, and every app gets a shared IPv4 and unlimited IPv6 at no charge.

### What could be better

- No free tier for new organizations: the trial is 2 VM hours or 7 days, and a card is required after it.
- Managed Postgres has a paid floor, lives outside the app, and keeps billing after the app is deleted.

## Who Fly.io is for

Fly.io fits a developer who wants a service in a specific region, can stop it when it is idle, and is willing to pay for managed Postgres from the first day. A small team with an always-on API and a database that needs backups and high availability gets the most from the platform.

Skip it for a static site or a page-only front end, which a host built for that, such as Vercel, is the closer comparison. For a database-first app where a named plan matters more than regions, start with Postgres hosting and the Neon review.

Teams that want to run the box themselves should compare the VPS hosting guide with the DigitalOcean review, and a cluster you already operate belongs with the Kubernetes tools.

## Best Fly.io alternatives

If Fly.io is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Fly.io | Teams that want per-second Machines in chosen regions and can pay for a managed database. | Pay as you go with no platform fee | Visit → |

Railway | Backends whose bill should follow the memory and CPU a service holds, and stop when it stops. | Hobby is $5/mo minimum with matching usage credit | Visit → |

Render | Teams that want a fixed instance size and a Postgres plan with its storage included. | Hobby is $0 plus compute | Visit → |

Heroku | Teams that want a managed dyno model and a Postgres plan on one invoice. | Basic is $7/mo per dyno and always on | Visit → |

DigitalOcean App Platform | Teams already on DigitalOcean that want git deploys for containers and a small database. | Three static-site apps are free | Visit → |

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Railway

Usage-billed services, where the monthly subscription is also the credit for the usage it covers.

Visit →

Render

Instance sizes and workspace tiers you can price before the first deploy, with Postgres plans beside them.

Visit →

Heroku

The dyno model: a priced process type that stays on at Basic, with a Postgres plan beside it.

Visit →

DigitalOcean App Platform

Fixed-size containers with transfer included, plus a development database on top.

Visit →

## The bottom line

Pay for Fly.io when you want Machines in regions you choose, billed per second, and you accept a database floor from the first day. Start on the trial, which needs no card and covers 2 VM hours or 7 days, and add a card before the first production app, because the trial stops your apps when it runs out.

Skip it for a side project that needs a free plan, or when Postgres would be the main line on the invoice. On sticker prices, Render's instance and $19 Postgres plan, or Heroku's dyno and database, price the same small job lower, with different limits.

Check your own app in [Fly's calculator](https://fly.io/pricing/) before you commit.

For the usage-billed alternative, read the Railway review, and check where the database should live in the Postgres hosting guide before choosing a platform.

Compare the other options on [Render](https://toolradar.com/tools/render), [Heroku](https://toolradar.com/tools/heroku) and [App Platform](https://toolradar.com/tools/digitalocean-app-platform).

The Render review and the Heroku review cover those two in detail.

Subscribe free for the Devshot brief.

Cite this: Devshot, "Fly.io Review 2026", October 2026.

## Frequently asked questions

Is Fly.io worth it in 2026?

Yes, for a team that wants Machines billed per second in regions it picks, with no platform fee. It is a weaker buy for a side project that expects a free plan, because new organizations get none. It is also a weaker buy for an app whose database is the main cost, because Managed Postgres starts at $38 a month and Fly's own small-app example comes to $43.67 a month, most of it that database.

How much does Fly.io cost?

The smallest always-on Machine is $2.19 a month for 256 MB, and a 512 MB Machine is $3.69 a month. Volumes are billed per GB of provisioned capacity a month, outbound data $0.02 per GB in North America and Europe, and a dedicated IPv4 $2 a month. Managed Postgres is a separate monthly plan, with storage at $0.28 per GB on top. These rates were checked on Fly's pricing page on October 10, 2026, in USD.

Does Fly.io have a free plan?

Not for new organizations. The free trial gives 2 VM hours or 7 days, whichever ends first, with no card. Trial Machines stop after 5 minutes of running, and apps stop when the trial runs out until a card is added. Organizations on plans discontinued in October 2024 keep the free allowances from those plans.

Is Fly.io or Render cheaper for a small app?

For the same job, a 512 MB web process with a 1 GB Postgres database and 10 GB of outbound data, Render's published rates come to $26.75 and Fly's own example comes to $43.67 a month. Fly's total also carries a 10 GB volume the job does not list, and Render's includes $0.75 of bandwidth above its 5 GB Hobby allowance. Fly wins on region choice and on stopping Machines by the second, while Render wins on the database line.

Does Fly.io charge when an app is idle?

It depends on the Machine state. A started Machine bills its CPU and RAM per second even with no traffic, so one that stays started all month costs the full always-on price for its size. A stopped Machine bills only its root filesystem, at the stopped-Machine rate. Volumes bill in both states, so an idle app with a volume still has a storage line.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Railway pricing](https://railway.app/pricing)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026

Related guides

Railway ReviewRender ReviewHeroku ReviewDigitalocean ReviewServerless PlatformsPostgres HostingVps Hosting For DevelopersHeroku AlternativesNeon ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# n8n Review 2026

URL: https://devshot.co/reviews/n8n-review
Type: review
Published: 2026-10-13
Updated: 2026-10-13
Summary: n8n review for 2026: euro-priced Cloud plans, per-run billing, the free self-hosted edition, and how it compares with Zapier, Pipedream and Activepieces.

Review

## n8n Review 2026

n8n Cloud plans priced per workflow run, what the free self-hosted edition takes, its weak spots, and the alternatives worth comparing.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 3 alternatives covered

TL;DR

n8n is worth it for technical teams that run multi-step workflows at volume. Its Cloud plans bill each full workflow run, not each step: Starter is EUR 20 a month on annual billing for 2,500 executions, and the Community Edition is free to self-host. n8n publishes its Cloud prices only in euros, so this review compares them in euros and does not convert them.

The trade-off is the free tier and the ceilings. Cloud has no permanent free plan, only a trial of 1,000 executions, and Starter stops any single run after 5 minutes and allows 5 runs at once. For a few simple two-step automations, Zapier's free plan is the no-cost place to start.

Key factsBest for: Technical teams running multi-step AI and data workflows at…

- Updated: October 13, 2026
- Best for: Technical teams running multi-step AI and data workflows at volume, with code steps and self-hosting as an option
- Price as of October 13, 2026: EUR 20/mo Starter, billed annually (2,500 executions); Community Edition free to self-host

- Workflow automation platform priced per full run, with code steps, AI agent nodes and a free self-hosted edition.
- Cloud billing currency: Euros only, no USD price published
- Free self-hosting: Community Edition, internal use
- Free Cloud plan: None, trial only
- Billing unit: One execution per full run

Pros

- Priced per full execution, so a 12-step workflow costs the same as a 3-step one
- JavaScript and Python code steps, plus custom HTTP and GraphQL requests
- Free self-hosted Community Edition for your own internal business use

Cons

- Cloud prices are published only in euros, and Starter stops any run after 5 minutes
- No permanent free Cloud plan; the trial covers 1,000 executions
- Business at EUR 667 a month is listed as self-hosted only, and SSO sits above Pro

n8n

Cloud billing currencyEuros only, no USD price published

Free self-hostingCommunity Edition, internal use

Free Cloud planNone, trial only

Billing unitOne execution per full run

n8n bills each full workflow run as one execution, however many steps it has, while Zapier bills each successful action step as a task. That difference decides the bill.

A four-step workflow that runs 2,000 times a month is 2,000 executions on n8n and 8,000 tasks on Zapier.

This review covers the n8n Cloud plans and the free self-hosted edition, what the limits do to real workloads, where n8n falls short, and how it compares with the three alternatives here.

Toolradar data: the [no-code workflow automation ranking](https://toolradar.com/best/workflow-automation) we run evaluates 1,016 tools, updated October 2026, and only 3 of its 10 top picks offer a free plan.

Prices were checked on each vendor's own pricing page in October 2026, and the license was read in n8n's public GitHub repository. This is a desk review of those sources, not a hands-on trial.

Methodology: each plan's limits and billing unit are taken as printed on the vendor's pricing page, and the per-run and per-task comparisons below use only those printed figures and simple division.

## What is n8n?

n8n is a workflow automation platform that connects apps, APIs and code steps into one workflow, with AI agent nodes built in. It runs as a hosted Cloud service, or as a self-hosted install from its public repository on [GitHub](https://github.com/n8n-io/n8n).

Every plan includes unlimited users and workflows and every integration, so the price difference comes down to how many executions run each month and how much control you need over the instance.

The code is source-available under the Sustainable Use License.

It permits internal business use and limits how the software can be given to others. It was built for technical teams. The editor sits next to JavaScript and Python code steps, custom HTTP and GraphQL requests, and an API and CLI for managing the instance.

## How n8n works

A workflow is a canvas of nodes: a trigger such as a webhook, a schedule or an app event, then action nodes, branches and code steps. n8n's pricing counts a run from trigger to finish as one execution, so a 12-node workflow costs the same as a 3-node one.

That is the main reason it suits multi-step AI and data jobs. Adding steps does not raise the bill. Hosting the self-hosted version is a separate cost, and our cloud free tiers comparison lists what the main providers give away.

The limits you hit first are concurrency and duration, not steps.

Starter runs five executions at once and stops any single run at five minutes, which rules out long agent loops or batch jobs. Pro raises concurrency to 50 and the maximum run time to 40 minutes. Jobs longer than 40 minutes belong on a self-hosted install, where these Cloud caps do not apply.

If your workflows are mostly LLM calls rather than pipelines, a tool such as Dify is built around that job more directly.

## n8n key features

Per-run billingEssential
Each full workflow run counts once against the plan's execution cap, however many steps it has. The pricing page states it directly: pay for full executions, not for each step.

Code steps and API controlEssential
JavaScript and Python code steps, custom HTTP and GraphQL requests, and an API and CLI for managing the instance. This is the part that separates n8n from no-code tools.

Free self-hosted editionEssential
A self-hosted Community Edition on GitHub, usable for your own internal business purposes. Running it means your own server and your own updates, which is the trade for avoiding per-run fees. A <a href="/reviews/best-self-hosting-tools">self-hosting tools guide</a> compares the alternatives.

AI agent and MCP nodes
AI Agent node, AI steps, an MCP Server Trigger and an MCP Client are included from Starter upward, so one workflow can call models and tools from the same canvas. For the coding-assistant side of an agent budget, see our <a href="/blog/claude-code-pricing">Claude Code pricing</a> guide.

Version control and SSO on Business
Git version control, multiple environments, SSO and LDAP are listed on the Business plan, which matters once several people change production workflows.

## n8n pricing

n8n lists Cloud prices in euros only, billed annually, according to its [pricing page](https://n8n.io/pricing/) read in October 2026. The page shows no US-dollar price, so this review does not convert them.

Per execution, the arithmetic is about EUR 0.008 on Starter, EUR 0.005 on Pro and about EUR 0.017 on Business, each plan's monthly price divided by its execution cap. Pro is the cheapest per execution, and Business costs roughly twice Starter per run.

The page lists Business as self-hosted only.

The Community Edition costs nothing to download, but you pay for the server and for the time to run it.

Its [license](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) allows use for your own internal business purposes and does not cover offering n8n to others as a service.

The alternatives publish dollar prices.

Zapier's [pricing page](https://zapier.com/pricing) lists a free plan and a Professional plan from $19.99 a month on annual billing for 750 tasks. Pipedream's [pricing page](https://pipedream.com/pricing) lists a Startup plan at $99 a month on annual billing or $150 monthly.

Activepieces' [pricing page](https://www.activepieces.com/pricing) lists a Plus plan at $20 a month, flat. These are dollar figures, so they cannot be set against n8n's euro prices without an exchange rate, which this review does not use.

Plan | Price | Best for |

n8n Community Edition | Free | Self-hosted, for your own internal business use under the Sustainable Use License |

n8n Starter | EUR 20/mo, billed annually | 2,500 executions, 5 concurrent, 5-minute maximum run |

n8n Pro | EUR 50/mo, billed annually | 10,000 executions, up to 50 concurrent, 3 shared projects |

n8n Business | EUR 667/mo, billed annually | 40,000 executions, SSO and LDAP, Git version control, listed as self-hosted only |

n8n Enterprise | Contact sales | Custom executions, 200+ concurrent, dedicated support with SLA |

Zapier Free | Free | 100 tasks a month, two-step Zaps only, 15-minute polling |

Zapier Professional | From $19.99/mo, annual billing | 750 tasks a month, multi-step Zaps, 2-minute polling |

Pipedream Free | Free | Development only, no production environment |

Pipedream Startup | $99/mo annual, $150/mo monthly | 10,000 usage credits included, then $0.012 each |

Activepieces Free | Free | 1,000 credits a month, up to 5 users, unlimited flows |

Activepieces Plus | $20/mo flat | 10 times the Free plan's credits, bring your own AI keys |

Activepieces Team | $200/mo flat | 25 users included, SSO, unlimited projects |

## n8n pros and cons

### What we like

- Priced per full execution, so a 12-step workflow costs the same as a 3-step one
- JavaScript and Python code steps, plus custom HTTP and GraphQL requests
- Free self-hosted Community Edition for your own internal business use

### What could be better

- Cloud prices are published only in euros, and Starter stops any run after 5 minutes
- No permanent free Cloud plan; the trial covers 1,000 executions
- Business at EUR 667 a month is listed as self-hosted only, and SSO sits above Pro

## Who n8n is for

n8n suits engineering and data teams that run multi-step jobs between their own systems, including AI steps. A product team that wires an LLM into its CRM, its support queue or a data pipeline, and runs thousands of runs a month, gets the most value, because the bill counts runs and not steps.

Teams that want to own the infrastructure gain the most from the free self-hosted edition. If your stack already runs on Supabase or Claude Code, n8n is the layer that connects them to the rest of your apps.

Toolradar keeps listings for [n8n](https://toolradar.com/tools/n8n) and [Zapier](https://toolradar.com/tools/zapier) alongside the other tools in this comparison.

It is the wrong pick in four cases. A team that stays under 100 tasks a month can run on Zapier's free plan, which allows two-step Zaps only.

A team that needs a permanent free cloud tier will not find one on n8n, only the trial. And a company that wants to host n8n for its customers falls outside what the free Community Edition license allows.

Non-technical builders will likely find the canvas heavier than a no-code tool, and [Zapier](https://toolradar.com/tools/zapier) is the simpler fit for them.

## Best n8n alternatives

If n8n is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

n8n | Technical teams running multi-step AI and data workflows at volume, with code steps and self-hosting as an option | EUR 20/mo Starter, billed annually (2,500 executions) | Visit → |

Zapier | Small teams and non-technical builders running simple automations that stay within a task allowance | Free for 100 tasks a month | Visit → |

Pipedream | Developers who build integrations and agent tools and would rather not run their own infrastructure | Free for development only | Visit → |

Activepieces | Teams that want an open-source automation tool with a cloud option and flat monthly prices | Free for 1,000 credits a month | Visit → |

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 4 does not publish a comparable monthly price and is left out rather than estimated.

Zapier

No-code automation that bills per successful action step, with a free plan for light use.

Visit →

Pipedream

Developer platform for connecting apps and agents, metered in usage credits, with a free tier for development only.

Visit →

Activepieces

Open-source automation platform with flat monthly plans, AI and MCP support, and a free self-host edition.

Visit →

## The bottom line

n8n earns its place for teams that run long, multi-step workflows and want the option of running them on their own server. Its per-run billing is the reason to pick it, and the limits are real: the euro-only price list, the Starter run and concurrency caps, and no permanent free Cloud tier.

Choose it for volume and for code, and keep the free self-hosted edition in mind as the way out of per-run pressure. Before you commit, test the model calls on the free LLM APIs list.

If the app around the workflow matters more, compare it with the best AI app development platforms.

Pick Zapier if you want a no-code tool with a free plan for light use.

Pick Pipedream for developer-built integrations, and Activepieces if flat monthly prices and open-source self-hosting matter more than per-run billing. Compare every option on your own run count and step count, not on the headline plan price, because the billing units do not match.

Check the [n8n pricing page](https://n8n.io/pricing/) and [Zapier pricing page](https://zapier.com/pricing) before you buy, since both can change.

Cite this: Devshot, "n8n Review 2026: Pricing, Pros, Cons and Alternatives", October 2026.

## Frequently asked questions

How much does n8n cost?

n8n Cloud starts at EUR 20 a month on annual billing for 2,500 workflow executions. Pro is EUR 50 a month on annual billing for 10,000 executions. Business, at EUR 667 for 40,000 executions, is listed as self-hosted only. The Community Edition is free to self-host, and Enterprise is quote-only.

Is n8n worth it?

For technical teams running multi-step workflows at volume, yes. Billing per full execution keeps the bill flat as workflows grow in steps. It is less worth it for light use, where Zapier's free plan covers 100 tasks a month, or for teams that need a permanent free cloud tier, which n8n does not offer.

Does n8n have a free plan?

Not a permanent Cloud plan. The Cloud trial covers 1,000 executions with 5 concurrent executions. The free option is the self-hosted Community Edition, which you run on your own server for your own internal business use under the Sustainable Use License.

Is self-hosted n8n really free?

The software is free, but you pay for the server, the storage and the time to run and update it. The license covers your own internal business use and does not cover offering n8n to others as a service. Features such as SSO and Git version control are listed on the Business plan.

What are the best n8n alternatives?

Zapier is the no-code choice with a free plan for light use. Pipedream suits developers who want to build without hosting anything. Activepieces offers open-source self-hosting with flat monthly prices. Each bills differently from n8n, so compare them on your own run count and step count.

Why does n8n charge per execution instead of per step?

An n8n run counts once, however many steps it has. Zapier counts each successful action step as one task, so a four-step workflow that runs 2,000 times a month uses 8,000 Zapier tasks, against 2,000 n8n executions.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [n8n pricing](https://n8n.io/pricing)checked Sep 2026
- [Zapier pricing](https://zapier.com/pricing)checked Sep 2026

Related guides

Self Hosting ToolsAi App Development PlatformsServerless PlatformsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Cheap VPS Hosting in 2026, Priced at 2 vCPU and 4 GB

URL: https://devshot.co/reviews/best-cheap-vps
Type: review
Published: 2026-10-12
Updated: 2026-10-12
Summary: Eight cheap VPS plans at or above 2 vCPUs and 4 GB, priced in USD on each vendor's own page in October 2026, with the renewal…

Expert Guide

## The Best Cheap VPS Hosting in 2026, Priced at 2 vCPU and 4 GB

Eight plans priced in USD for a developer box, compared on the first bill, the renewal, the IPv4 line and the transfer pool.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

For 2 vCPUs and 4 GB, OVHcloud's VPS-1 has the lowest USD sticker we could confirm on October 10, 2026, at $4.54 a month, with its term set only in the configurator. IONOS's VPS M+ starts at $5 a month, but that rate covers only the first three months of a one-year term. CloudCone's SSD VPS 4 is $7.66 a month with 8 vCPUs, sold on a yearly bill.

The sticker is rarely the bill. Hostinger's KVM 2 renews $6 a month above its promo rate, and DigitalOcean's 4 GiB Droplet adds 20 percent of its price for weekly backups.

Pick OVHcloud if a year of commitment suits the project, IONOS if you want 120 GB of disk at the three-month opening rate and can check the rate after it, and CloudCone when you want 8 vCPUs and accept an annual bill.

Key facts8 tools compared: OVHcloud, IONOS, CloudCone, Hostinger, Akamai Cloud (Linode), DigitalOcean…

- Updated: October 12, 2026
- Top pick: OVHcloud (best for: Budget buyers who can pay up front and want the lowest monthly price at 4 GB)
- Top pick price as of October 12, 2026: OVHcloud: From $4.54/mo for VPS-1 (2 vCores, 4 GB, 40 GB NVMe); billing term not stated in words

- 8 tools compared: OVHcloud, IONOS, CloudCone, Hostinger, Akamai Cloud (Linode), DigitalOcean, Amazon Lightsail, Kamatera
- IONOS (best for: Buyers who want the most disk per dollar in the opening months and will check the rate before month four): From $5/mo for VPS M+ (2 vCPU, 4 GB, 120 GB NVMe), for 3 months on a 1-year term
- CloudCone (best for: Developers who accept an annual bill and want 8 vCPUs on 4 GB of RAM): $7.66/mo for SSD VPS 4 (8 vCPU, 4 GB, 120 GB SSD), billed $92 per year in USD
- Hostinger (best for: A panel-first buyer who wants more RAM than this spec needs and will budget the renewal): KVM 2 from $8.99/mo promo (2 vCPU, 8 GB, 100 GB NVMe), renewing at $14.99/mo for 2 years

For a small API, a Postgres container and a reverse proxy, two vCPUs and 4 GB is a sensible starting size.

Every plan here is priced at that spec or at the smallest plan above it, which keeps the comparison between like boxes, not between a 512 MB row and a 4 GB one.

The wider VPS hosting for developers guide covers the larger plans and the operating differences between hosts.

This page answers a narrower question: which plan costs the least in USD for the same box, once the renewal, the address and the transfer pool are on the same sheet.

Toolradar data: Toolradar tracks 6 VPS and cloud infrastructure tools out of 11,051 in its directory, as of October 10, 2026.

That count describes the directory, not prices, so every figure below comes from the vendor's own pricing page, read the same day.

## Top Picks

Based on features, real-world fit, and value for money.

Best Cheap VPS Hosting in 2026: Priced at 2 vCPU and 4 GB: 8 tools compared, updated Oct 2026

Tool | Pricing | Best for |

OVHcloud | From $4.54/mo for VPS-1 (2 vCores, 4 GB, 40 GB NVMe); billing term not stated in words | Budget buyers who can pay up front and want the lowest monthly price at 4 GB |

IONOS | From $5/mo for VPS M+ (2 vCPU, 4 GB, 120 GB NVMe), for 3 months on a 1-year term | Buyers who want the most disk per dollar in the opening months and will check the rate before month four |

CloudCone | $7.66/mo for SSD VPS 4 (8 vCPU, 4 GB, 120 GB SSD), billed $92 per year in USD | Developers who accept an annual bill and want 8 vCPUs on 4 GB of RAM |

Hostinger | KVM 2 from $8.99/mo promo (2 vCPU, 8 GB, 100 GB NVMe), renewing at $14.99/mo for 2 years | A panel-first buyer who wants more RAM than this spec needs and will budget the renewal |

Akamai Cloud (Linode) | $24/mo for the 4 GB shared-CPU plan (2 vCPU, 80 GB SSD, 4 TB transfer) | Teams whose box serves files, where the per-GB overage is the number that matters |

[DigitalOcean](https://toolradar.com/tools/digitalocean) | $24/mo for the Basic 4 GiB Droplet (2 vCPU, 80 GiB SSD, 4,000 GiB transfer) | Developers who want IPv4 included, free inbound transfer and a backup line they can budget |

Amazon Lightsail | $24/mo for the General Purpose 4 GB bundle (2 vCPU, 80 GB SSD, 4 TB transfer, public IPv4) | Teams already on AWS that want one fixed bundle with an IPv4 address |

Kamatera | $39/mo for Type B Pro (2 vCPU, 4 GB, 20 GB NVMe, 5 TB traffic) | Buyers who want a configurable cloud server and can live with a small disk and a high price per GB |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### OVHcloud

Top Pick

Best for: Budget buyers who can pay up front and want the lowest monthly price at 4 GB

PricingFrom $4.54/mo for VPS-1 (2 vCores, 4 GB, 40 GB NVMe); billing term not stated in words

+Per GB of RAM the 4 GB plan works out to about $1.14, the lowest of the 4 GB plans here and the lowest monthly price on the list.

+Traffic is unlimited on a 500 Mbps public port, so a download-heavy box does not carry a metered line on the page.

+The ladder is public: VPS-2 is $8.50 a month for 4 vCores, 8 GB and 75 GB NVMe.

−The page does not spell out the billing term. Its configure link carries an upfront 12-month setting, so the checkout total is the figure to budget.

−40 GB of NVMe fills quickly once a container image, logs and a database share it.

Visit OVHcloud →

2

### IONOS

Best for: Buyers who want the most disk per dollar in the opening months and will check the rate before month four

PricingFrom $5/mo for VPS M+ (2 vCPU, 4 GB, 120 GB NVMe), for 3 months on a 1-year term

+The 120 GB of NVMe matches CloudCone's disk for less than CloudCone charges, so the opening rate buys the most storage per dollar here.

+Unlimited traffic and one IPv4 address are included, which keeps the address off a second line for a single box.

−The page gives a three-month rate on a one-year term and does not state the rate after that, so month four is the number to ask about before you sign.

−Each extra public address is a flat monthly fee, so a second IPv4 address doubles the opening rate.

Visit IONOS →

3

### CloudCone

Best for: Developers who accept an annual bill and want 8 vCPUs on 4 GB of RAM

Pricing$7.66/mo for SSD VPS 4 (8 vCPU, 4 GB, 120 GB SSD), billed $92 per year in USD

+10 TB of transfer a month and one IPv4 address sit inside the $7.66 figure, with no second line for either.

+The plan has 8 vCPUs against the 2 this spec asks for, so a parallel build or a burst job has headroom at the same RAM.

+The page states a 1 Gb/s network speed on all plans and USA datacenters, which suits a box that serves one US region.

−The page shows only a yearly bill and no monthly option, so a one-month test costs the full $92 at once.

−The page lists no backup or snapshot price, so the restore plan is a cost you have to size yourself.

Visit CloudCone →

4

### Hostinger

Best for: A panel-first buyer who wants more RAM than this spec needs and will budget the renewal

PricingKVM 2 from $8.99/mo promo (2 vCPU, 8 GB, 100 GB NVMe), renewing at $14.99/mo for 2 years

+8 GB of RAM at the promo rate is double the memory this spec asks for, which works out to about $1.12 per GB.

+The plan card lists 8 TB of bandwidth and free weekly backups, so a restore point is not a separate line.

+The FAQ says every KVM client gets a dedicated IP address, though the page does not name that address as IPv4.

−The renewal is $6 a month above the promo, a rise of about 67 percent on the promo rate.

−All plans are paid upfront, so the checkout total is larger than the monthly figure, and the page does not state how long the promo runs.

Visit Hostinger →

5

### Akamai Cloud (Linode)

Best for: Teams whose box serves files, where the per-GB overage is the number that matters

Pricing$24/mo for the 4 GB shared-CPU plan (2 vCPU, 80 GB SSD, 4 TB transfer)

+The overage is $0.005 per GB, so 100 GB past the pool costs $0.50, which keeps the worst month of a file host small.

+Every plan on the page is priced in US dollars, so the figure you read is the figure you are billed in.

+The shared-CPU family runs from 1 GB to 192 GB on one page, so the next size is easy to price before you need it.

−The page lists no price for an extra IPv4 address, so a second public address is a line to ask about before you assume it is free.

−Shared CPU shares cores with other tenants, so a CPU-heavy job should move to dedicated cores or a larger plan.

Visit Akamai Cloud (Linode) →

6

### DigitalOcean

Best for: Developers who want IPv4 included, free inbound transfer and a backup line they can budget

Pricing$24/mo for the Basic 4 GiB Droplet (2 vCPU, 80 GiB SSD, 4,000 GiB transfer)

+The 4 GiB Basic Droplet includes 4,000 GiB of outbound transfer, and inbound traffic is free, so uploads do not spend the pool.

+A public IPv4 is included at no extra cost on the bundled plans, so the address does not become a second line on the bill.

+Weekly backups are 20 percent of the Droplet price, which is $4.80 a month on this plan, and snapshots are $0.06 per GB a month.

−IPv4 is billed separately on v5 Droplets, so the plan family decides whether the address is free.

−The page does not say whether a powered-off Droplet keeps billing, so ask before you park a test machine.

Visit DigitalOcean →

7

### Amazon Lightsail

Best for: Teams already on AWS that want one fixed bundle with an IPv4 address

Pricing$24/mo for the General Purpose 4 GB bundle (2 vCPU, 80 GB SSD, 4 TB transfer, public IPv4)

+The public IPv4 bundle is one price, so the address, the disk and the transfer sit on one invoice row.

+The page prices the next step in the same table: $44 a month for 8 GB on 2 vCPUs, with 5 TB of transfer.

+The Compute Optimized 4 GB bundle is $42 a month on 160 GB of SSD, the step up when the workload needs faster cores.

−The IPv6-only bundle is a separate, lower-priced product, and it cannot serve an A record for a site that needs IPv4.

−Bundles are fixed, so a 2 vCPU box that needs 6 GB of RAM moves up to the 8 GB row rather than sitting between two sizes.

Visit Amazon Lightsail →

8

### Kamatera

Best for: Buyers who want a configurable cloud server and can live with a small disk and a high price per GB

Pricing$39/mo for Type B Pro (2 vCPU, 4 GB, 20 GB NVMe, 5 TB traffic)

+5 TB of traffic is included, more than the 4 TB pools on the Akamai and Lightsail 4 GB plans, and extra traffic is $0.01 per GB.

+Ubuntu Server 24.04 LTS (64-bit) is named as the image on this configuration, so the operating system is set before checkout.

−Per GB of RAM the 4 GB Pro plan costs $9.75, the highest per-GB figure in this ranking.

−The page does not say whether an IPv4 address is included, and it prints dollar figures without a currency code.

Visit Kamatera →

## What it is

A VPS is a virtual Linux server with its own public address, rented by the month and billed whether the CPU is busy or idle.

You get root access, which also means you patch it, open the firewall and run the stack you choose.

It is a different purchase from a serverless platform, which bills per request, and from a managed database.

If the box holds a database you will also operate, the Postgres hosting comparison prices the managed route next to it.

## Why it matters

The sticker hides three lines: the renewal, the IPv4 address and the transfer allowance. The four lowest stickers all carry a condition: OVHcloud's term sits in its configurator, IONOS's rate lasts three months, CloudCone bills yearly and Hostinger's renewal costs more.

The 4 GB plans from DigitalOcean, Akamai and Amazon Lightsail read as plain monthly list prices on their pages.

A cheap VPS also charges for things a managed platform bundles. DigitalOcean prices a backup as a share of the machine, and Akamai bills egress by the gigabyte past the pool.

A team comparing a server with a hosted platform should read the cheaper Vercel alternatives page, which prices hosting for a three-developer team.

## Key features to look for

A fixed spec
Every plan is priced at 2 vCPUs and 4 GB of RAM, or at the smallest plan above that spec. Where the smallest qualifying row is larger, such as Hostinger's 8 GB or CloudCone's 8 vCPUs, the extra capacity is noted rather than hidden.

The billing term
Some stickers hold for one month, some for a year paid upfront, and one for three months only. The term decides the first-year bill more than the monthly figure does.

The address line
A public IPv4 is included on some of these plans, charged separately on others, and not stated on some pages. Count it on the sheet before you compare the RAM.

The transfer pool
Included transfer runs from 4 TB to unlimited across these plans. Overage rates are printed on some pages and missing on others, so the worst month is not always visible on the card.

Backups and snapshots
A restore point is a separate line on some bills and free on others. DigitalOcean prices weekly backups as a share of the Droplet, while Hostinger includes weekly backups on its KVM plans.

Disk size
The disk runs from 20 GB at Kamatera to 120 GB at IONOS and CloudCone. A container image, logs and a database share that space, so the smallest disk is often the first limit a box reaches.

## Pricing

How we compared: each vendor's own USD pricing page, read on October 10, 2026. We ranked the smallest plan with at least 2 vCPUs and 4 GB of RAM by its monthly sticker, and we say where a page leaves out the term or the renewal.

Vendors that publish only euros, or without a confirmed current USD price for this spec, are left out.

The arithmetic: OVHcloud's sticker times 12 is $54.48 a year, CloudCone's yearly bill is $92, and Hostinger's promo comes to $107.88 for a year against $179.88 at renewal.

Linode, DigitalOcean and Lightsail at the same 4 GB tier come to $288 a year each, and Kamatera's $39 is $468 a year. Those are sticker totals, before backups, snapshots, the extra address or any overage.

Contabo, Netcup and Scaleway publish euro prices and no USD list, so their figures are not converted here.

Hetzner, RackNerd and Vultr are missing because no current USD price for this spec is confirmed.

The developer VPS guide carries earlier figures for Vultr and Hetzner, and the Hetzner Cloud review covers the product itself.

IONOS does not state what month four costs, and OVHcloud's term appears only in its configurator, so read the checkout total before you budget.

Toolradar keeps profiles for [OVHcloud](https://toolradar.com/tools/ovhcloud), [IONOS](https://toolradar.com/tools/ionos), [Hostinger](https://toolradar.com/tools/hostinger), [Linode](https://toolradar.com/tools/linode) and [DigitalOcean](https://toolradar.com/tools/digitalocean).

Vendor pages read for this page: [OVHcloud VPS](https://us.ovhcloud.com/vps/), [IONOS VPS](https://www.ionos.com/servers/vps), [CloudCone VPS](https://cloudcone.com/vps/), [Hostinger VPS](https://www.hostinger.com/vps-hosting), [Akamai Cloud shared CPU](https://www.akamai.com/cloud/pricing/cpu), [DigitalOcean Droplets](https://www.digitalocean.com/pricing/droplets), [Amazon Lightsail](https://aws.amazon.com/lightsail/pricing/) and [Kamatera](https://www.kamatera.com/pricing/). Kamatera prints dollar figures without a currency code, and we read them as USD.

Plan | Price | Best for |

OVHcloud VPS-1 | $4.54/mo | 2 vCores, 4 GB, 40 GB NVMe, unlimited traffic; term set in the configurator |

OVHcloud VPS-2 | $8.50/mo | 4 vCores, 8 GB, 75 GB NVMe, 1 Gbps public port |

IONOS VPS M+ opening rate | $5/mo | 2 vCPU, 4 GB, 120 GB NVMe; first 3 months of a 1-year term |

IONOS VPS M+ after month 3 | Not stated | The page does not state the rate after the opening period |

CloudCone SSD VPS 4 | $7.66/mo | 8 vCPU, 4 GB, 120 GB SSD, 10 TB a month; billed $92 a year |

Hostinger KVM 2 promo | $8.99/mo | 2 vCPU, 8 GB, 100 GB NVMe, 8 TB; all plans paid upfront |

Hostinger KVM 2 renewal | $14.99/mo | Renews for 2 years; free weekly backups on the plan card |

Akamai egress overage | $0.005/GB | Past the transfer pool on the shared-CPU plans, in US dollars |

DigitalOcean weekly backups | 20% of Droplet | Daily backups are 30 percent of the Droplet price |

DigitalOcean snapshots | $0.06/GB/mo | Stored snapshots, billed separately from backups |

Amazon Lightsail Compute Optimized 4 GB | $42/mo | 2 vCPU, 4 GB, 160 GB SSD, 5 TB transfer, public IPv4 |

Amazon Lightsail General Purpose 8 GB | $44/mo | 2 vCPU, 8 GB, 160 GB SSD, 5 TB transfer, public IPv4 |

Kamatera Type B Pro | $39/mo | 2 vCPU, 4 GB, 20 GB NVMe, 5 TB traffic |

Kamatera extra traffic | $0.01/GB | Beyond the 5 TB included with Type B Pro |

Mistakes to avoid

×Ordering the first price on the grid without reading the term. IONOS's opening rate holds for three months of a one-year term, and the page does not state what follows, so price month four before you sign.

×Budgeting Hostinger at the promo rate. The KVM 2 renewal is $14.99 a month for two years, $6 above the $8.99 opener, and that gap is $144 across 24 months.

×Forgetting the address. IONOS charges a flat monthly fee for each extra public address, which doubles the opening rate for a second one, and the Akamai page lists no IPv4 price at all.

×Treating backups as free. DigitalOcean's weekly backups are 20 percent of the Droplet, which is $4.80 a month on the 4 GiB plan, before snapshots at $0.06 per GB a month.

×Paying for a container host with a 20 GB disk. Kamatera's Pro plan has 20 GB of NVMe, so the disk is the first limit a Docker stack meets.

×Buying a test box on the yearly plan. CloudCone's page shows a yearly bill and no monthly option, so a box you abandon in month two still costs the full year.

Expert tips

→Price the year, not the month. Multiply the sticker by twelve, then add the backup percentage, the second address and an expected transfer overage. Once those lines sit on the same sheet, the cheapest plan can change.

→Size the stack before you shop. A 2 vCPU, 4 GB plan is the usual home for a small API, Postgres and a reverse proxy. If the database grows, the <a href="/reviews/best-postgres-hosting">Postgres hosting comparison</a> prices the managed option next to a box you patch yourself.

→Read the powered-off rule before you create a test machine. The DigitalOcean page does not say whether a stopped Droplet bills, so ask support before you park one. If the app serves files, read the egress line first, since that is where a quiet box turns into a large bill. The <a href="/reviews/digitalocean-review">DigitalOcean review</a> covers the Droplet itself.

## The bottom line

OVHcloud is the lowest sticker at this spec, and it is the buy if its upfront 12-month term and 40 GB of disk suit the project.

IONOS gives the most disk per dollar at its opening rate: a three-month rate on a one-year term, one IPv4 address and 120 GB of NVMe, with the rate after month three to confirm before you order.

CloudCone is for a team that wants 8 vCPUs and accepts an annual bill.

Hostinger is for a panel buyer who has budgeted the renewal and wants 8 GB of RAM.

DigitalOcean is the pick when you want the IPv4 address included and a weekly backup you can budget. Akamai is the pick when the overage per gigabyte is the number you worry about.

Kamatera only makes sense when a configurable server matters more than the price.

Cite this: Devshot, "Best Cheap VPS Hosting in 2026", October 2026.

## Frequently asked questions

What is the cheapest VPS with 2 vCPUs and 4 GB of RAM in 2026?

On the vendor pages read on October 10, 2026, OVHcloud's VPS-1 has the lowest sticker at $4.54 a month, with 40 GB of NVMe and unlimited traffic. Its page does not state the billing term in words, so read the checkout total. IONOS's VPS M+ is $5 a month for the first three months of a one-year term, and the page does not state the rate after that.

How much does a 4 GB VPS cost per month?

Across the seven 4 GB plans here, the sticker runs from OVHcloud's VPS-1 to $39 at Kamatera. Per GB of RAM that is about $1.14 to $9.75, and Hostinger's 8 GB plan works out to $1.12 on its promo, below that range. The lowest prices come with a term, a promo or an annual bill, so read the term before you read the number.

Is the Hostinger VPS price the same after the first term?

No. Hostinger's KVM 2 is $8.99 a month on the promo and $14.99 a month at renewal for two years, so the renewal is $6 a month more. The page says all plans are paid upfront, so the checkout total is larger than the monthly figure.

Does a cheap VPS include an IPv4 address?

Some do and some do not say. CloudCone includes one IPv4 address, IONOS includes one and charges a flat monthly fee for each extra, and DigitalOcean includes a public IPv4 at no extra cost on its bundled plans. Akamai's page lists no IPv4 price, and Kamatera's page does not say whether one is included.

Why are Vultr and Hetzner missing from the ranking?

No current USD price for the 2 vCPU and 4 GB spec is confirmed for Vultr, Hetzner or RackNerd, so no figure is quoted for them. Contabo, Netcup and Scaleway publish euro prices only. The developer VPS guide carries earlier figures for Vultr and Hetzner.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [DigitalOcean pricing](https://digitalocean.com/pricing)checked Sep 2026

Related guides

Self Hosting ToolsPostgres HostingServerless PlatformsDigitalocean ReviewHetzner ReviewCheaper Vercel Alternatives in 2026Object Storage Pricing Compared (2026)AI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Free AI Coding Tools in 2026

URL: https://devshot.co/reviews/best-free-ai-coding-tools
Type: review
Published: 2026-10-12
Updated: 2026-10-12
Summary: Free plans for AI coding tools compared on printed request caps, model access and the lowest paid step, checked on vendor pricing pages in October 2026.

Expert Guide

## The Best Free AI Coding Tools in 2026

Eight AI coding tools compared on their free allowance, the cap that stops a working day, and the cheapest paid step.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

Gemini CLI has the largest free allowance printed on a vendor page: 60 requests a minute and 1,000 a day on a personal Google account. GitHub Copilot Free gives 2,000 completions and 50 chat requests a month, and its Pro plan costs $10 a month. Zed keeps its personal plan free forever. Cline is free for individual developers, and Aider runs on your own key under an Apache 2.0 licence. With both open-source tools, the bill is your model provider's. Claude Code is not on Claude's free plan at all.

Key facts8 tools compared: Gemini CLI, GitHub Copilot Free, Zed (Personal), Amazon Q Developer Free…

- Updated: October 12, 2026
- Top pick: Gemini CLI (best for: Developers who want the largest printed daily allowance and an open-source terminal agent)
- Top pick price as of October 12, 2026: Gemini CLI: Free with a personal Google account: 60 requests a minute and 1,000 a day

- 8 tools compared: Gemini CLI, GitHub Copilot Free, Zed (Personal), Amazon Q Developer Free, Cline, Aider, Cursor Hobby, OpenAI Codex on ChatGPT
- GitHub Copilot Free (best for: Developers who want completions inside their editor on a fixed monthly quota): Free: 2,000 completions and 50 chat requests a month, with a paid Pro plan above it
- Zed (Personal) (best for: Developers who want an editor with a permanent free plan and their own API keys): Personal, free forever: 2,000 accepted edit predictions; Pro with $5 of tokens included
- Amazon Q Developer Free (best for: Developers working in AWS who want chat and agent help in the IDE and the CLI): Free: 50 agentic requests a month; Pro $19 per user a month

Most free AI coding tools are a meter, not a trial. The meter decides whether a free plan survives a working day, so this page ranks eight tools on the largest free allowance each vendor prints, then on whether the free plan stays free.

A tool that prints no number ranks lower, even when it is generous in practice.

Toolradar data: of the 156 AI code editors we evaluate, 133 (85%) have a free plan, as of October 2026 ([AI code editors ranking](https://toolradar.com/best/ai-code-editors)).

Free access is the norm in this category, so the useful comparison is the cap, not the sticker price.

Methodology: we read each vendor's own pricing page in October 2026 and recorded every free figure it prints, in US dollars where the page labels the currency. A cap the vendor does not print is marked as not printed. We do not estimate it.

## Top Picks

Based on features, real-world fit, and value for money.

Best Free AI Coding Tools in 2026: Free Limits Compared: 8 tools compared, updated Oct 2026

Tool | Pricing | Best for |

Gemini CLI | Free with a personal Google account: 60 requests a minute and 1,000 a day | Developers who want the largest printed daily allowance and an open-source terminal agent. |

GitHub Copilot Free | Free: 2,000 completions and 50 chat requests a month, with a paid Pro plan above it | Developers who want completions inside their editor on a fixed monthly quota. |

[Zed (Personal)](https://toolradar.com/tools/zed) | Personal, free forever: 2,000 accepted edit predictions; Pro with $5 of tokens included | Developers who want an editor with a permanent free plan and their own API keys. |

Amazon Q Developer Free | Free: 50 agentic requests a month; Pro $19 per user a month | Developers working in AWS who want chat and agent help in the IDE and the CLI. |

[Cline](https://toolradar.com/tools/cline) | Free for individual developers and open source; you pay your model provider for inference | Developers who want no subscription and will pay their model provider per token. |

[Aider](https://toolradar.com/tools/aider) | No tool price published; Apache 2.0 licence on GitHub; you bring your own provider key | Terminal users who want a command-line coding assistant and will supply their own provider key. |

Cursor Hobby | Hobby: free with no card, limited agent requests (no number printed); Pro $20 a month | Developers who want to try an AI editor before paying, and accept a cap that is not printed. |

OpenAI Codex on ChatGPT | Free plan at $0 (usage limit not printed); Go $8 a month; Plus $20 a month | Developers already on ChatGPT who want a coding agent on the same subscription. |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

1

### Gemini CLI

Top Pick

Best for: Developers who want the largest printed daily allowance and an open-source terminal agent.

PricingFree with a personal Google account: 60 requests a minute and 1,000 a day

+The only daily cap printed in this list, and the largest free allowance we verified

+Apache 2.0 licence, so the client itself costs nothing to run

−The page names the Gemini 3 family, not the individual models on the free tier

−No paid step is priced on its GitHub page, so the cost after the cap is unknown

Visit Gemini CLI →

2

### GitHub Copilot Free

Best for: Developers who want completions inside their editor on a fixed monthly quota.

PricingFree: 2,000 completions and 50 chat requests a month, with a paid Pro plan above it

+Both caps are printed in the plans FAQ, so the quota can be planned around

+The page says the free plan supports CLI and agent mode

−50 chat requests a month is a small quota for agent work

−The page prints no separate number for premium or agent requests on the free plan

Visit GitHub Copilot Free →

3

### Zed (Personal)

Best for: Developers who want an editor with a permanent free plan and their own API keys.

PricingPersonal, free forever: 2,000 accepted edit predictions; Pro with $5 of tokens included

+The page states the personal plan is free forever, with no trial clock

+Unlimited use with your own API keys or external agents on the free plan

−2,000 accepted edit predictions with no time period stated, so the cap is hard to size

−Pro bills usage beyond the included $5 at list price plus 10%, so heavy use costs more than the sticker price

Visit Zed (Personal) →

4

### Amazon Q Developer Free

Best for: Developers working in AWS who want chat and agent help in the IDE and the CLI.

PricingFree: 50 agentic requests a month; Pro $19 per user a month

+The free pool covers agentic coding in the IDE and the CLI, not only chat

+Pro is priced per user, so the cost scales with headcount at $19 a user a month

−Pro's agentic requests are listed as included with limits, with no number printed

−No completion cap is printed for the free tier, so heavy autocomplete use is hard to size

Visit Amazon Q Developer Free →

5

### Cline

Best for: Developers who want no subscription and will pay their model provider per token.

PricingFree for individual developers and open source; you pay your model provider for inference

+The page describes the tool as free for individual developers, with no printed usage cap

+You choose the model, so the cost follows the model you pick

−The page prints no token prices, so the bill cannot be estimated from it

−Enterprise is custom-priced, so the page prints no price for it

Visit Cline →

6

### Aider

Best for: Terminal users who want a command-line coding assistant and will supply their own provider key.

PricingNo tool price published; Apache 2.0 licence on GitHub; you bring your own provider key

+The repository is Apache 2.0 licensed, so the licence terms are published

+No seat price: the cost is what your provider bills for the tokens you send

−The project page prints no price for the tool and no token price for any model

−You set up and monitor the provider key and its billing yourself

Visit Aider →

7

### Cursor Hobby

Best for: Developers who want to try an AI editor before paying, and accept a cap that is not printed.

PricingHobby: free with no card, limited agent requests (no number printed); Pro $20 a month

+The Hobby plan needs no credit card, as the pricing page states

+Access to Composer is listed on the Hobby plan

−Agent requests are capped with no number printed, so the cap cannot be sized

−Teams is $40 per user a month, so a team's bill grows with every seat

Visit Cursor Hobby →

8

### OpenAI Codex on ChatGPT

Best for: Developers already on ChatGPT who want a coding agent on the same subscription.

PricingFree plan at $0 (usage limit not printed); Go $8 a month; Plus $20 a month

+The lowest printed paid entry in this list is the Go plan

+The page describes the Free plan as a way to explore Codex on quick coding tasks

−No usage limit is printed for Free or Go, so their allowance cannot be compared with the others

−The pricing page's usage table covers Plus, but not Go, so Go is the step a buyer cannot size

Visit OpenAI Codex on ChatGPT →

## What it is

A free AI coding tool is one of two things. It is either a plan with a published quota (completions, chat requests, agent requests or tokens), or an open-source client that you run with your own model key. The first kind is capped by the vendor.

The second is capped by your provider's bill. For the broader ranking of assistants, including the paid ones, see our best AI coding assistants guide.

## Why it matters

Whether a free plan works for a developer depends on its cap. The two kinds of number describe different days: 50 chat requests a month is a small quota for agent work, while 1,000 requests a day is a large one. Only Gemini CLI, GitHub Copilot and Amazon Q Developer print a time period with their figure.

Zed's edit predictions have none, so that cap is harder to size. Claude Code fails the free test. It is not included on Claude's Free plan, and our Claude Code review covers what the paid plans include.

## Key features to look for

Printed capEssential
The number the vendor gives for the free plan, per minute, per day or per month. Gemini CLI prints a per-minute and a per-day cap, GitHub Copilot Free prints monthly completions and chat, and Amazon Q Developer prints a monthly count of agentic requests. A tool with no printed number ranks lower.

Permanent or timedEssential
Whether the free plan stays free. Zed states that its personal plan is free forever. Cursor's Hobby plan needs no credit card, but no end date is printed for it. Cursor prints no number for its agent cap, so the cap cannot be planned around from the page.

Your own key
Open-source clients such as Cline and Aider run on the provider key you supply, so the bill is the provider's token charge. Neither project page prints a token price for the tool, so that bill cannot be worked out from the vendor's own page.

Lowest paid step
The cheapest published subscription that lifts the free cap. The lowest printed entry is Codex Go, but its usage limit is not printed. Copilot Pro and Zed Pro are the lowest paid plans that print what they include.

## Pricing

Cost after the free cap, for one developer on a monthly plan over 12 months. Codex Go: $8 x 12 = $96 a year, but its usage limit is not printed, so the cap cannot be checked. Copilot Pro: $10 x 12 = $120 a year. Zed Pro: $10 x 12 = $120 a year, with $5 of tokens included and usage beyond billed at API list price plus 10%.

Amazon Q Developer Pro: $19 x 12 = $228 a year. Cursor Pro and Codex Plus: $20 x 12 = $240 a year each. Claude Pro: [$200 a year billed up front](https://claude.com/pricing), or $240 a year billed monthly, with Claude Code included.

Cline and Aider have no subscription, so their cost is the token bill from your provider.

Neither vendor page prices a token, so this page does not compute that bill. Where the order changes: the arithmetic is for one developer on a flat plan. Team plans reorder the list, with Cursor Teams at $40 per user a month and Zed Business at $30 per seat a month.

The Cursor pricing and Claude Code pricing guides cover the full plan ladders.

Plan | Price | Best for |

Gemini CLI, personal Google account | Free | 60 requests a minute and 1,000 a day; Gemini 3 models with a 1M-token context window |

GitHub Copilot Free | Free | 2,000 completions and 50 chat requests a month, including Copilot Edits |

GitHub Copilot Pro | $10/mo | The paid tier directly above Free on GitHub's plans page |

Zed Personal | Free forever | 2,000 accepted edit predictions; unlimited use with your own API keys or external agents |

Zed Pro | $10/mo | Includes $5 of tokens; usage beyond billed at API list price plus 10% |

Amazon Q Developer Free | Free | 50 agentic requests a month, covering Q&A chat and agentic coding in the IDE and CLI |

Amazon Q Developer Pro | $19/user/mo | Agentic requests included with limits the page does not print |

Cline | Free | Open source for individual developers; you pay your model provider for inference |

Aider | No tool price | Apache 2.0 licence on GitHub; you bring your own provider key |

Cursor Hobby | Free | Limited agent requests with no number printed; no credit card needed |

OpenAI Codex Go | $8/mo | Lightweight coding tasks; the usage limit is not printed on the pricing page |

OpenAI Codex Plus | $20/mo | The usage table on OpenAI's pricing page covers Plus |

Claude Pro | $20/mo | Claude Code included; not on the Free plan. $17/mo billed annually |

Mistakes to avoid

×Reading a daily cap as a monthly one. Gemini CLI prints 1,000 requests a day, while GitHub Copilot Free prints a monthly quota. Put every cap on one clock before you compare them.

×Assuming the free plan has no limit on agent mode. GitHub's plans page says the free plan supports CLI and agent mode but prints no number for either, so plan around the monthly completions and chat quota instead.

×Expecting Claude Code on the free tier. Claude's pricing page marks Claude Code as not included on the Free plan, and it shares the usage limits of whichever paid plan you choose.

×Forgetting the provider bill on open-source tools. Cline and Aider have no subscription, but every token is billed by the provider whose key you use, and neither project page prints a token price.

Expert tips

→Count one real task before you commit. Run a typical refactor, note how many requests it used, then compare that number with the printed cap.

→Pick the paid step before you need it. Decide the trigger in advance, such as the first week you hit the cap, rather than choosing a plan under pressure in the middle of a sprint.

→Decide which account the tool runs under. The Gemini CLI allowance applies to the personal Google account you sign in with, so set that account up before the first session.

→If you need API access rather than an editor tool, the <a href="/blog/free-llm-apis">free LLM APIs</a> guide compares the caps on free API routes.

## The bottom line

There is no single best free AI coding tool, only the one whose cap you hit last. For the largest printed daily allowance, Gemini CLI leads. For a fixed monthly quota inside your editor, GitHub Copilot Free is the clearest number on this list.

For the no-subscription route, Cline and Aider put the bill on your model provider, and that bill cannot be sized from the vendor's page. Claude Code is not on a free plan.

Read the full plan ladders for [Gemini CLI](https://toolradar.com/tools/gemini-cli) and [GitHub Copilot](https://toolradar.com/tools/github-copilot) on Toolradar, or compare the category in our best code editors guide.

Cite this: Devshot, "Best Free AI Coding Tools in 2026", checked October 2026.

## Frequently asked questions

What is the best free AI coding tool in 2026?

Gemini CLI has the largest printed free allowance we found: 60 requests a minute and 1,000 a day on a personal Google account, per its GitHub page checked in October 2026. For a fixed monthly quota inside your editor, GitHub Copilot Free prints 2,000 completions and 50 chat requests a month.

Is GitHub Copilot free?

Yes. The Free plan prints 2,000 completions and 50 chat requests a month, including Copilot Edits, on GitHub's plans page checked in October 2026. The Pro plan is the paid step up.

Is Claude Code free?

No. Claude Code is not included on Claude's Free plan, according to claude.com/pricing checked in October 2026. The Pro plan costs $200 a year billed up front, and Claude Code shares the usage limits of the plan you choose.

What is the cheapest paid step after a free AI coding tier?

Codex Go is the lowest printed paid entry, but OpenAI prints no usage limit for it on the pricing page. Copilot Pro and Zed Pro are the paid plans that print what they include. Cline and Aider have no subscription, so their cost is your provider's token bill.

Do free AI coding tools have daily limits?

Some do. Gemini CLI prints a daily cap of 1,000 requests. GitHub Copilot Free and Amazon Q Developer print monthly caps, and Cursor Hobby prints no number for its agent requests.

Is Cline free?

Cline is free for individual developers and open source, according to cline.bot/pricing checked in October 2026. You pay your model provider for inference, and the page prints no token prices.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Zed pricing](https://zed.dev/pricing)checked Sep 2026
- [Cline pricing](https://github.com/cline/cline)checked Sep 2026
- [Aider pricing](https://aider.chat)checked Sep 2026

Related guides

Ai Coding AssistantsFree Database HostingGithub Copilot ReviewCursor ReviewClaude Code pricingGitHub Copilot pricingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Transactional Email APIs in 2026

URL: https://devshot.co/reviews/best-transactional-email-apis
Type: review
Published: 2026-10-12
Updated: 2026-10-12
Summary: Seven transactional email APIs priced at 50,000 emails a month, with free-tier caps and dedicated IP costs, checked on vendor pricing pages on 10 October 2026.

Expert Guide

## The Best Transactional Email APIs in 2026

Seven transactional email APIs compared on free tiers, the monthly bill at 50,000 emails and the cost of a dedicated IP.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 7 tools compared

TL;DR

For 50,000 transactional emails a month, Amazon SES is the cheapest at about $5 on list rates, but it means running an AWS account and your own bounce handling. Among managed services that publish a 50,000-email price, SendGrid Essentials is lowest at $19.95 a month, Resend Pro and Mailtrap Basic tie at $20 a month, and Mailgun's Foundation plan costs $35.

For free sending, Mailtrap has the largest stated monthly allowance, 4,000 emails with a 150-a-day cap, and Brevo has the highest daily cap at 300 a day after account approval. SendGrid's current pricing page lists a 60-day trial and no permanent free plan, so budget it as a paid service from the start.

Key facts7 tools compared: Amazon SES, Mailtrap, Resend, Mailgun, Postmark, Brevo, SendGrid

- Updated: October 12, 2026
- Top pick: Amazon SES (best for: Engineering teams already on AWS that send high volume and can run their own bounce handling)
- Top pick price as of October 12, 2026: Amazon SES: $0.10 per 1,000 emails, about $5 at 50,000; no monthly plan; dedicated IP $24.95/mo

- 7 tools compared: Amazon SES, Mailtrap, Resend, Mailgun, Postmark, Brevo, SendGrid
- Mailtrap (best for: Teams that want the largest stated free monthly allowance and a flat monthly plan for 50,000 emails): Free 4,000 emails/mo (150/day); Basic $20/mo for 50,000 emails; Business from $85/mo
- Resend (best for: Small product teams that want a free monthly tier and a flat plan price at 50,000 emails): Free 3,000 emails/mo (100/day); Pro $20/mo for 50,000; extra $0.90 per 1,000
- Mailgun (best for: Developers who want a 30-day trial on a paid plan before committing to 50,000 emails a month): Free 100/day; Foundation $35/mo for 50,000 after a 30-day trial; extra from $1.30 per 1,000

A transactional email API sends the message your app fires on its own: a signup confirmation, a password reset, an invoice.

Every service on this list bills that sending, but the free tiers, daily caps and dedicated IP add-ons differ enough that the same 50,000 emails can cost from the cheapest option here to $87 a month, depending on the vendor and the plan you land on.

Toolradar data: the [transactional email guide on Toolradar](https://toolradar.com/best/transactional-email) evaluated 13 transactional email tools, updated October 2026, and 12 of the 13 have a free plan.

For the wider category, open [Amazon SES](https://toolradar.com/tools/amazon-ses), [Resend](https://toolradar.com/tools/resend), [Mailtrap](https://toolradar.com/tools/mailtrap), [Mailgun](https://toolradar.com/tools/mailgun), [Postmark](https://toolradar.com/tools/postmark), [Brevo](https://toolradar.com/tools/brevo) and [SendGrid](https://toolradar.com/tools/sendgrid) there, and use the prices below for the bill.

Disclosure: Dupple's own newsletters are sent on beehiiv, which is not on this list.

How we ranked: seven APIs, each priced on the vendor's own pricing page on 10 October 2026.

The order follows the monthly list cost at 50,000 emails, and ties go to the larger free allowance. A vendor whose page does not show a 50,000-email price sits last, and we do not estimate it. Where a plan shows only a base quota and a per-1,000 overage, we compute the 50,000-email figure from those published rates and say so.

## Top Picks

Based on features, real-world fit, and value for money.

Best Transactional Email APIs in 2026: Free Tiers and Cost per 50,000 Emails: 7 tools compared, updated Oct 2026

Tool | Pricing | Best for |

[Amazon SES](https://toolradar.com/tools/amazon-ses) | $0.10 per 1,000 emails, about $5 at 50,000; no monthly plan; dedicated IP $24.95/mo | Engineering teams already on AWS that send high volume and can run their own bounce handling |

Mailtrap | Free 4,000 emails/mo (150/day); Basic $20/mo for 50,000 emails; Business from $85/mo | Teams that want the largest stated free monthly allowance and a flat monthly plan for 50,000 emails |

Resend | Free 3,000 emails/mo (100/day); Pro $20/mo for 50,000; extra $0.90 per 1,000 | Small product teams that want a free monthly tier and a flat plan price at 50,000 emails |

[Mailgun](https://toolradar.com/tools/mailgun) | Free 100/day; Foundation $35/mo for 50,000 after a 30-day trial; extra from $1.30 per 1,000 | Developers who want a 30-day trial on a paid plan before committing to 50,000 emails a month |

[Postmark](https://toolradar.com/tools/postmark) | Free 100/mo; Basic $15 for 10,000, extra $1.80 per 1,000; dedicated IP from $50/mo | Teams sending under 10,000 emails a month who want the lowest paid base plan and can accept overage above it |

[Brevo](https://toolradar.com/tools/brevo) | Free 300/day after approval; Starter $9/mo from 5,000 emails; no 50,000 price shown | Teams that want campaign email, SMS and transactional sends from one account |

SendGrid | 60-day trial, 100 emails/day; Essentials $19.95/mo at 50,000 emails; Pro from $89.95/mo | Teams already on Twilio that can budget a paid plan from the first month |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Amazon SES

Top Pick

Best for: Engineering teams already on AWS that send high volume and can run their own bounce handling

Pricing$0.10 per 1,000 emails, about $5 at 50,000; no monthly plan; dedicated IP $24.95/mo

+Sending 50,000 emails costs about $5 before attachment data

+No monthly plan to outgrow: the bill follows the sends each month

+A standard dedicated IP is $24.95 a month, below the $50 Postmark starts at

−No fixed monthly email quota on the pricing page; new AWS accounts get up to $200 of Free Tier credits instead

−Attachments add $0.12 per GB of outbound data on top of the per-email rate

Visit Amazon SES →

2

### Mailtrap

Best for: Teams that want the largest stated free monthly allowance and a flat monthly plan for 50,000 emails

PricingFree 4,000 emails/mo (150/day); Basic $20/mo for 50,000 emails; Business from $85/mo

+The free plan allows 4,000 emails a month at up to 150 a day

+Basic lists unlimited daily sends at 50,000 emails a month

+Business plans from $85 a month for 100,000 emails include a dedicated IP

−The overage rates are not mapped to plans on the page, so a bill above the cap is hard to forecast

−The dedicated IP price on Business is not stated

Visit Mailtrap →

3

### Resend

Best for: Small product teams that want a free monthly tier and a flat plan price at 50,000 emails

PricingFree 3,000 emails/mo (100/day); Pro $20/mo for 50,000; extra $0.90 per 1,000

+Pro includes 50,000 emails a month with no daily limit listed

+The free plan allows 3,000 emails a month at up to 100 a day

+Scale is $90 a month for 100,000 emails, with the same $0.90 overage rate

−Self-serve plans are billed monthly only, with no annual discount on the page

−A dedicated IP costs $30 a month and needs the Scale plan for senders above 3,000 emails a day

Visit Resend →

4

### Mailgun

Best for: Developers who want a 30-day trial on a paid plan before committing to 50,000 emails a month

PricingFree 100/day; Foundation $35/mo for 50,000 after a 30-day trial; extra from $1.30 per 1,000

+Foundation covers 50,000 emails a month, after a 30-day trial

+Basic is the entry plan for 10,000 emails a month

+The free plan allows 100 emails a day

−A dedicated IP is $59 per IP a month, the highest published add-on price here

−Overage on Foundation is from $1.30 per 1,000 emails, so a month above 50,000 adds up fast

Visit Mailgun →

5

### Postmark

Best for: Teams sending under 10,000 emails a month who want the lowest paid base plan and can accept overage above it

PricingFree 100/mo; Basic $15 for 10,000, extra $1.80 per 1,000; dedicated IP from $50/mo

+Basic is $15 a month for 10,000 emails, and the free plan covers 100 emails a month

+Pro is $16.50 a month for 10,000 emails, with overage at $1.30 per 1,000

+Platform is $18 a month for 10,000 emails, with overage at $1.20 per 1,000, the lowest Postmark overage rate

−At 50,000 emails the computed cost is $66 on Platform and $87 on Basic, the highest computed figure on this list

−The page offers no annual billing, only monthly plans

Visit Postmark →

6

### Brevo

Best for: Teams that want campaign email, SMS and transactional sends from one account

PricingFree 300/day after approval; Starter $9/mo from 5,000 emails; no 50,000 price shown

+The free plan allows up to 300 emails a day once the account is approved

+Starter is $9 a month, or $96.96 a year billed annually

+Starter includes sending of transactional messages

−No 50,000-email price is shown: Starter starts at 5,000 emails and Professional at 150,000

−Professional is $499 a month, a large step up from Standard at $18

Visit Brevo →

7

### SendGrid

Best for: Teams already on Twilio that can budget a paid plan from the first month

Pricing60-day trial, 100 emails/day; Essentials $19.95/mo at 50,000 emails; Pro from $89.95/mo

+The trial allows 100 emails a day for 60 days

+Pro includes one dedicated IP, with more available as an add-on

+Both paid plans list a monthly starting price on the page

−No permanent free plan appears on the page, so the free sending ends after 60 days

−Each plan's email quota follows a volume selector that prices only fixed stops, such as 50,000 and 100,000 emails on Essentials

Visit SendGrid →

## What it is

A transactional email API is a send endpoint your code calls once per event. The price follows the volume you send, not the size of your subscriber list, and that is what separates it from a marketing platform.

Resend's marketing plans, for example, are priced by contact on a separate list, so a newsletter's bill follows the subscriber count rather than the send count.

The category splits into two billing shapes.

Per-message services such as [Amazon SES](https://toolradar.com/tools/amazon-ses) charge a flat rate per 1,000 emails with no monthly plan. Plan-based services sell a monthly bundle of emails and charge overage above it.

A bundle looks cheap until the month runs past it, which is why every option here is priced at the same volume.

A team building on a serverless platform or a backend-as-a-service such as Supabase sees this bill grow with signups, because each signup triggers a send.

## Why it matters

At 50,000 emails a month, the spread is wide. Amazon SES is the lowest at list rates, SendGrid Essentials comes next at $19.95, Mailtrap Basic and Resend Pro follow at the same $20, Mailgun's Foundation plan follows, and Postmark costs the most of the plans with a 50,000 figure once overage is added.

The gap is the monthly sending bill, not the feature list.

Free tiers differ in shape too. Mailtrap's 4,000 emails a month is the largest stated monthly allowance here, while Brevo's 300 a day is the highest daily cap. The daily cap matters when a launch or a burst of password resets sends most of a month in one afternoon.

The cloud free tiers comparison covers the compute side of the same launch budget.

Dedicated IPs add a second line. Resend's add-on is $30 a month, Mailgun lists $59 per IP, Postmark starts at $50 per IP, and SES's is the lowest of the four.

Resend reserves its add-on for senders above 3,000 emails a day on the Scale plan, so most small senders never see that price on the way in.

## Key features to look for

The billing unitEssential
Amazon SES bills each message, while the other services sell a monthly bundle. Price the bundle at your real monthly volume, because the overage rates that are published run from $0.90 to $1.80 per 1,000 emails.

The daily cap on free sendingEssential
Resend and Mailgun cap free sending at 100 a day, Mailtrap at 150 and Brevo at 300 once the account is approved. SendGrid's trial allows 100 a day for 60 days. A cap of 100 a day can stop a launch even when the monthly total looks small.

Trial versus permanent free planEssential
SendGrid's current pricing page lists a 60-day trial and no permanent free plan. Postmark's free plan is 100 emails a month. Mailgun's Foundation plan carries a 30-day trial before its $35 price starts.

Dedicated IP
A per-IP add-on, priced from $24.95 a month on SES to $59 on Mailgun. Resend's $30 add-on is reserved for senders above 3,000 emails a day on Scale, and SendGrid includes one IP on its Pro plan.

Annual billing
Brevo's yearly price on Starter is about 10% below its monthly price. Resend's self-serve plans are billed monthly only, and Postmark lists no annual plans on its pricing page.

## Pricing

Start with the flat per-message rate, because it converts cleanly to a monthly bill. Amazon SES charges $0.10 per 1,000 emails, so 50,000 sends cost the least of any option here before attachment data.

Plan-based services bill a monthly base, and the cost at 50,000 depends on where the plan ends.

Resend Pro and Mailtrap Basic both include 50,000 emails at the same monthly price, and Mailgun Foundation covers that volume after a 30-day trial.

Postmark plans include 10,000 emails, so 50,000 comes to $66 to $87 once overage is added, computed from its published per-1,000 rates.

Brevo shows no 50,000-email price, so we leave that figure out rather than estimate it.

SendGrid's volume selector does show one: $19.95 a month for 50,000 emails. Brevo's entry plan starts at $9 a month for Starter, and that entry price is the comparison we can make for Brevo.

Plan | Price | Best for |

Amazon SES | $0.10 per 1,000 emails | About $5 at 50,000 emails; attachment data adds $0.12 per GB |

Amazon SES dedicated IP | $24.95/mo per IP | Standard IP on the account, billed monthly |

Resend Free | Free | 3,000 emails a month, up to 100 a day |

Resend Pro | $20/mo | 50,000 emails a month; extra emails at $0.90 per 1,000 |

Resend Scale | $90/mo | 100,000 emails a month; dedicated IP add-on for senders above 3,000 a day |

Mailtrap Free | Free | 4,000 emails a month, up to 150 a day |

Mailtrap Basic | $20/mo | 50,000 emails a month with unlimited daily sends |

Mailtrap Business | $85/mo | 100,000 emails a month; dedicated IP included, IP price not stated |

Mailgun Free | Free | 100 emails a day |

Mailgun Basic | $15/mo | 10,000 emails a month; extra emails from $1.80 per 1,000 |

Mailgun Foundation | $35/mo | 50,000 emails a month after a 30-day trial; extra from $1.30 per 1,000 |

Postmark Free | Free | 100 emails a month |

Postmark Basic | $15/mo | 10,000 emails a month; extra emails at $1.80 per 1,000 |

Postmark Platform | $18/mo | 10,000 emails a month; extra emails at $1.20 per 1,000 |

Brevo Free | Free | Up to 300 emails a day once the account is approved |

Brevo Starter | $9/mo | From 5,000 emails a month; transactional sends included |

Brevo Professional | $499/mo | From 150,000 emails a month; 10 seats included |

SendGrid Free trial | Free for 60 days | 100 emails a day; no permanent free plan on the page |

SendGrid Essentials | From $19.95/mo | 50,000 emails a month at $19.95 and 100,000 at $34.95; overage rate not stated |

SendGrid Pro | From $89.95/mo | One dedicated IP included; add-on price not stated |

Mistakes to avoid

×Comparing a free plan with a paid entry price as if they gave the same allowance. SendGrid's trial ends after 60 days, and Brevo's 300 a day starts only once the account is approved.

×Pricing Postmark at its base price for a 50,000-email month. The base covers 10,000 emails, so overage takes the bill to $66 or more.

×Assuming a dedicated IP comes free. Resend charges for one on Scale, Mailgun lists $59 per IP, and SES charges for one too, so the IP adds to the bill.

Expert tips

→Price the peak day as well as the month. Resend's free tier allows 100 emails a day and Mailtrap's 150, so a launch-day burst can hit the cap while the monthly total stays low. Mailtrap's Basic plan lists unlimited daily sends, and Resend Pro lists no daily limit. For alerts on a send spike, see the <a href="/reviews/best-monitoring-tools">monitoring tools guide</a>.

→Wire up bounce handling before trusting SES's low price. SES reports bounces and complaints through AWS notifications that your code has to receive and act on, so the lowest bill also carries the most setup work. The <a href="/reviews/best-error-tracking-tools">error tracking tools guide</a> covers the alerting side, and the <a href="/reviews/best-api-testing-tools">API testing tools guide</a> covers testing the send path before launch.

→Check the first invoice, not the headline. Mailgun's Foundation plan starts after a 30-day trial, Brevo's yearly price on Starter is about 10% below monthly, and Postmark's base covers only 10,000 emails, so compare each vendor at 50,000 emails.

## The bottom line

Default to Resend Pro or Mailtrap Basic when you want a managed service at the same flat monthly price for 50,000 emails.

Pick Amazon SES when the lowest bill matters more than the setup, and your team can run bounce handling on its own.

Choose Mailgun if its 30-day trial on Foundation fits your launch timing. Choose Postmark only if your sending stays close to 10,000 emails a month, where its base price is the real cost.

For the wider API stack around a send, the API gateways guide covers the next layer. Cite this: Devshot, "Best Transactional Email APIs in 2026", October 2026.

## Frequently asked questions

What is the cheapest transactional email API in 2026?

For 50,000 emails a month, Amazon SES has the lowest list price at about $5, from $0.10 per 1,000 emails, checked on AWS's pricing page on 10 October 2026. It needs an AWS account and your own bounce handling. Among managed services that publish a 50,000-email price, SendGrid Essentials is lowest at $19.95, and Resend Pro and Mailtrap Basic tie at $20.

Is there a free transactional email API?

Yes. Mailtrap's free plan gives 4,000 emails a month at up to 150 a day, Resend gives 3,000 a month at up to 100 a day, and Brevo gives up to 300 a day once the account is approved. Postmark's free plan is 100 emails a month. SendGrid's current page lists a 60-day trial with no permanent free plan.

How much does SendGrid cost in 2026?

SendGrid's Essentials plan starts at $19.95 a month and Pro at $89.95, both on the Twilio pricing page checked on 10 October 2026. The calculator prices Essentials at $19.95 a month for 50,000 emails and $34.95 for 100,000.

Is Resend cheaper than SendGrid?

Resend Pro has a published price of $20 for 50,000 emails a month, and SendGrid Essentials has a published price of $19.95 at the same volume, so Resend is not cheaper there. The gap is five cents a month.

Does Amazon SES have a free tier?

SES lists no fixed monthly email quota on its pricing page. New AWS accounts get up to $200 of AWS Free Tier credits. After that, sending is billed per 1,000 emails at the rate on AWS's page.

Do I need a dedicated IP for transactional email?

Resend's own page reserves its $30 a month dedicated IP for senders above 3,000 emails a day on Scale. Mailgun lists $59 per IP a month, SES charges $24.95, and Postmark starts at $50 per IP. Each price is on the vendor's pricing page checked on 10 October 2026.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Amazon SES pricing](https://aws.amazon.com/ses/pricing)checked Sep 2026
- [Mailgun pricing](https://mailgun.com/pricing)checked Sep 2026
- [Postmark pricing](https://postmarkapp.com/pricing)checked Sep 2026
- [Brevo pricing](https://www.brevo.com/pricing)checked Sep 2026

Related guides

Serverless PlatformsBackend As A ServiceError Tracking ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Postman Alternatives in 2026, Priced for Teams

URL: https://devshot.co/reviews/postman-alternatives
Type: review
Published: 2026-10-12
Updated: 2026-10-12
Summary: Postman alternatives priced for 3, 10 and 25 users, with USD team costs checked on each vendor's pricing page in October 2026.

Expert Guide

## The Best Postman Alternatives in 2026, Priced for Teams

Seven Postman alternatives, the seat cost for three, ten and twenty-five users where a vendor lists one, and each free tier's cap.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 7 tools compared

TL;DR

Bruno Pro is the lowest paid per-seat team plan here. It costs $6 per user a month billed annually, or $216 a year for three people. Postman Team costs $19 per user a month billed annually, which is $684 a year for the same three people. Postman's Free plan includes one user, so the saving starts with the first teammate you add.

Insomnia is the free start for a team of three, because Essentials covers Git Sync for up to three users at $0. Yaak suits developers who need a licence for work and want a local-first desktop app. Thunder Client fits people who live in VS Code, and Kreya is the choice when gRPC is the protocol you test most.

Stay on Postman when its hosted workspace, monitors and API catalog are part of the job. This list prices the client seat, not those extras.

Key facts7 tools compared: Bruno, Insomnia, Yaak, Thunder Client, Kreya, HTTPie, Hurl

- Updated: October 12, 2026
- Top pick: Bruno (best for: Teams that want collections in Git as plain files and a low seat bill)
- Top pick price as of October 12, 2026: Bruno: Open Source free; Pro $6/user/mo billed annually; Ultimate $11/user/mo billed annually

- 7 tools compared: Bruno, Insomnia, Yaak, Thunder Client, Kreya, HTTPie, Hurl
- Insomnia (best for: Small teams that want a free start, then Git or cloud sync when the team grows): Essentials free (Git Sync for up to 3 users); Pro $12/user/mo; Enterprise $45/user/mo
- Yaak (best for: Developers who want a local-first desktop client with licences they can transfer): Free for personal use; Individual $79/yr; Business $149/user/yr, billed annually
- Thunder Client (best for: Developers who test APIs inside VS Code or JetBrains without leaving the editor): Free in VS Code; Starter $3, Business $7, Enterprise $16 per user/mo, billed annually

Postman's Free plan includes one user, so the first teammate you add is the point where the bill starts. The decision is a seat decision. The seat price, the billing term and the free-tier cap matter more than the feature list.

This page prices five clients for three, ten and twenty-five people, using each vendor's own pricing page, and lists two command-line tools without a seat figure.

Toolradar data: Of the 479 API tools we track as of 10 October 2026, 74% offer a free, freemium or open-source plan.

The median cheapest paid plan across the 123 tools that publish a price is $29 a month. Postman has a free plan, so the switch is about seats, not the entry fee. Our [API tools ranking](https://toolradar.com/best/api-tools) covers the wider field, and [Postman has its own Toolradar profile](https://toolradar.com/tools/postman).

## Top Picks

Based on features, real-world fit, and value for money.

Best Postman Alternatives in 2026: Cost for 3, 10 and 25 Users: 7 tools compared, updated Oct 2026

Tool | Pricing | Best for |

[Bruno](https://toolradar.com/tools/bruno) | Open Source free; Pro $6/user/mo billed annually; Ultimate $11/user/mo billed annually | Teams that want collections in Git as plain files and a low seat bill |

[Insomnia](https://toolradar.com/tools/insomnia) | Essentials free (Git Sync for up to 3 users); Pro $12/user/mo; Enterprise $45/user/mo | Small teams that want a free start, then Git or cloud sync when the team grows |

Yaak | Free for personal use; Individual $79/yr; Business $149/user/yr, billed annually | Developers who want a local-first desktop client with licences they can transfer |

Thunder Client | Free in VS Code; Starter $3, Business $7, Enterprise $16 per user/mo, billed annually | Developers who test APIs inside VS Code or JetBrains without leaving the editor |

Kreya | Free forever; Pro $5/mo or $50/yr for one person; Enterprise for companies | Engineers who test gRPC services, where REST-only clients fall short |

HTTPie | Open-source terminal client; no seat price listed for the desktop app | Engineers who want one open-source terminal client for quick requests |

Hurl | Command-line tool with no seat price listed | Teams that keep API checks in CI and want them readable as text |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 7 do not publish a comparable monthly price and are left out rather than estimated.

1

### Bruno

Top Pick

Best for: Teams that want collections in Git as plain files and a low seat bill

PricingOpen Source free; Pro $6/user/mo billed annually; Ultimate $11/user/mo billed annually

+Pro is the lowest paid team seat price in this list at the annual rate shown above

+Open Source is free with two workspaces, and Pro removes that cap for unlimited workspaces

−The free tier reaches public Git providers only, so a private repo needs Pro

−Bruno's 14-day trial covers Ultimate, not Pro, and needs no card

Visit Bruno →

2

### Insomnia

Best for: Small teams that want a free start, then Git or cloud sync when the team grows

PricingEssentials free (Git Sync for up to 3 users); Pro $12/user/mo; Enterprise $45/user/mo

+Essentials is free with Git Sync for up to three users, so a team of three pays no seat cost

+Pro adds unlimited users and role-based access control, with 10,000 mock requests a month

−Pro is per user, so ten people cost $1,440 a year on the monthly rate

−Enterprise is self-serve up to 50 users, and larger teams go through sales

Visit Insomnia →

3

### Yaak

Best for: Developers who want a local-first desktop client with licences they can transfer

PricingFree for personal use; Individual $79/yr; Business $149/user/yr, billed annually

+Business licences can be reassigned to other team members at any time

+Business adds SSO, user provisioning and priority email support

−Licences are billed annually only, so there is no monthly option

−Work use needs a licence even for one person, so the Individual plan is the floor

Visit Yaak →

4

### Thunder Client

Best for: Developers who test APIs inside VS Code or JetBrains without leaving the editor

PricingFree in VS Code; Starter $3, Business $7, Enterprise $16 per user/mo, billed annually

+Business is $7 per user a month billed annually, with centralized billing

+The free version covers individual testing in VS Code at no cost

−Starter is a single-user licence, so a team has to start at Business

−It is an editor extension, not a standalone app, so a tester who never opens the IDE will not use it

Visit Thunder Client →

5

### Kreya

Best for: Engineers who test gRPC services, where REST-only clients fall short

PricingFree forever; Pro $5/mo or $50/yr for one person; Enterprise for companies

+The free plan covers gRPC, REST and GraphQL at no cost

+Pro adds scripts, tests, snapshots and collections for one person

−Free request history keeps only the latest request

−Free support runs through GitHub on a best-effort basis

Visit Kreya →

6

### HTTPie

Best for: Engineers who want one open-source terminal client for quick requests

PricingOpen-source terminal client; no seat price listed for the desktop app

+The terminal client is open source, so it has no seat price to compare

+Installs through common package managers such as Homebrew, apt and pip

−A team that works in a graphical client will find the terminal version narrow

−The desktop and web app is not priced on this page, so the team cost table leaves it out

Visit HTTPie →

7

### Hurl

Best for: Teams that keep API checks in CI and want them readable as text

PricingCommand-line tool with no seat price listed

+Requests are written in a plain text format, so they can be reviewed like code

+It chains requests and captures values between steps

−No graphical client, so exploratory debugging and a shared request library are not its job

−With no seat price to compare, Hurl stays out of the team cost table

Visit Hurl →

## What it is

A Postman alternative is an API client that sends requests, saves collections and runs tests, chosen for a seat price, a storage model or a workflow that Postman does not fit. The seven clients below cover the same core job. They differ on who pays per seat, where collections live and what the free tier caps.

Toolradar also keeps profiles for [Bruno](https://toolradar.com/tools/bruno) and [Insomnia](https://toolradar.com/tools/insomnia). For the wider category, start with the API testing tools guide.

Postman's own plans set the benchmark.

Free includes one user and 50 AI credits a month. Solo is $9 a month billed annually for one user, and Team is a per-seat plan. Enterprise is quote-only, so this page does not price it.

## Why it matters

The seat count drives the invoice, so this page prices three scenarios: three, ten and twenty-five users, billed annually where the vendor offers that term, except Insomnia, which is priced at its default monthly rate. Postman Team comes to $684 a year for three users, $2,280 for ten and $5,700 for twenty-five.

Bruno Pro comes to $216, $720 and $1,800, which saves $468 at three users, $1,560 at ten and $3,900 at twenty-five.

Thunder Client Business is $7 per user a month billed annually, so $252, $840 and $2,100. Yaak Business is $149 per user a year, so $447, $1,490 and $3,725.

Bruno Ultimate is $11 per user a month billed annually, so $396, $1,320 and $3,300. Insomnia Pro is $12 per user a month on monthly billing, so $432, $1,440 and $3,600.

A team of three can start at no seat cost.

[Insomnia Essentials](https://insomnia.rest/pricing) covers Git Sync for up to three users, so that headcount is the one scenario with a free line. Past three users the seat price decides the order, and the cheapest paid team seat is Bruno Pro.

Docs-first teams should also read the API documentation tools guide, which prices the docs side separately.

## Key features to look for

Seat price versus client price
Postman Team, Bruno Pro, Bruno Ultimate, Insomnia Pro, Thunder Client Business and Yaak Business bill per user. Yaak Individual and Kreya Pro are single-person prices, so they do not scale to a team.

The free tier's user cap
Postman's Free plan includes one user. Insomnia's Essentials covers Git Sync for up to three users. Bruno's Open Source tier allows two workspaces and public Git providers only.

Where the collections live
Bruno offers Native Git Integration, and Hurl keeps requests in plain-text files. Insomnia offers local, Git and cloud storage. Postman's collections live in its workspaces, the model a team plans around when it switches.

The billing term
Postman Team is $19 per user billed annually and $23 billed monthly. Insomnia's annual toggle is 17% lower than monthly. Yaak licences are billed annually and cannot be paid monthly.

Private Git access
Bruno's Open Source tier supports public Git providers only. Bruno Pro adds private ones, so a team with private repos pays from the first seat.

Where the client runs
Thunder Client is an extension inside VS Code and JetBrains. Yaak and Kreya are desktop apps. HTTPie and Hurl are command-line tools, which suits scripted checks more than shared exploration.

## Pricing

Methodology: we checked each vendor's own pricing page in October 2026 and quote USD. Team cost is the seat price times users times twelve, at the annual rate where the vendor bills annually. It covers the licence only.

AI credits, monitors and support tiers are not added to any figure.

Postman's monthly rate is $23 per user, which is $4 more than the annual rate. That is 21% more on the same team, or $480 a year at ten users.

Solo covers one user only, so a team cannot use it.

Insomnia's cost uses its monthly rate, the default on its pricing page, and its annual toggle is 17% lower. Bruno's Pro and Ultimate rates are billed annually. Yaak bills annually only.

Kreya's Pro price is for one person, and its Enterprise plan for companies has no per-seat price, so Kreya stays out of the team table. Git-first teams can also check the Git clients guide, and VS Code users the code editors guide.

Plan | Price | Best for |

Postman Free | Free | One user, 50 AI credits a month, API client and core tools |

Postman Solo | $9/mo billed annually | One user, 400 AI credits a month, data-driven testing with exports |

Postman Team | $19/user/mo billed annually | Team collaboration, basic role-based access control, 400 AI credits per user |

Bruno Open Source | Free | Two workspaces, public Git providers, unlimited collection runner |

Bruno Pro | $6/user/mo billed annually | Unlimited workspaces, private Git providers, 48-hour support SLA |

Bruno Ultimate | $11/user/mo billed annually | Private Git, 24-hour support SLA, 14-day trial with no card |

Insomnia Essentials | Free | Git Sync for up to 3 users, unlimited collection runs, 1,000 mock requests a month |

Insomnia Pro | $12/user/mo | Unlimited users, role-based access control, 10,000 mock requests a month |

Insomnia Enterprise | $45/user/mo | SSO and SCIM, vault integrations, self-serve up to 50 users |

Yaak Individual | $79/yr | One person using Yaak for work, with a lifetime fallback licence |

Yaak Business | $149/user/yr | Transferable licences, SSO and user provisioning, priority email support |

Thunder Client Free | Free | VS Code extension with the free version limits |

Thunder Client Business | $7/user/mo billed annually | Team features, CLI and CI/CD, centralized billing with invoice payment |

Kreya Free | Free | gRPC, REST, GraphQL and WebSockets, with the latest request in history |

Kreya Pro | $5/mo or $50/yr | One individual, with scripts, tests, snapshots and collections |

Mistakes to avoid

×Comparing entry prices instead of seats: Postman's Free plan costs nothing for one user, so the first paid cost arrives when the second person joins.

×Choosing monthly billing for Postman Team adds $4 per user a month over the annual rate, which is $480 a year at ten users.

×Assuming Insomnia Essentials covers a team of ten, because its free Git Sync is capped at three users.

×Treating Yaak as free at work: a licence is required for work use, and Individual is $79 a year.

Expert tips

→Price three headcounts before you compare features, because the seat count decides the order more than any single feature.

→Keep collections in Git if your team already reviews code, because Bruno's plain files and Hurl's text requests diff cleanly in a pull request.

→Run the free tier with your real collection count for one sprint before paying, since the caps (one Postman user, three Insomnia Git Sync users, two Bruno workspaces) are what forces an upgrade.

## The bottom line

For a team of three or more, Bruno Pro is the lowest paid team seat price here, at the annual rate shown in the table.

Keep Postman when its hosted workspace, monitors or API catalog are part of the job, because this list prices the client seat only.

Insomnia Essentials suits a team of three that wants no seat cost, and Yaak Business suits a team that wants transferable licences and SSO.

Thunder Client fits developers who never leave VS Code. Kreya is the pick for gRPC work, Hurl for CI-only checks, and HTTPie for a terminal-first team.

For GraphQL endpoints, see the GraphQL tools guide, and for load tests the load testing guide.

Cite this: Devshot, "Best Postman Alternatives in 2026: Cost for 3, 10 and 25 Users", October 2026.

## Frequently asked questions

What is the best Postman alternative in 2026?

For a team of three or more, Bruno Pro is the lowest paid per-seat team plan at $6 per user a month billed annually. For a team of three that wants no seat cost, Insomnia Essentials covers Git Sync for up to three users at $0. Both prices were checked on each vendor's pricing page in October 2026.

How much does Postman cost for a team?

Postman Team is $19 per user a month billed annually, so three users cost $684 a year, ten cost $2,280 and twenty-five cost $5,700. Monthly billing is $23 per user. Postman's Free plan includes one user, so every teammate after the first is a paid seat.

Is there a free Postman alternative?

Yes. Bruno Open Source is free with two workspaces and public Git providers, Insomnia Essentials is free with Git Sync for up to three users, and Thunder Client's free version costs nothing inside VS Code. The Bruno and Insomnia free tiers have caps that force a paid plan once the team grows.

Is Bruno cheaper than Postman for a team?

Yes, on list prices. Bruno Pro is $216 a year for three users, against $684 for Postman Team. At twenty-five users the gap is $3,900 a year. The comparison covers the seat price only, not Postman's AI credits or monitors.

Can a team use Yaak for free?

Not at work. Yaak is free for personal use, but work use needs a licence. The Individual licence is $79 a year for one person, and Business is $149 per user a year, billed annually.

Is Insomnia free for a team of four or more?

Essentials stops at three Git Sync users. A team of four that needs Git Sync for everyone moves to Pro at $12 per user a month on monthly billing, which adds unlimited users and role-based access control. Enterprise is $45 per user a month, with SSO and SCIM, and is self-serve up to 50 users.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Bruno pricing](https://usebruno.com/pricing)checked Sep 2026
- [Insomnia pricing](https://insomnia.rest/pricing)checked Sep 2026

Related guides

Api Testing ToolsApi Documentation ToolsGraphql ToolsLoad Testing ToolsCi Cd ToolsGit ClientsCode EditorsSelf Hosting ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Sentry Review: Is It Worth It in 2026?

URL: https://devshot.co/reviews/sentry-review
Type: review
Published: 2026-10-12
Updated: 2026-10-12
Summary: Sentry review for 2026: what the free Developer plan covers, Team and Business prices, overage math at 50K and 500K errors, pros, cons and four alternatives.

Review

## Sentry Review: Is It Worth It in 2026?

What each Sentry plan really costs, where the free tier runs out, and how it compares with Honeybadger, Rollbar, Raygun and GlitchTip.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Short answer: Sentry is worth it for product teams that want errors, traces and replays in one account, and the free Developer plan covers 5,000 errors for one user. Team is $26 a month when billed annually, with 50,000 errors and unlimited users. Above that quota, Team bills per error in bands, so 500,000 errors a month comes to about $132. Once your volume passes the quota, compare it with GlitchTip and Honeybadger before you commit.

Key factsBest for: Product teams that want errors, traces and replays in one…

- Updated: October 12, 2026
- Best for: Product teams that want errors, traces and replays in one account
- Price as of October 12, 2026: Free Developer plan (5,000 errors, one user); Team $26/mo billed annually

- The broadest error-tracking bundle here, with unlimited users on Team and overage billed per error, at rates that fall as volume grows.
- Free plan: 5,000 errors, one user
- Team plan: 50,000 errors and unlimited users, billed annually
- Business plan: SAML, SCIM and advanced quota management
- At 500K errors a month: About five times the Team base fee

Pros

- The free Developer plan includes 5,000 errors and 50 session replays, enough to test a real app before paying
- Team includes unlimited users, so the bill grows with errors rather than with seats
- Tracing, logs, replays, uptime and cron monitoring sit in one account instead of separate tools

Cons

- Overage is billed per error in bands, so 500,000 errors costs about five times the Team base fee
- Seer, the AI debugging agent, needs its own subscription on Team and is not listed on Developer
- Developer keeps events for 30 days and allows one user, which suits solo work rather than a team

Sentry

Free plan5,000 errors, one user

Team plan50,000 errors and unlimited users, billed annually

Business planSAML, SCIM and advanced quota management

At 500K errors a monthAbout five times the Team base fee

Sentry is one of the error trackers developers reach for first, and its pricing is easy to misread. The free plan looks generous until you count how many errors a busy app sends, and the paid tiers look cheap until volume grows.

Toolradar data: 7 of the 10 debugging tools we rank have a free plan, out of 91 we evaluated in October 2026 ([Toolradar's debugging ranking](https://toolradar.com/best/debugging)).

## What is Sentry?

Sentry is an application monitoring platform built around error tracking. Its SDKs catch exceptions in web, mobile and backend code, and each error arrives with a stack trace, the release it came from and the context around it.

Performance tracing, logs, session replay, uptime checks, cron monitors and Seer, an AI debugging agent (see our AI debugging tools guide), sit alongside it. For most teams the error count is the line that moves the bill.

For the tool profile, see [Sentry on Toolradar](https://toolradar.com/tools/sentry).

## How Sentry works

You add the SDK for your language, and errors start arriving in a project. Sentry groups identical errors into one issue, so ten thousand copies of the same crash show up as one issue with a count. Issues themselves do not count toward the quota; the errors inside them do, and that is the detail most people miss when they estimate usage.

Alerts go to email on the free plan. The site lists GitHub, Slack and a Sentry MCP server among its integrations. If you need a dedicated log store, the log management tools guide compares the options.

## Sentry key features

Error monitoring and grouping
Stack traces, release tracking and issue grouping, so a spike after a deploy shows as one issue with a count instead of thousands of rows.

Tracing
Tracing includes 5 million spans on Developer, Team and Business, and UI profiling needs pay-as-you-go billing.

Session replay
50 replays on Developer and Team.

Logs and application metrics
5GB of each included on Developer, Team and Business, then $0.50 per extra GB on Team and Business.

Uptime and cron monitoring
One uptime monitor and one cron monitor included on Developer, Team and Business. Extra uptime alerts cost $1.00 each and extra cron monitors $0.78 each on Team and Business.

SSO and provisioning
SAML and SCIM provisioning on Business, with advanced quota management and up to a 90-day lookback plus sampled retention.

## Sentry pricing

Sentry publishes list prices for Developer, Team and Business, and Enterprise is a custom quote. Developer is free for one user. Team and Business are priced per month when billed annually, and each includes a pre-set monthly quota of errors.

Anything above the quota is billed at pay-as-you-go rates, and that is where the bill grows.

Prices checked October 2026 on [Sentry's pricing page](https://sentry.io/pricing/) and in its [pricing documentation](https://docs.sentry.io/pricing/).

Team's per-error rates fall as volume rises: $0.0003625 for errors from 50,000 to 100,000, $0.0002188 up to 500,000, and $0.0001875 up to 10 million.

At 100,000 errors a month, Team comes to about $44. At 500,000 errors a month, Team comes to about $132: the base fee, plus 50,000 errors at $0.0003625 ($18.13) and 400,000 errors at $0.0002188 ($87.52).

Business at the same volume is about $386, because its band rates are higher.

Sentry's pricing page lists the bands but does not say whether a band applies only to the errors inside it. We price each band separately, which is the higher reading.

If Sentry bills the whole month at a single band rate, the Team bill is lower. The rates are shown rounded, so treat these totals as estimates and check the pricing calculator before you budget.

Two other costs sit on the page.

Seer needs its own subscription on Team, and overage is capped by the pay-as-you-go budget you set. Sentry's documentation says data sent after the reserved volume and the budget run out is dropped and not charged.

A spike therefore loses events rather than running up the bill, and you should decide which trade you want before launch.

The ranking changes with volume. At 50,000 errors a month, Sentry Team's base fee is level with Honeybadger Team on monthly billing, and GlitchTip's Small plan covers 100,000 events for $15.

At 500,000 errors, GlitchTip's Medium plan covers the whole volume. Honeybadger Team comes to $161 with overage billing switched on, at $0.0003 for each notification above 50,000. Raygun's Business plan reserves 1,000,000 errors for $400 a month billed annually.

Sources: [Honeybadger plans](https://www.honeybadger.io/plans/), [GlitchTip pricing](https://glitchtip.com/pricing), [Raygun pricing](https://raygun.com/pricing) and [Rollbar pricing](https://rollbar.com/pricing/), all read October 10, 2026.

Toolradar also profiles [Honeybadger](https://toolradar.com/tools/honeybadger) and [GlitchTip](https://toolradar.com/tools/glitchtip). For the wider field, see the monitoring tools guide.

Plan | Price | Best for |

Sentry Developer | Free | 5,000 errors, one user, 50 session replays, 30-day lookback |

Sentry Team | $26/mo billed annually | 50,000 errors and unlimited users, pay-as-you-go above the quota, up to 90-day lookback |

Sentry Business | $80/mo billed annually | 50,000 errors, SAML and SCIM, advanced quota management, sampled retention |

Sentry Enterprise | Custom | Custom volume, retention and support terms |

Honeybadger Developer | Free | 5,000 errors a month, built for solo developers on side projects |

Honeybadger Team | $286/yr billed annually | 50,000 errors a month, unlimited users, 90-day error retention, overage billing optional |

Honeybadger Business | $880/yr billed annually | 50,000 errors a month, 180-day error retention, enterprise security and compliance |

Rollbar Free | Free | 5K occurrences and 1K session replays, real-time feed and alerts |

Rollbar Essentials | Usage slider | 4K credits, up to 90-day retention, fine-grained rate limiting |

Rollbar Advanced | Usage slider | 8K credits, up to 180-day retention, SCIM provisioning |

Raygun Basic | $40/mo billed annually | 100,000 reserved errors a month, on-demand at $0.001 per error above the reserve |

Raygun Team | $80/mo billed annually | 200,000 reserved errors a month, inbound filters, user roles and permissions |

Raygun Business | $400/mo billed annually | 1,000,000 reserved errors a month, SAML SSO |

GlitchTip Free | Free | 1,000 events a month, unlimited projects and team members |

GlitchTip Small | $15/mo | Up to 100,000 events a month, support access |

GlitchTip Medium | $50/mo | Up to 500,000 events a month, priority email and live chat support |

GlitchTip Large | $250/mo | Up to 3 million events a month, development support, BAA on request |

## Sentry pros and cons

### What we like

- The free Developer plan includes 5,000 errors and 50 session replays, enough to test a real app before paying
- Team includes unlimited users, so the bill grows with errors rather than with seats
- Tracing, logs, replays, uptime and cron monitoring sit in one account instead of separate tools
- Business adds SAML, SCIM provisioning and advanced quota management for larger teams

### What could be better

- Overage is billed per error in bands, so 500,000 errors costs about five times the Team base fee
- Seer, the AI debugging agent, needs its own subscription on Team and is not listed on Developer
- Developer keeps events for 30 days and allows one user, which suits solo work rather than a team
- Business costs $80 a month, about three times Team, and that step is what adds SSO

## Who Sentry is for

Sentry suits product teams with a live web or mobile app that need stack traces, release tracking and replays in one tool. The free Developer plan works for one person testing an app. Team is the realistic first paid plan for a small team, and it stays predictable while errors stay near 50,000 a month.

For context on the category, see our error tracking ranking and Datadog review, which covers the heavier observability option.

It is a weaker fit for teams sending hundreds of thousands of errors a month, where GlitchTip costs less than half of Sentry Team, and for small teams that need SSO, since SAML and SCIM start at Business.

## Best Sentry alternatives

If Sentry is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Sentry | Product teams that want errors, traces and replays in one account | Free Developer plan (5,000 errors, one user) | Visit → |

Honeybadger | Small teams that want errors, uptime and logs at a lower annual price than Sentry Team | Free Developer plan (5,000 errors a month) | Visit → |

Rollbar | Teams that want real-time error alerts and fine-grained rate limiting | Free plan at $0 (5K occurrences, 1K session replays) | Visit → |

Raygun | Mobile and web teams that want crash reporting alongside real user monitoring | Crash Reporting Basic $40/mo billed annually for 100K errors | Visit → |

GlitchTip | Budget teams and open-source projects, and anyone who wants to self-host | Free to 1,000 events a month | Visit → |

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

Honeybadger

Errors, uptime, logs and status pages in one plan, with unlimited users on Team and a flat per-notification overage.

Visit →

Rollbar

A real-time error feed with intelligent grouping on the free plan, and paid tiers that add retention, credits and SCIM.

Visit →

Raygun

Crash reporting, real user monitoring and APM, priced by reserved errors each month with on-demand usage above the reserve.

Visit →

GlitchTip

The lowest published prices at both 50,000 and 500,000 errors here, with unlimited team members on the free plan and a self-hosted option.

Visit →

## The bottom line

Sentry is worth paying for once a product has real users and needs errors with full context, provided the error count stays near the Team quota. Start on the free Developer plan to measure how many errors your app really sends, then move to Team once you know the volume.

Model the overage before you pass 50,000 errors, because the bands make 500,000 errors about five times the base fee.

Choose GlitchTip if the lowest bill at high volume matters most, and Honeybadger if annual billing at the base price matters more than per-error overage.

Methodology: prices read on the vendor pages linked on this page, checked October 2026.

Overage totals are computed band by band and rounded.

Cite this: Devshot, 'Sentry Review 2026: Pricing, Pros, Cons and Alternatives', October 2026.

## Frequently asked questions

Is Sentry worth it in 2026?

For most product teams, yes, provided the error count stays near the Team quota. The free Developer plan covers 5,000 errors for one user, and Team at $312 a year billed annually adds 50,000 errors, unlimited users and tracing in one account. Past the quota, per-error bands raise the bill quickly: 500,000 errors a month comes to about $132 on Team. Above that volume, price GlitchTip and Honeybadger first.

How much does Sentry cost?

Developer is free for one user and 5,000 errors. Team is $312 a year billed annually and includes 50,000 errors, unlimited users and up to a 90-day lookback. Business costs $80 a month billed annually and adds SAML, SCIM and advanced quota management. Enterprise is a custom quote. Overage on Team is billed per error in bands, starting at $0.0003625 per error between 50,000 and 100,000 errors.

Is there a free Sentry plan?

Yes. The Developer plan costs nothing and includes 5,000 errors, one user, 50 session replays and a 30-day lookback. Seer, the AI debugging agent, is not listed on Developer and is sold as a subscription on Team. The plan suits a side project or a solo app. Once a second person needs access, you need Team, which is where unlimited users start.

Is Sentry cheaper than Honeybadger?

It depends on volume. On the base price, Honeybadger's annual Team plan costs $286 a year against Sentry Team's $312 a year, and both include 50,000 errors. At 500,000 errors a month, Sentry Team comes to about $132 with band-by-band overage, while Honeybadger Team billed monthly with overage billing on comes to $161. Honeybadger is cheaper on the annual base, and Sentry is cheaper at 500,000 errors.

What is the cheapest alternative to Sentry?

GlitchTip has the lowest published prices in this comparison. Small is $15 a month for up to 100,000 events, and Medium is $50 a month for up to 500,000. Its events include uptime checks and performance transactions, so the counts are not like-for-like with Sentry's errors. Rollbar sizes its paid plans with a usage slider, so run its calculator at your volume before you decide.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Sentry pricing](https://sentry.io/pricing)checked Sep 2026
- [Rollbar pricing](https://rollbar.com/pricing)checked Sep 2026

Related guides

Error Tracking ToolsMonitoring ToolsDatadog ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Free Database Hosting in 2026

URL: https://devshot.co/reviews/best-free-database-hosting
Type: review
Published: 2026-10-11
Updated: 2026-10-11
Summary: Free Postgres, MySQL, MongoDB and Redis hosting compared on storage caps, pause and expiry rules, backups and the cheapest paid upgrade…

Expert Guide

## The Best Free Database Hosting in 2026

Ten free database tiers compared on storage, compute, pause and expiry rules, backups and the first paid step.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 10 tools compared

TL;DR

Short answer: Neon is the best free Postgres for a side project. Its free plan is permanent, needs no credit card and gives 100 CU-hours and 1 GB per project. Supabase is the pick if you also need auth and file storage, but its free projects pause after one week of inactivity. For MySQL, TiDB Cloud Starter is free up to 25 GiB of row storage. MongoDB Atlas M0 is free forever at 512 MB, and Redis Cloud and Upstash cover Redis at 30 MB and 250 MB.

Key facts10 tools compared: Neon, Supabase, Aiven, Prisma Postgres, TiDB Cloud Starter, MongoDB Atlas…

- Updated: October 11, 2026
- Top pick: Neon (best for: Side projects and preview databases that need Postgres and no credit card)
- Top pick price as of October 11, 2026: Neon: Free plan: 100 CU-hours and 1 GB per project. Launch: $0.106 per CU-hour, $0.35 per GB-month

- 10 tools compared: Neon, Supabase, Aiven, Prisma Postgres, TiDB Cloud Starter, MongoDB Atlas (M0), Upstash Redis, Redis Cloud, Turso, Render Postgres
- Supabase (best for: Prototypes that want Postgres with auth and file storage in one project): Free plan: 500 MB database, 1 GB file storage, 2 active projects. Pro: from $25 a month
- Aiven (best for: Teams that want managed Postgres with no time limit and can live with 1 GB): PostgreSQL free plan: 1 GB RAM and 1 GB storage on one node. Hobbyist from $12 a month
- Prisma Postgres (best for: Prototypes and serverless apps that want Postgres without entering a card): Free plan: 200,000 operations a month, up to 50 databases. Starter: $10 a month with backups

Free database tiers differ in the parts that break a project: how much you can store, what pauses or deletes an idle database, and whether a backup exists. Toolradar data: among the 23 managed database tools we rank, 14 have a free plan.

A free tier is the norm in this category, not the exception ([managed databases ranking](https://toolradar.com/best/managed-databases), October 2026).

Methodology: we checked each free plan and the cheapest paid step on the vendor's own pricing page and docs in October 2026.

Vendors print prices with a dollar sign, and we read them as US dollars. Items are ranked on three things: how long the free tier runs without a deadline, how much data it holds, and what the first paid step costs. Where a vendor does not print a limit, the table says so rather than estimating one.

If you already pay for a database, our Postgres hosting comparison covers billing models instead. If you would rather run the database yourself, see our self-hosting tools roundup.

## Top Picks

Based on features, real-world fit, and value for money.

Best Free Database Hosting in 2026: Postgres, MySQL, MongoDB, Redis: 10 tools compared, updated Oct 2026

Tool | Pricing | Best for |

[Neon](https://toolradar.com/tools/neon) | Free plan: 100 CU-hours and 1 GB per project. Launch: $0.106 per CU-hour, $0.35 per GB-month | Side projects and preview databases that need Postgres and no credit card. |

[Supabase](https://toolradar.com/tools/supabase) | Free plan: 500 MB database, 1 GB file storage, 2 active projects. Pro: from $25 a month | Prototypes that want Postgres with auth and file storage in one project. |

Aiven | PostgreSQL free plan: 1 GB RAM and 1 GB storage on one node. Hobbyist from $12 a month | Teams that want managed Postgres with no time limit and can live with 1 GB. |

Prisma Postgres | Free plan: 200,000 operations a month, up to 50 databases. Starter: $10 a month with backups | Prototypes and serverless apps that want Postgres without entering a card. |

TiDB Cloud Starter | Free per organization: 25 GiB row and 25 GiB columnar storage, 250M request units a month | MySQL-style apps that need a large free allowance and can accept usage-based billing later. |

MongoDB Atlas (M0) | M0 free forever: 512 MB storage, shared RAM and vCPU. M10 from $56.94 a month | MongoDB prototypes and small apps that can live without backups and tolerate idle pauses. |

Upstash Redis | Free plan: 250 MB per database and 500K commands a month. Fixed 250 MB: $10 a month | Small apps that need Redis for caching, sessions or rate limits, with a monthly command budget. |

Redis Cloud | Always free: 30 MB single database. Essentials: from $5 a month | Tiny caches and session stores that fit in 30 MB and grow slowly. |

Turso | Free plan: 5 GB storage, 100 databases, 500M rows read a month. Developer: $4.99 a month | Apps that run many small databases, such as one per customer, on a SQLite-style engine. |

Render Postgres | Free: 1 GB storage, expires 30 days after creation. Basic: $6 a month, 256 MB RAM | A 30-day demo or hackathon database, with a plan to upgrade or export before it expires. |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

1

### Neon

Top Pick

Best for: Side projects and preview databases that need Postgres and no credit card.

PricingFree plan: 100 CU-hours and 1 GB per project. Launch: $0.106 per CU-hour, $0.35 per GB-month

+The free plan is permanent and needs no credit card, according to the pricing page

+Compute suspends after five minutes of inactivity, so a quiet database burns no CU-hours

+Launch has no monthly minimum, with storage at $0.35 per GB-month

−One gigabyte per project is small for anything beyond a prototype

−Free backups keep six hours of history and one manual snapshot

Visit Neon →

2

### Supabase

Best for: Prototypes that want Postgres with auth and file storage in one project.

PricingFree plan: 500 MB database, 1 GB file storage, 2 active projects. Pro: from $25 a month

+The free plan covers 50,000 monthly active users, far more than a prototype needs

+Two active free projects per organization, so a staging copy can run beside production

+Pro includes 8 GB of disk per project, then $0.125 per GB

−Free projects pause after one week of inactivity, so a quiet demo can be down when a visitor arrives

−The free plan includes no backups and caps egress at 5 GB

Visit Supabase →

3

### Aiven

Best for: Teams that want managed Postgres with no time limit and can live with 1 GB.

PricingPostgreSQL free plan: 1 GB RAM and 1 GB storage on one node. Hobbyist from $12 a month

+Free plans have no time limitation, according to the pricing page FAQ

+One account can cover Postgres, MySQL and Valkey, which suits a side project that uses more than one engine

+The paid PostgreSQL Hobbyist plan starts at $12 a month, or $0.02 an hour

−The free plan prints no limits for its MySQL or Valkey plans

−Aiven may shut down services it judges unused, so a quiet free service is at risk

Visit Aiven →

4

### Prisma Postgres

Best for: Prototypes and serverless apps that want Postgres without entering a card.

PricingFree plan: 200,000 operations a month, up to 50 databases. Starter: $10 a month with backups

+Up to 50 databases in one workspace on the free plan

+Idle databases cost nothing on the free plan

+Starter at $10 a month adds daily backups kept for seven days

−The free plan has no backups at all, so a free database needs its own export routine

−No storage cap is printed for the free plan, so check the limit before you load real data

Visit Prisma Postgres →

5

### TiDB Cloud Starter

Best for: MySQL-style apps that need a large free allowance and can accept usage-based billing later.

PricingFree per organization: 25 GiB row and 25 GiB columnar storage, 250M request units a month

+Overage is $0.20 per additional GiB of storage and $0.10 per additional million request units

+The pricing page says the Starter tier scales down to zero

+The allowance is per organization, so a team can share one quota across its projects

−New connections are refused once the free quota is used up, and the page gives no connection number

−Essential, the usage-based step up, has no starting price on the pricing page

Visit TiDB Cloud Starter →

6

### MongoDB Atlas (M0)

Best for: MongoDB prototypes and small apps that can live without backups and tolerate idle pauses.

PricingM0 free forever: 512 MB storage, shared RAM and vCPU. M10 from $56.94 a month

+Free forever on the pricing page, with 512 MB of storage that includes indexes

+M0 allows 500 connections per the docs, enough for a small app's connection pool

+The first dedicated tier, M10, is $0.08 an hour on the pricing page

−No backups on the free cluster, so you need your own mongodump routine

−Operations are capped at 100 per second, and a free cluster allows only one per project

Visit MongoDB Atlas (M0) →

7

### Upstash Redis

Best for: Small apps that need Redis for caching, sessions or rate limits, with a monthly command budget.

PricingFree plan: 250 MB per database and 500K commands a month. Fixed 250 MB: $10 a month

+Pay-as-you-go is $0.20 per 100K commands once you pass the free 500K

+Fixed plans make the bill flat: 1 GB for $20 a month, 5 GB for $100

+The free plan allows requests up to 10 MB

−The free plan caps each database at 250 MB, so a cache with long-lived data outgrows it quickly

−Per-command billing grows with traffic, so a busy cache can cost more than a fixed plan

Visit Upstash Redis →

8

### Redis Cloud

Best for: Tiny caches and session stores that fit in 30 MB and grow slowly.

PricingAlways free: 30 MB single database. Essentials: from $5 a month

+Essentials at $5 a month adds SAML SSO, role-based access control and encryption at rest

+Essentials lists uptime of up to 99.99% on its shared deployment

+The free database is listed as always free on the pricing page

−30 MB is small enough that most production caches outgrow it quickly

−Free support is community only, with a best-effort SLA

Visit Redis Cloud →

9

### Turso

Best for: Apps that run many small databases, such as one per customer, on a SQLite-style engine.

PricingFree plan: 5 GB storage, 100 databases, 500M rows read a month. Developer: $4.99 a month

+The free plan allows 100 databases and 500 million rows read a month

+Developer at $4.99 a month adds 9 GB of storage and unlimited databases

+The free plan allows 10 million rows written and 3 GB of syncs each month

−The free plan keeps only one day of point-in-time restore, against 10 days on Developer

−It is a SQLite-family engine, so a Postgres or MySQL app needs porting before it runs here

Visit Turso →

10

### Render Postgres

Best for: A 30-day demo or hackathon database, with a plan to upgrade or export before it expires.

PricingFree: 1 GB storage, expires 30 days after creation. Basic: $6 a month, 256 MB RAM

+Until it expires, the free instance has the same RAM and connection limits as the $6 plan

+The Basic plan at $6 a month is the cheapest paid step, with expandable storage at $0.30 per GB

+The free instance gives 1 GB of storage at no cost

−A free database expires 30 days after creation, and Render deletes it after a 14-day grace period

−Free databases have no backups and no connection pooling, according to Render's docs

Visit Render Postgres →

## What it is

A free database tier is a managed database with no monthly fee and fixed limits on storage, compute or requests. Most vendors cap usage rather than time, but some pause or expire idle projects, and that matters more than the cap for anything that does not run all day.

Four engines dominate the list: Postgres, MySQL, MongoDB and Redis. Turso is included for its storage allowance, although it describes itself as the next evolution of SQLite rather than as one of the four.

For the managed platforms that wrap a database in auth and APIs, see our backend-as-a-service roundup.

## Why it matters

The free tier decides what a project costs at its first real size, not at zero users. Three numbers do most of the work. The storage cap usually ends the free tier first: the Postgres allowances start at 500 MB on Supabase and 1 GB on Neon, Aiven and Render.

The pause or expiry rule decides whether a staging or demo database is still there on demo day. The first paid step sets the real bill: Neon charges $0.106 per CU-hour, Supabase Pro starts at $25 a month and MongoDB M10 starts at $56.94 a month.

Our Supabase review walks through the full paid ladder.

## Key features to look for

Storage capEssential
The limit that usually ends a free tier first. Supabase gives 500 MB of database, Neon, Aiven and Render give 1 GB, and TiDB Cloud Starter gives 25 GiB of row storage. Measure your dump size before you pick.

Compute or request allowanceEssential
Neon meters 100 CU-hours per project, Prisma Postgres counts 200,000 operations, TiDB Cloud Starter gives 250M request units and Upstash gives 500K commands. The units differ, so compare each one against your own query volume rather than against each other.

Pause and expiry ruleEssential
What happens when the project sits idle. Supabase pauses a free project after one week, MongoDB Atlas pauses an M0 after 30 days with zero connections, and Render expires a free database after 30 days. Neon's five-minute suspension stops compute, not storage.

Backups
Most free tiers have none. Supabase, Prisma Postgres, MongoDB Atlas M0 and Render include no backups on the free plan. Neon keeps a six-hour history with one manual snapshot, and Turso keeps one day of point-in-time restore.

Connection limits
Only a few vendors print a number: Render allows 100 connections and MongoDB Atlas M0 allows 500. Neon and TiDB Cloud Starter do not publish a cap, and TiDB refuses new connections once its usage quota is used up.

First paid step
Compare the cheapest paid plan that keeps your data, not the free one. Turso Developer is $4.99 a month, Render Basic is $6, Prisma Postgres Starter is $10, and Atlas M10 is $56.94.

## Pricing

Every plan in the table is priced on the vendor's own pricing page as read in October 2026. Two well-known names have no free tier to rank. PlanetScale lists no free plan, with its cheapest PS-5 at $5 a month in USD, and CockroachDB lists a 30-day trial with $400 in credit instead of a free plan.

Neither appears in the ranking.

Keeping a 5 GB database costs very different amounts. Turso's free plan holds 5 GB, so it costs nothing at that size. TiDB Cloud Starter holds 25 GiB free per organization while request units stay under 250M a month.

Neon's Launch plan adds $1.75 a month for 5 GB of storage at $0.35 per GB-month, plus compute. Supabase Pro costs $25 a month with 8 GB of disk included, and Upstash's fixed 5 GB plan costs $100 a month. Our Supabase pricing breakdown covers the paid tiers in more depth.

For the platform behind Render Postgres, see our Render review.

Plan | Price | Best for |

Neon Free | Free | 100 CU-hours and 1 GB storage per project, permanent plan |

Neon Launch | $0.106/CU-hour | Plus $0.35 per GB-month storage and no monthly minimum |

Supabase Free | Free | 500 MB database, 1 GB file storage, 50,000 monthly active users |

Supabase Pro | From $25/month | 8 GB disk per project, then $0.125 per GB |

Aiven PostgreSQL Free | Free | 1 GB RAM, 1 GB storage, one node, no time limit |

Aiven PostgreSQL Hobbyist | From $12/month | Entry paid PostgreSQL plan, or $0.02 an hour |

Prisma Postgres Free | Free | 200,000 operations a month, up to 50 databases, no backups |

Prisma Postgres Starter | $10/month | Daily backups kept seven days |

Render Free | Free | 1 GB storage and 256 MB RAM, expires 30 days after creation |

Render Basic | From $6/month | 256 MB RAM and 100 connections, storage at $0.30 per GB |

TiDB Cloud Starter | Free | 25 GiB row and 25 GiB columnar storage, 250M request units a month per organization |

TiDB Cloud Essential | Usage-based | Billed on storage and request capacity units, no starting price printed |

MongoDB Atlas M0 | Free | 512 MB storage, shared RAM and vCPU, 500 connections, no backups |

MongoDB Atlas M10 | From $56.94/month | Dedicated cluster, billed at $0.08 an hour |

Upstash Redis Free | Free | 250 MB per database and 500K commands a month |

Upstash Fixed 250 MB | $10/month | Flat fee, while pay-as-you-go is $0.20 per 100K commands |

Redis Cloud Free | Free | 30 MB single database on a shared deployment |

Redis Cloud Essentials | From $5/month | SAML SSO, role-based access, encryption at rest, up to 99.99% uptime |

Turso Free | Free | 5 GB storage, 100 databases, 500M rows read a month |

Turso Developer | $4.99/month | 9 GB storage, unlimited databases, 10-day point-in-time restore |

Mistakes to avoid

×Treating a 30-day expiry as a footnote. Render's free Postgres expires 30 days after creation, and Render deletes it after a 14-day grace period.

×Building on a free database with no backup. Supabase Free, Prisma Postgres Free, MongoDB M0 and Render Free all include no backups, so a manual export is your only restore point.

×Ignoring idle pauses before a demo. Supabase pauses free projects after one week of inactivity, and Atlas pauses a free cluster after 30 days with zero connections. Check the project the day before the meeting, not during it.

Expert tips

→Size against the storage cap first. Neon, Aiven and Render give 1 GB and Supabase gives 500 MB, so measure your real dump size before you pick, then leave room for indexes.

→Pick by the rule, not the price. A free database with a 30-day clock can cost more in migration time than the $6 Basic plan saves, so decide now when you will upgrade.

→Keep an export on a schedule. Run pg_dump for Postgres or mongodump for Atlas M0 on a cron, so a free tier is never your only copy.

## The bottom line

There is no single best free database, only the one whose limit you hit last. For Postgres that is Neon, because its free plan is permanent and its limits are the easiest to plan around. For MySQL, TiDB Cloud Starter's 25 GiB is the largest printed free allowance on this list.

For MongoDB, M0 is free forever, provided you can live without backups and with a 30-day idle pause. Render is the cheapest way to trial Postgres for a month and the easiest way to lose a database you forgot about.

Full profiles on Toolradar: [Neon](https://toolradar.com/tools/neon), [MongoDB Atlas](https://toolradar.com/tools/mongodb-atlas) and [Redis Cloud](https://toolradar.com/tools/redis-cloud).

For the paid side, read our Neon review, and once you have picked a host, our database clients roundup covers the tools for querying it.

Cite this: Devshot, "Best Free Database Hosting in 2026", October 2026.

## Frequently asked questions

What is the best free Postgres hosting in 2026?

Neon, for most side projects. Its free plan is permanent, needs no credit card and gives 100 CU-hours and 1 GB per project, as printed on the Neon pricing page in October 2026. Choose Supabase instead if you need auth and file storage, and accept that its free projects pause after one week of inactivity.

Is there a free MySQL hosting option?

Yes. TiDB Cloud Starter is free each month with 25 GiB of row storage, 25 GiB of columnar storage and 250M request units per organization. That comes from its pricing page, checked in October 2026. Aiven lists a free MySQL plan, but its pricing page does not print the limits. PlanetScale lists no free plan at all.

Does MongoDB Atlas have a free tier?

Yes, the M0 cluster, which the pricing page lists as free forever. It has 512 MB of storage, 500 connections and one free cluster per project. It has no backups, and Atlas pauses it after 30 days with zero connections. The first dedicated tier, M10, costs $0.08 an hour, and the pricing page lists it from $56.94 a month.

Is there a free Redis?

Two options. Redis Cloud's free database is always free at 30 MB, and Upstash gives 250 MB and 500K commands a month. After the free allowance, Upstash's pay-as-you-go rate is $0.20 per 100K commands, and Redis Cloud's Essentials plan starts at $5 a month.

Do free database tiers delete your data?

Some do. Render deletes a free Postgres database after its 30-day life and a 14-day grace period, and it offers no backups on the free instance. Supabase and MongoDB Atlas pause idle projects instead.

How much does it cost to keep a 5 GB database?

Between nothing and $100 a month, depending on the vendor. Turso's free plan holds 5 GB, and TiDB Cloud Starter holds 25 GiB free. Neon's Launch plan adds $1.75 a month for storage, plus compute. Supabase Pro costs $25 a month with 8 GB of disk, and Upstash's fixed 5 GB plan costs $100 a month.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Neon pricing](https://neon.tech/pricing)checked Sep 2026
- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026

Related guides

Postgres HostingBackend As A ServiceNeon ReviewSupabase ReviewRender ReviewSupabase pricingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best LLM APIs in 2026, Priced per Million Tokens

URL: https://devshot.co/reviews/best-llm-apis
Type: review
Published: 2026-10-11
Updated: 2026-10-11
Summary: LLM API prices per million tokens in 2026 for OpenAI, Anthropic, Google, DeepSeek, Mistral, Together AI and xAI, costed at one usage level.

Expert Guide

## The Best LLM APIs in 2026, Priced per Million Tokens

Seven LLM APIs compared on what one shared month of traffic costs, with batch, cache and peak-hour discounts, checked in October 2026.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 7 tools compared

TL;DR

Short answer: for 10 million input and 2 million output tokens a month, Gemini 2.5 Flash-Lite is the cheapest model priced on this page at $1.80, and GPT-6 Luna and Claude Haiku 5.5 come in at the same total as the cheapest models in the GPT-6 and Claude 5.5 families. Prices checked on vendor pricing pages in October 2026.

The same traffic costs $40.00 on GPT-6.1 Sol or Claude Sonnet 5.5. Keep small models on classification and routing, send offline work to batch at half price, and reserve the flagship for the steps that need it. Coding-tool subscriptions are a different bill, covered in our Claude Code pricing guide.

Key facts7 tools compared: Google Gemini API, OpenAI API, Anthropic API, DeepSeek API, Mistral AI API…

- Updated: October 11, 2026
- Top pick: Google Gemini API (best for: High-volume classification, extraction and tagging, where the lowest bill matters most)
- Top pick price as of October 11, 2026: Google Gemini API: Gemini 2.5 Flash-Lite $0.10 in, $0.40 out per 1M; Gemini 3.8 Flash $0.75 in, $3.75 out until Dec 31, 2026

- 7 tools compared: Google Gemini API, OpenAI API, Anthropic API, DeepSeek API, Mistral AI API, Together AI, xAI API
- OpenAI API (best for: Teams already on the OpenAI API that want a current small model with batch and caching built in): GPT-6 Luna $0.10 in, $0.50 out per 1M; GPT-6.1 Sol $2.00 in, $10.00 out (short prompts)
- Anthropic API (best for: Teams that route with a small model and call a larger one only for the hard requests): Claude Haiku 5.5 $0.10 in, $0.50 out per 1M (prompts to 100k tokens); Sonnet 5.5 $2.00 in, $10.00 out
- DeepSeek API (best for: Cost-sensitive apps that can move batch jobs and bulk work into off-peak UTC hours): DeepSeek V4.1 Flash $0.15 in, $0.60 out per 1M off-peak; double at peak (UTC weekday windows)

Most model comparisons list prices and stop there. A developer needs the monthly bill. Every vendor here bills input and output tokens at separate rates, and output costs three to six times input, so one shared usage level makes the rates comparable.

We price 10 million input and 2 million output tokens a month at standard rates, unless a row says otherwise.

Toolradar data: we evaluated 479 API tools in our API category in October 2026, and 131 of the 200 we rank (66%) list a free plan.

That category covers [API development, testing and documentation tools](https://toolradar.com/best/api-tools), not model APIs, so this ranking uses the per-token rate instead. For the wider API picks, see best API testing tools and best API gateways.

## Top Picks

Based on features, real-world fit, and value for money.

Best LLM APIs in 2026: Price per Million Tokens Compared: 7 tools compared, updated Oct 2026

Tool | Pricing | Best for |

Google Gemini API | Gemini 2.5 Flash-Lite $0.10 in, $0.40 out per 1M; Gemini 3.8 Flash $0.75 in, $3.75 out until Dec 31, 2026 | High-volume classification, extraction and tagging, where the lowest bill matters most |

OpenAI API | GPT-6 Luna $0.10 in, $0.50 out per 1M; GPT-6.1 Sol $2.00 in, $10.00 out (short prompts) | Teams already on the OpenAI API that want a current small model with batch and caching built in |

Anthropic API | Claude Haiku 5.5 $0.10 in, $0.50 out per 1M (prompts to 100k tokens); Sonnet 5.5 $2.00 in, $10.00 out | Teams that route with a small model and call a larger one only for the hard requests |

DeepSeek API | DeepSeek V4.1 Flash $0.15 in, $0.60 out per 1M off-peak; double at peak (UTC weekday windows) | Cost-sensitive apps that can move batch jobs and bulk work into off-peak UTC hours |

Mistral AI API | Mistral Small 2603 $0.15 in, $0.60 out per 1M; Mistral Medium 3.5 $1.50 in, $7.50 out per 1M | Teams that want low-cost small models, a 50% batch discount and cheap repeated prompts |

Together AI | gpt-oss-120B $0.15 in, $0.60 out per 1M on serverless inference, open-weight models from several labs | Teams that want open-weight models such as gpt-oss-120B on one bill and one API |

xAI API | Grok 4.7 $2.00 in, $6.00 out per 1M tokens; 500k-token context window | Agent builders who want a 500k-token window and the Web Search and X Search tools in one API |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

1

### Google Gemini API

Top Pick

Best for: High-volume classification, extraction and tagging, where the lowest bill matters most

PricingGemini 2.5 Flash-Lite $0.10 in, $0.40 out per 1M; Gemini 3.8 Flash $0.75 in, $3.75 out until Dec 31, 2026

+Gemini 2.5 Flash-Lite costs $1.80 at this usage level, the lowest total on this page

+Batch is half the standard rate across the paid tiers

+Gemini 3.1 Pro Preview's output rate on prompts up to 200k tokens is below Claude Opus 5.5's

−Gemini 2.5 Flash-Lite is a previous-generation model, so the cheapest line is not the newest one

−Gemini 3.8 Flash input and output double on January 1, 2027, which breaks any budget set in 2026

Visit Google Gemini API →

2

### OpenAI API

Best for: Teams already on the OpenAI API that want a current small model with batch and caching built in

PricingGPT-6 Luna $0.10 in, $0.50 out per 1M; GPT-6.1 Sol $2.00 in, $10.00 out (short prompts)

+GPT-6 Luna is the cheapest GPT-6 model and ties Claude Haiku 5.5 on total

+Batch halves both models, so the GPT-6.1 Sol scenario drops by half

+Cached input has its own price line, so the cache discount shows up separately on the bill

−Long-context prompts cost more, and GPT-6.1 Sol's long-context output rate is one and a half times its short-context rate

−GPT-6-astra is the most expensive OpenAI model on the page, at five times Sol's output rate

Visit OpenAI API →

3

### Anthropic API

Best for: Teams that route with a small model and call a larger one only for the hard requests

PricingClaude Haiku 5.5 $0.10 in, $0.50 out per 1M (prompts to 100k tokens); Sonnet 5.5 $2.00 in, $10.00 out

+Batch halves input and output for Sonnet 5.5 and the other listed models

+A cache hit costs 5% of the standard input price, which is $0.10 per 1M on Sonnet 5.5

+The price page lists cache writes separately, so you can cost a long-lived cache before you build it

−Haiku 5.5 changes price above 100,000 prompt tokens, to $0.50 in and $2.50 out

−Claude Fable 5.1 costs $10.00 in and $50.00 out, one of the most expensive Claude rates on the page

Visit Anthropic API →

4

### DeepSeek API

Best for: Cost-sensitive apps that can move batch jobs and bulk work into off-peak UTC hours

PricingDeepSeek V4.1 Flash $0.15 in, $0.60 out per 1M off-peak; double at peak (UTC weekday windows)

+Off-peak V4.1 Flash ties the Mistral Small and gpt-oss-120B entries for the scenario total

+A cache hit on the flash model is $0.003 per 1M off-peak, the lowest cached rate here

+One pricing table lists off-peak and peak columns, so the budget is easy to model by hour

−Peak pricing doubles every rate, so an unscheduled job that runs in a weekday peak costs twice as much

−V4 Pro costs about four times the Flash bill for the same month, even off-peak

Visit DeepSeek API →

5

### Mistral AI API

Best for: Teams that want low-cost small models, a 50% batch discount and cheap repeated prompts

PricingMistral Small 2603 $0.15 in, $0.60 out per 1M; Mistral Medium 3.5 $1.50 in, $7.50 out per 1M

+Mistral Small 2603 ties the DeepSeek and gpt-oss-120B entries for the scenario total

+The pricing FAQ states the batch and cache discounts in plain terms, which makes budgeting simple

+Mistral Medium 3.5 is priced below GPT-6.1 Sol and Claude Sonnet 5.5 at the scenario total

−Mistral Medium 3.5's output rate is about twelve times Small's, so the bill moves fast when you step up

−The pricing FAQ's worked example uses rates that differ from Mistral's model cards, so budget from the model card rates

Visit Mistral AI API →

6

### Together AI

Best for: Teams that want open-weight models such as gpt-oss-120B on one bill and one API

Pricinggpt-oss-120B $0.15 in, $0.60 out per 1M on serverless inference, open-weight models from several labs

+gpt-oss-120B ties the cheapest DeepSeek and Mistral Small entries for the scenario total

+The serverless table lists dozens of open-weight models with input and output prices in one place

+A Batch API price column sits beside the serverless prices on the same page

−Prices change by model and by host, so the same open-weight model can cost more on another provider

−Some listed promotions have end dates, so check the date on any promo line before you budget against it

Visit Together AI →

7

### xAI API

Best for: Agent builders who want a 500k-token window and the Web Search and X Search tools in one API

PricingGrok 4.7 $2.00 in, $6.00 out per 1M tokens; 500k-token context window

+Its output rate is below the output rate of GPT-6.1 Sol and Claude Sonnet 5.5

+Web Search and X Search are listed as tools in the same API docs, so real-time retrieval sits on one API

−Input is priced the same as GPT-6.1 Sol and Claude Sonnet 5.5, so the saving is on output only

−Batch is not available on every model, so check the model page before you plan bulk work

Visit xAI API →

## What it is

An LLM API bills by the token: a token can be a word, a number or a punctuation mark. Each request is charged for the input tokens you send and the output tokens the model returns, at separate rates.

## Why it matters

Output is the expensive side. On these price lists, output costs three to six times input. Claude Sonnet 5.5 and GPT-6.1 Sol both price output at five times input, and Google's Gemini 3.1 Pro Preview at six times. A support bot that writes long answers pays more for its replies than for its questions, so measure both sides before you pick a model.

The same two-rate shape applies to voice pricing, covered in best text-to-speech APIs.

Price the scenario, not the sticker. A cheap input rate and a cheap output rate look harmless on their own. At 2 million output tokens, the output line is half of the GPT-6 Luna bill.

## Key features to look for

Input and output ratesEssential
Every model here bills the two separately, and output costs three to six times input. A model that writes long answers can cost more than one that reads long prompts, so measure both.

Batch processing
OpenAI, Anthropic, Google and Mistral all list batch at half the standard rate. Use it for overnight jobs, evaluations and backfills, where a delay of hours costs nothing.

Prompt caching
Repeated prompt prefixes bill at a fraction of the input rate. Anthropic charges 5% of the input price for a cache hit, OpenAI charges 5% on GPT-6.1 Sol, and Mistral says cached input is up to 90% cheaper.

Time-of-day pricing
DeepSeek charges half its peak rate off-peak. Peak is 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, so a scheduled job outside those windows costs half as much per token.

Context length
Context windows differ: Mistral Small and Medium list 256k tokens, Grok 4.7 lists 500k and DeepSeek lists 1M. Long prompts often move a request to a higher price tier, so check the tier before you rely on the window.

## Pricing

Methodology: we read each vendor's own pricing page in October 2026 and took the entry-level current text model on each list, plus a flagship for context. For Together AI we priced gpt-oss-120B as the reference open-weight model, because its serverless list also carries cheaper small models.

Items are ordered by the entry model's cost at 10 million input and 2 million output tokens a month.

The scenario cost is the input volume times the input rate, plus the output volume times the output rate, with no caching or batch unless the row says so.

Google's [Gemini API](https://toolradar.com/tools/google-gemini) has the cheapest entry rate, from [Gemini 2.5 Flash-Lite](https://ai.google.dev/gemini-api/docs/pricing) on Google's price list.

Its newer Gemini 3.8 Flash is priced at one rate through December 31, 2026 and rises on January 1, 2027, so the same month of traffic costs $15.00 this year and $30.00 next year.

[OpenAI](https://toolradar.com/tools/openai-api) and [Anthropic](https://toolradar.com/tools/anthropic-api) price their current small models at the same scenario total, which is why they tie.

Claude Haiku 5.5 moves to a higher rate above 100,000 prompt tokens.

Gateways add their own markup on top of these rates; our OpenRouter alternatives page covers those fees.

[DeepSeek's price list](https://api-docs.deepseek.com/quick_start/pricing) sets V4.1 Flash at a lower off-peak rate and doubles it in the UTC peak windows.

The [Mistral Small model card](https://docs.mistral.ai/models/model-cards/mistral-small-4-0-26-03) gives its rates, and the [Mistral](https://toolradar.com/tools/mistral-ai) pricing FAQ confirms the 50% batch discount.

For coding-tool bills rather than API bills, see our Cursor pricing guide.

Plan | Price | Best for |

Gemini 2.5 Flash-Lite | $1.80 a month | Previous-generation model and the lowest total on the list |

GPT-6 Luna | $2.00 a month | Current small model; batch halves the rate |

Claude Haiku 5.5 | $2.00 a month | Current small model; prompts over 100k tokens cost more |

DeepSeek V4.1 Flash | $2.70 off-peak, $5.40 at peak | Peak hours double the rate |

gpt-oss-120B on Together AI | $2.70 a month | Open-weight model on serverless hosting |

Mistral Small 2603 | $2.70 a month | 256k context window; batch at half the rate |

Gemini 3.8 Flash | $15.00 to 2026, $30.00 from 2027 | Input and output rates double on January 1, 2027 |

GPT-6.1 Sol | $40.00 a month, $20.00 on batch | Flagship-class OpenAI model |

Claude Sonnet 5.5 | $40.00 a month, $20.00 on batch | Mid-tier Claude model; cache hits cost 5% of input |

Grok 4.7 | $32.00 a month | 500k-token context window |

Mistral Medium 3.5 | $30.00 a month | 256k context window |

DeepSeek V4 Pro | $10.56 off-peak, $21.12 at peak | The Pro tier of DeepSeek V4 |

Claude Opus 5.5 | $80.00 a month | For long-running agentic coding and knowledge work |

Claude Fable 5.1 | $200.00 a month | For demanding reasoning and long-horizon agentic work |

Mistakes to avoid

×Comparing per-token rates with monthly plan prices. Convert every line to the same unit before you put two numbers side by side.

×Budgeting DeepSeek at the off-peak rate. Peak is 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, and it doubles the bill for scheduled jobs that run in those windows.

×Missing the January 1, 2027 step on Gemini 3.8 Flash, where input and output both double. A budget signed off in 2026 misses it.

×Sending long documents without checking the tier. Claude Haiku 5.5 and Gemini 3.1 Pro both change price above a prompt-size threshold, and GPT-6.1 Sol lists a separate long-context rate.

×Paying for cache writes without repeat traffic. Anthropic bills a 1-hour cache write at twice the input rate, so caching pays only when the same prefix is read again and again.

Expert tips

→Price the scenario with your own token counts. Log input and output tokens for a week, then run them through each rate card, because the output share often decides the ranking.

→Send offline work to batch. OpenAI, Anthropic, Google and Mistral all list batch at half the standard rate, so evaluations, backfills and nightly enrichment cost half as much.

→Route by task. Use a small model for classification and extraction and escalate only the hard cases. At this usage level, a flagship on every call costs twenty times more than a small model.

→Re-check the price list before you commit. Gemini's January 2027 step-up and DeepSeek's peak windows show that the same model can cost a different amount on another date.

## The bottom line

Default to Gemini 2.5 Flash-Lite or GPT-6 Luna for high-volume jobs with short outputs, and send anything that can wait to batch. Step up to GPT-6.1 Sol or Claude Sonnet 5.5 only for the calls that need it, because the same traffic costs twenty times more.

If the calls run on a serverless platform, the hosting bill is separate; see best serverless platforms.

Watch the calendar. Gemini 3.8 Flash doubles on January 1, 2027, and DeepSeek charges double in its UTC peak windows.

For another per-unit cost comparison in the same format, read object storage pricing compared. Cite this: Devshot, "Best LLM APIs in 2026: Price per Million Tokens Compared", October 2026.

## Frequently asked questions

What is the cheapest LLM API in 2026?

For 10 million input and 2 million output tokens a month, Gemini 2.5 Flash-Lite is the cheapest model priced on this page at $1.80, checked on Google's pricing page in October 2026. GPT-6 Luna and Claude Haiku 5.5 come out at the same total and are the cheapest models in the GPT-6 and Claude 5.5 families.

How much does GPT-6.1 Sol cost per million tokens?

GPT-6.1 Sol lists $10.00 per 1M output tokens for short prompts, and half that on the batch tier, per OpenAI's pricing page in October 2026. Long prompts pay a higher input and output rate, so check the tier before you budget a long document.

Is the batch API really half price?

Yes, for the vendors that publish it. OpenAI, Anthropic and Google price batch at 50% of the standard rate on input and output, and Mistral states a 50% reduction in its pricing FAQ. DeepSeek's discount comes from off-peak hours instead.

Does prompt caching cut the bill?

On repeated prompt prefixes, yes. Anthropic charges 5% of the input rate for a cache hit, OpenAI charges 5% on GPT-6.1 Sol, and Mistral says cached input is up to 90% cheaper. Cache writes cost extra on Anthropic, so measure your repeat rate before you enable caching.

Is DeepSeek cheaper than OpenAI?

Not at this usage level. DeepSeek V4.1 Flash costs $2.70 off-peak and $5.40 at peak, which is more than GPT-6 Luna and Claude Haiku 5.5 at the same usage level. The answer depends on which models you compare and when your jobs run, so cost the scenario with your own token counts.

Related guides

Openrouter AlternativesApi GatewaysText To Speech ApisApi Testing ToolsServerless PlatformsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best OpenRouter Alternatives in 2026

URL: https://devshot.co/reviews/openrouter-alternatives
Type: review
Published: 2026-10-09
Updated: 2026-10-09
Summary: OpenRouter alternatives in 2026 for developers who want a lower fee, a self-hosted gateway or stronger controls, with USD prices from each vendor.

Expert Guide

## The Best OpenRouter Alternatives in 2026

Eight LLM gateways and model hosts compared on the fee they add to model tokens, the free tier and who runs the server, with prices checked in October 2026.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

[Vercel AI Gateway](https://toolradar.com/tools/vercel) is the closest hosted replacement for OpenRouter in 2026. It adds no markup and no platform fee on tokens, even with your own keys, while OpenRouter lists a 5.5% platform fee on its Standard plan. Use [LiteLLM](https://toolradar.com/tools/litellm) to self-host for $0, [Cloudflare AI Gateway](https://toolradar.com/tools/cloudflare) if the app already runs on Cloudflare, or [Portkey](https://toolradar.com/tools/portkey) from $49/mo for guardrails and logs. Prices were checked in October 2026.

Stay on [OpenRouter](https://toolradar.com/tools/openrouter) if one credit balance across 500+ models is worth the fee to you. Among the alternatives, only Requesty (a flat 5% markup) and Cloudflare (5% on credits bought through Unified Billing) add a percentage to what providers charge.

Key facts8 tools compared: Vercel AI Gateway, LiteLLM, Cloudflare AI Gateway, Portkey, Requesty, Kong…

- Updated: October 9, 2026
- Top pick: Vercel AI Gateway (best for: Hosted replacement for teams that want the lowest fee and a base URL swap)
- Top pick price as of October 9, 2026: Vercel AI Gateway: No markup or platform fee on tokens; $5/mo free credit, then pay-as-you-go credits

- 8 tools compared: Vercel AI Gateway, LiteLLM, Cloudflare AI Gateway, Portkey, Requesty, Kong AI Gateway, Together AI, Amazon Bedrock
- LiteLLM (best for: Teams that must keep prompts and keys on their own infrastructure): Open source $0 self-hosted; Enterprise is a sales quote sized to request volume, never per token
- Cloudflare AI Gateway (best for: Apps already running on Cloudflare that need caching and rate limits): Core features free on all plans; 5% fee only on credits bought via Unified Billing
- Portkey (best for: Production teams that want routing, guardrails and traces in one product): Free for prototyping; Production $49/mo with 100k recorded logs; Enterprise custom

A team that sends every model call through one aggregator eventually reads the fee line. At OpenRouter's Standard rate, $10,000 of monthly model spend carries $550 of fee, and Business costs more.

Whether that is fair depends on what you would otherwise run yourself, and often the better answer is a thinner gateway or none at all.

An OpenRouter alternative is a gateway or model host that puts many providers behind one OpenAI-compatible endpoint, with its own fee model: a markup, a flat subscription, or nothing when you bring provider keys.

We split the picks on that fee, because token prices are the same almost everywhere and the fee is what differs.

Devshot data: Toolradar's closest category, [AI observability](https://toolradar.com/best/ai-observability), evaluates 43 tools as of October 2026, and Portkey sits in its top 10.

Of the 8 alternatives below, 3 charge no platform fee on tokens when you bring your own keys (Vercel AI Gateway, LiteLLM and Cloudflare AI Gateway), and 1 adds a flat markup to every token (Requesty).

If you want a ranking that is not built around leaving OpenRouter, Dupple's [LLM gateways guide](https://dupple.com/learn/best-llm-gateways) covers the wider field.

REST gateways do a different job and live in the API gateways guide.

## Top Picks

Based on features, real-world fit, and value for money.

Best OpenRouter Alternatives in 2026 (LLM Gateways): 8 tools compared, updated Oct 2026

Tool | Pricing | Best for |

Vercel AI Gateway | No markup or platform fee on tokens; $5/mo free credit, then pay-as-you-go credits | Hosted replacement for teams that want the lowest fee and a base URL swap |

LiteLLM | Open source $0 self-hosted; Enterprise is a sales quote sized to request volume, never per token | Teams that must keep prompts and keys on their own infrastructure |

Cloudflare AI Gateway | Core features free on all plans; 5% fee only on credits bought via Unified Billing | Apps already running on Cloudflare that need caching and rate limits |

Portkey | Free for prototyping; Production $49/mo with 100k recorded logs; Enterprise custom | Production teams that want routing, guardrails and traces in one product |

Requesty | 5% markup on model cost, no minimum; free tier on free models at 200 requests a day | Teams that want a hosted router with EU data residency |

Kong AI Gateway | Plus from $25/mo plus usage; $100/mo per proxied LLM; fully self-hosted is a custom quote | Enterprises already running Kong for API traffic who want LLM traffic governed alongside it |

Together AI | Per-token serverless pricing, no minimum; gpt-oss-120b is $0.15 in and $0.60 out per 1M tokens | Teams whose traffic is mostly open-weight models |

Amazon Bedrock | Per-model token rates by region; batch is 50% lower on select models; no single list price | Teams already on AWS that want models billed to the same account |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Oct 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 4 of 8 do not publish a comparable monthly price and are left out rather than estimated.

1

### Vercel AI Gateway

Top Pick

Best for: Hosted replacement for teams that want the lowest fee and a base URL swap

PricingNo markup or platform fee on tokens; $5/mo free credit, then pay-as-you-go credits

+Tokens are billed at provider list rates with zero markup, and BYOK adds no fee on the paid tier

+Failover moves a degraded request to the same model on another provider

+Speaks OpenAI Chat Completions, Responses and Anthropic Messages, so existing SDK code needs a new base URL

−BYOK needs the paid tier and purchased credits, and the free tier covers only a subset of models

−Optional add-ons such as custom reporting carry surcharges beyond token rates, and payment processing fees are yours

Visit Vercel AI Gateway →

2

### LiteLLM

Best for: Teams that must keep prompts and keys on their own infrastructure

PricingOpen source $0 self-hosted; Enterprise is a sales quote sized to request volume, never per token

+Virtual keys, budgets, rate limits, spend tracking and fallbacks are in the free build

+Vendor count is 140+ providers and 1,800+ models behind one API

+Self-hosted, so your data and your keys never leave your own infrastructure

−You deploy, scale and patch the gateway, so the real bill is the server and the engineer

−SSO, SCIM and audit logs sit in Enterprise, which publishes no list price

Visit LiteLLM →

3

### Cloudflare AI Gateway

Best for: Apps already running on Cloudflare that need caching and rate limits

PricingCore features free on all plans; 5% fee only on credits bought via Unified Billing

+Dashboard analytics, caching and rate limiting are core features offered for free

+Provider routes cover Workers AI, OpenAI, Anthropic, Google, Bedrock, Groq and more

+Data loss prevention scanning is free on all plans, with two predefined profiles without a Zero Trust subscription

−Guardrails run on Workers AI and bill as token inference, so safety checks add cost

−Dynamic routing, guardrails and key storage are labeled beta in the docs

Visit Cloudflare AI Gateway →

4

### Portkey

Best for: Production teams that want routing, guardrails and traces in one product

PricingFree for prototyping; Production $49/mo with 100k recorded logs; Enterprise custom

+Production adds LLM and partner guardrails, simple and semantic caching, role-based access control and 30-day logs

+An open-source gateway exists for teams that want to host the routing layer themselves

+Unlimited prompt templates and versioning on Production

−The free Developer plan is labeled not suitable for production and keeps logs for 3 days

−Overage is $9 per additional 100k requests, and SSO is Enterprise only

Visit Portkey →

5

### Requesty

Best for: Teams that want a hosted router with EU data residency

Pricing5% markup on model cost, no minimum; free tier on free models at 200 requests a day

+600+ models behind one OpenAI-compatible endpoint, with 20+ providers listed

+EU data residency is included on every plan, including the free one

+Routing policies, caching, fallbacks, spend limits and an MCP gateway on pay as you go

−The flat markup is the kind of fee you may be leaving: $10 per 1M tokens becomes $10.50

−SSO, full RBAC and audit logs are Enterprise features

Visit Requesty →

6

### Kong AI Gateway

Best for: Enterprises already running Kong for API traffic who want LLM traffic governed alongside it

PricingPlus from $25/mo plus usage; $100/mo per proxied LLM; fully self-hosted is a custom quote

+A serverless AI gateway control plane is $25/mo, and a hybrid one is $200/mo

+Paid plugins include semantic cache, semantic prompt guard, sanitizer and advanced rate limiting

+One platform covers API gateways and LLM traffic for teams that already standardize on Kong

−Plus charges for each LLM proxied and caps it at 5 models, so a wide catalog gets expensive

−It is a gateway and not a marketplace, so you keep each provider account and bill

Visit Kong AI Gateway →

7

### Together AI

Best for: Teams whose traffic is mostly open-weight models

PricingPer-token serverless pricing, no minimum; gpt-oss-120b is $0.15 in and $0.60 out per 1M tokens

+Serverless has no provisioning and no minimum cost, and you pay only for tokens processed

+The catalog lists DeepSeek, GLM, Qwen, Kimi, MiniMax and gpt-oss with input, cached and output rates

+Fine-tuning and dedicated inference sit under the same account

−The serverless chat catalog has no Claude models, so closed-model traffic needs another route

−There is no multi-provider failover, since it only serves what it hosts

Visit Together AI →

8

### Amazon Bedrock

Best for: Teams already on AWS that want models billed to the same account

PricingPer-model token rates by region; batch is 50% lower on select models; no single list price

+Models from Anthropic, Meta, Mistral, Amazon, DeepSeek, Qwen, OpenAI and others share one AWS account

+Select models offer batch inference at a 50% lower price than on-demand

+Charges land on the AWS bill and commitments you already manage

−Rates differ by model and region, so there is no single fee to compare with OpenRouter's

−It serves the models Bedrock hosts, not other vendors' APIs, so it is not a drop-in router

Visit Amazon Bedrock →

## What it is

[OpenRouter](https://toolradar.com/tools/openrouter) is a hosted marketplace and router: 500+ models from 80+ providers behind one API and one credit balance. Its free plan covers 25+ free models at 50 requests a day, which suits a prototype and nothing heavier.

The paid plans add auto-routing, budgets and spend controls, and the fee schedule is what sends developers shopping.

Bringing your own provider keys does not remove the fee entirely. Paid plans include $25,000 a month of list-price inference with no BYOK fee and then charge 5%, so heavy users still pay for the routing layer they lean on.

On August 19, 2026, OpenRouter [announced that it is joining Stripe](https://openrouter.ai/blog/announcements/openrouter-is-joining-stripe/) and said the product, name and roadmap continue as they are. That makes this page about fees and control, not a forced move.

Three fee models separate the alternatives.

A hosted router adds a percentage or nothing, a gateway you run yourself adds a server bill, and a cloud model host skips the gateway layer and bills tokens directly.

Pick the model before the vendor, because a free gateway that needs an engineer on call is not cheaper than a small markup for a team of two.

How we compared: we read each vendor's pricing page, documentation and public announcements in October 2026, in USD, and ranked the picks by how directly each replaces OpenRouter for a developer team.

We did not run these gateways, so the comparison rests on what vendors publish. Plans change often, so confirm the fee with the vendor before you commit.

## Why it matters

The fee grows with spend, and switching is the one way to change it. At $10,000 of monthly model spend, OpenRouter Standard adds $550 and Requesty adds $500. Cloudflare matches Requesty only if you buy credits through Unified Billing, because its fee applies to credits and not to tokens paid with your own provider keys.

Vercel AI Gateway and a self-hosted LiteLLM add no fee, so the real cost of LiteLLM is the server and the person who patches it.

Switching costs less than it sounds. Vercel, Requesty and LiteLLM all expose an OpenAI-compatible interface, so the move is mostly a base URL swap and a new key, then a week of watching error rates and spend.

Moving from a marketplace to your own provider keys is the bigger change, because you then hold rate limits, billing and data terms with each provider yourself.

Flat subscriptions suit teams with high call volume and few models.

Portkey's Production plan bills for 100k recorded logs, and Kong charges $100/mo for each LLM proxied through its AI plugins, so those bills follow logs and models, not tokens. That wins at scale and loses for a side project that makes a few thousand calls a month.

Where the gateway sits in your stack matters as much as its price.

If the app deploys on Vercel, start with the Vercel review and the serverless platforms guide.

If you are wiring models into a product, the AI app development platforms guide and the Dify review cover the layer above the gateway, and the AI coding assistants guide covers tools with their own routing built in.

Self-hosting a proxy belongs with the self-hosting tools guide, and spend tracking with the monitoring tools guide. Later price changes show up in the Devshot brief.

## Key features to look for

Who adds a fee on top of tokens
Vercel and a self-hosted LiteLLM add none, Requesty adds a flat percentage, and Cloudflare adds one only on credits bought through Unified Billing. Portkey and Kong bill for the platform, so their cost follows logs and proxied models, not tokens.

Hosted or self-hosted
Vercel, Cloudflare and Requesty run the gateway for you. LiteLLM runs on your infrastructure, which keeps prompts and keys with you but makes uptime yours. Portkey offers both, and Kong sells hosted control planes next to self-managed ones.

Whether your own provider keys are supported
Bring-your-own-key is how you keep your own provider rate limits and negotiated pricing. Vercel supports it on the paid tier, Requesty and LiteLLM list it, and Together and Bedrock skip the question because you buy from them directly.

Guardrails and logging depth
Portkey Production includes guardrails and 30 days of logs. Cloudflare scans for sensitive data for free but bills its Guardrails feature as Workers AI inference, and Kong sells prompt guards and sanitizers as paid plugins.

Model catalog and who hosts it
Requesty lists 600+ models and LiteLLM 1,800+ across 140+ providers, each by its own count, so compare the models you need and not the headline. Together hosts open models itself and Bedrock hosts its partners' models, so neither routes to other vendors' APIs.

Where your data lives
Requesty includes EU data residency on every plan. OpenRouter limits in-region routing to Business and Enterprise, and LiteLLM keeps traffic on your own servers, which is the strongest answer for a strict residency review.

## Pricing

Every figure was read on the vendor's own page in USD and checked in October 2026: [OpenRouter pricing](https://openrouter.ai/pricing), [Vercel AI Gateway pricing](https://vercel.com/docs/ai-gateway/pricing), [LiteLLM pricing](https://www.litellm.ai/pricing), [Cloudflare AI Gateway pricing](https://developers.cloudflare.com/ai-gateway/reference/pricing/), [Portkey pricing](https://portkey.ai/pricing), [Requesty pricing](https://www.requesty.ai/pricing), [Kong pricing](https://konghq.com/pricing), [Together AI models and pricing](https://docs.together.ai/docs/inference-models) and [Amazon Bedrock pricing](https://aws.amazon.com/bedrock/pricing/).

OpenRouter's fee terms also appear in its [FAQ](https://openrouter.ai/docs/faq).

Read the unit before the number. A percentage grows with every token, a flat subscription grows with logs or models, and a self-hosted gateway grows with your own cloud bill.

Vercel says payment processing fees are yours unless you pay by invoice on an Enterprise contract.

Cloudflare shows how a small print line changes the comparison: a $100 credit purchase through Unified Billing is a $105 charge, yet the gateway itself, with caching and rate limiting, is free.

Kong's serverless control plane is $25/mo and its hybrid plane is $200/mo, so the entry price says little until you count proxied models and requests.

Plan | Price | Best for |

OpenRouter Free | Free | 25+ free models and 4 free providers, capped at 50 requests a day |

OpenRouter Standard | 5.5% fee | 500+ models, 80+ providers, budgets, prompt caching and email support |

OpenRouter Business | 8% fee | Adds EU and US in-region routing and up to 1,000 workspaces |

OpenRouter BYOK | 5% fee | Applies after $25,000 a month of list-price inference with your own keys |

Vercel AI Gateway free tier | $5/mo credit | A subset of models at lower rate limits, with no BYOK |

Vercel AI Gateway paid | No fee | All models at provider list rates, BYOK and custom rate limits, bought as credits |

LiteLLM Open Source | $0 | Self-hosted gateway with virtual keys, budgets, rate limits and fallbacks |

LiteLLM Enterprise | Contact sales | Adds SSO, SCIM, audit logs and SLAs, sized to request capacity |

Cloudflare AI Gateway core | Free | Analytics, caching and rate limiting on every Cloudflare plan |

Cloudflare Unified Billing | 5% fee | Charged on credits purchased; token rates pass through at provider prices |

Portkey Developer | Free | 10k recorded logs a month with 3-day log retention, for prototypes only |

Portkey Production | $49/mo | 100k recorded logs, guardrails, semantic caching and role-based access |

Requesty Free | Free | Free models only, 200 requests a day, with routing and EU data residency |

Requesty Pay as you go | 5% markup | All 600+ models, your own keys, spend limits and an MCP gateway |

Kong AI Plus, serverless | $25/mo | Per control plane, up to 5 control planes and no SLA |

Kong AI Plus, hybrid | $200/mo | Per control plane, self-managed data plane, up to 2 control planes |

Together AI serverless | Per token | Open and partner models, no minimum, rates listed per model |

Amazon Bedrock on-demand | Per token | Rates vary by model and region, so there is no single list price |

Mistakes to avoid

×Comparing a percentage fee with a flat subscription at one volume misleads, because a percentage is cheap at $1,000 a month and costly at $100,000, while a subscription does the reverse.

×Calling a self-hosted gateway free ignores the server, the on-call rotation and the upgrade work that LiteLLM leaves with you.

×Assuming bring-your-own-key removes the fee is wrong for OpenRouter, which still charges beyond $25,000 of monthly list-price inference.

×Treating Together AI or Amazon Bedrock as routers fails, because each serves the models it hosts and does not fail over to other vendors' APIs.

Expert tips

→Pull last month's spend by provider before you shop, then price each fee model at that number and at three times it.

→Start with the hosted option that has no markup, and move to a self-hosted proxy only when data residency or a custom control forces it.

→Keep the OpenAI-compatible base URL and model names in config, so a second gateway is a deploy and not a rewrite.

## The bottom line

Choose [Vercel AI Gateway](https://toolradar.com/tools/vercel) when you want OpenRouter's one-endpoint convenience without a percentage fee, and choose [LiteLLM](https://toolradar.com/tools/litellm) when prompts and keys must stay on servers you run.

[Cloudflare AI Gateway](https://toolradar.com/tools/cloudflare) fits an app already on Cloudflare, and [Portkey](https://toolradar.com/tools/portkey) fits a team that wants guardrails and traces to ship together with routing.

Requesty is the nearest like-for-like swap if you want a hosted router with EU residency, and Kong fits a company that already governs API traffic with it.

[Together AI](https://toolradar.com/tools/together-ai) and [Amazon Bedrock](https://toolradar.com/tools/amazon-bedrock) suit teams that would rather buy tokens from one host than route at all.

Stay on [OpenRouter](https://toolradar.com/tools/openrouter) when its single credit balance is worth the fee to you.

Fees and plan limits were read from vendor pages and docs, and they move often. Cite this: Devshot, "Best OpenRouter Alternatives in 2026", October 2026.

## Frequently asked questions

What is the best OpenRouter alternative in 2026?

Vercel AI Gateway is the closest hosted replacement, since it charges provider list prices with no markup and no platform fee. LiteLLM is the pick for self-hosting at $0, Cloudflare AI Gateway suits apps already on Cloudflare, and Portkey from $49/mo suits teams that want guardrails and logs. Fees were checked on vendor pages in October 2026.

Is there a cheaper alternative to OpenRouter?

Yes, in fee terms. Vercel AI Gateway adds no fee on tokens, LiteLLM is free to self-host, and Cloudflare AI Gateway's core features are free, while OpenRouter charges 8% on Business. Requesty is cheaper than that tier, though a flat markup is still a markup. A self-hosted gateway moves the cost to your server and your engineers, so count them before calling it free.

Is there an open-source or self-hosted OpenRouter?

LiteLLM is the main one: an open-source proxy you host, with virtual keys, budgets, rate limits and fallbacks included, and 140+ providers by its own count. Portkey also offers an open-source gateway you can host. Neither gives you OpenRouter's single marketplace balance, because you bring your own provider accounts and keys.

What fee does OpenRouter charge?

OpenRouter's pricing page lists a platform fee of 5.5% on Standard and 8% on Business, with Enterprise fee discounts available, and its FAQ says provider token prices pass through without markup. With your own keys, the first $25,000 a month of list-price inference has no BYOK fee and 5% applies after that. The free plan is limited to 25+ free models and 50 requests a day.

OpenRouter vs LiteLLM vs Portkey: which should I pick?

Pick OpenRouter for a hosted marketplace with one balance, LiteLLM for a gateway you run with no license fee, and Portkey for a managed gateway with guardrails and traces on its Production plan. The first adds a percentage fee, the second adds a server, and the third adds a log-based subscription. Choose the one whose cost grows slowest with your traffic.

Which LLM gateway is best for production?

Production needs fallbacks, spend limits, logs and access control, and several options include them. Portkey Production bundles guardrails, caching, role-based access and 30-day logs. LiteLLM includes budgets and fallbacks for free, but SSO needs Enterprise. Vercel adds failover with no fee on tokens, and Kong suits companies that already run it. Choose by who owns uptime, not by feature count.

Related guides

Api GatewaysAi App Development PlatformsAi Coding AssistantsDify ReviewVercel ReviewServerless PlatformsMonitoring ToolsSelf Hosting ToolsClaude Code pricingAI for developersDevshot briefFree LLM APIs in 2026: Rate Limits and Training TermsLlm ApisAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Static Site Generators in 2026

URL: https://devshot.co/reviews/best-static-site-generators
Type: review
Published: 2026-10-08
Updated: 2026-10-08
Summary: Astro, Hugo, Eleventy, Next.js, Docusaurus, VitePress, Zola and Jekyll compared on version, GitHub stars and the free hosting limits that decide the real bill.

Expert Guide

## The Best Static Site Generators in 2026

Eight open-source generators compared on release freshness, GitHub stars and what the free tiers of Cloudflare Pages, Netlify and GitHub Pages let you ship.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Oct 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

Astro is the best static site generator for most content sites in 2026, and Hugo is the pick when build time is the constraint. All eight tools here are free and open source, so the bill is hosting: Cloudflare Pages and GitHub Pages have free tiers, and Netlify is free with 300 credits a month, then $9 on Personal.

Choose Docusaurus or VitePress for documentation, Eleventy for minimal JavaScript, and Next.js only if your team already ships React.

Key facts8 tools compared: Astro, Hugo, Eleventy (Build Awesome), Next.js (static export), Docusaurus…

- Updated: October 8, 2026
- Top pick: Astro (best for: Blogs, marketing sites and content sites that want almost no JavaScript)
- Top pick price as of October 8, 2026: Astro: Free and open source; 63,114 GitHub stars; astro 7.3.7 (October 7, 2026).

- 8 tools compared: Astro, Hugo, Eleventy (Build Awesome), Next.js (static export), Docusaurus, VitePress, Zola, Jekyll
- Hugo (best for: Large sites with thousands of pages where build time is the bottleneck): Free and open source (Apache-2.0); 90,057 GitHub stars; v0.167.0 (September 28, 2026).
- Eleventy (Build Awesome) (best for: Developers who want plain HTML output and a choice of template languages): Free and open source (MIT); 19,954 GitHub stars; v3.1.6 stable (June 2, 2026).
- Next.js (static export) (best for: React teams that want a marketing site in the same codebase as their app): Free and open source (MIT); 143,251 GitHub stars; v16.4.0 (October 7, 2026).

Nothing about choosing a generator costs money, which makes the choice easy to get wrong. The cost shows up later, as a build cap you hit, a framework migration, or a docs site that outgrows its host.

Two shifts matter this year.

Astro shipped version 7.0 on June 22, 2026, with Vite 8 and a new Rust compiler aimed at faster builds. Its own blog says Evil Martians moved a site off Gatsby to Astro in nine days without rewriting React components. Gatsby's latest release on GitHub is gatsby@5.16.1 from February 10, 2026, and it is not in the ranking.

Eleventy has been renamed Build Awesome, and its repository now lives at 11ty/buildawesome.

Devshot data: [Toolradar's static site generator ranking](https://toolradar.com/best/static-site-generators) evaluated 26 tools as of October 8, 2026, and 25 of them have a free plan.

We took eight of those 26 and read their GitHub repositories the same day: together they hold 470,394 stars, and five of the eight published a release since September 1, 2026 (Astro, Hugo, Next.js, Zola and VitePress).

If you are weighing where the site will live, pair this with our serverless platforms guide and the Netlify review.

## Top Picks

Based on features, real-world fit, and value for money.

Best Static Site Generators in 2026: 8 tools compared, updated Oct 2026

Tool | Pricing | Best for |

Astro | Free and open source; 63,114 GitHub stars; astro 7.3.7 (October 7, 2026). | Blogs, marketing sites and content sites that want almost no JavaScript |

Hugo | Free and open source (Apache-2.0); 90,057 GitHub stars; v0.167.0 (September 28, 2026). | Large sites with thousands of pages where build time is the bottleneck |

Eleventy (Build Awesome) | Free and open source (MIT); 19,954 GitHub stars; v3.1.6 stable (June 2, 2026). | Developers who want plain HTML output and a choice of template languages |

Next.js (static export) | Free and open source (MIT); 143,251 GitHub stars; v16.4.0 (October 7, 2026). | React teams that want a marketing site in the same codebase as their app |

Docusaurus | Free and open source (MIT); 66,430 GitHub stars; v3.10.2 (July 10, 2026). | Product documentation that needs versions, translations and search |

VitePress | Free and open source (MIT); 18,385 GitHub stars; 2.0.0-alpha.20 (September 4, 2026). | Vue and Vite developers writing documentation |

Zola | Free and open source (EUPL-1.2); 17,498 GitHub stars; v0.23.6 (September 12, 2026). | Personal sites and small docs where a Rust binary beats a Node toolchain |

Jekyll | Free and open source (MIT); 51,705 GitHub stars; v4.4.1 (January 29, 2025). | GitHub Pages sites where you want zero build configuration |

Pricing read from each vendor's own published pricing page, checked Oct 2026. Every vendor here publishes a price.

1

### Astro

Top Pick

Best for: Blogs, marketing sites and content sites that want almost no JavaScript

PricingFree and open source; 63,114 GitHub stars; astro 7.3.7 (October 7, 2026).

+Docs describe islands, so pages ship little client JavaScript by default

+Supports React, Preact, Svelte, Vue, Solid and web components in one project

+Astro 7.2 added experimental incremental static builds for large sites

−Fast release cadence: 7.0 landed June 22, 2026 and 7.3.7 followed by October 7, so pin versions

−The repository license shows as unrecognized on GitHub, so read the LICENSE file before adopting

Visit Astro →

2

### Hugo

Best for: Large sites with thousands of pages where build time is the bottleneck

PricingFree and open source (Apache-2.0); 90,057 GitHub stars; v0.167.0 (September 28, 2026).

+Project states sites build in seconds, often less, the strongest speed claim in this list

+Apache-2.0 license

+No Node.js toolchain: it is written in Go

−Version numbers are still 0.x, so minor releases can change behavior

−Its pitch is speed and flexibility, not a component model like Astro's UI framework support

Visit Hugo →

3

### Eleventy (Build Awesome)

Best for: Developers who want plain HTML output and a choice of template languages

PricingFree and open source (MIT); 19,954 GitHub stars; v3.1.6 stable (June 2, 2026).

+Supports HTML, Markdown, JavaScript, WebC, Nunjucks, Liquid, Handlebars, Mustache, EJS, HAML, Pug and Sass templates

+Both npm package names work, @awesome.me/buildawesome and the legacy @11ty/eleventy

+No client-side framework forced on you

−The newest stable build is from June, while 4.0 alphas ran through July 1, so the next major is not final

−The rename splits search results and tutorials between two names

Visit Eleventy (Build Awesome) →

4

### Next.js (static export)

Best for: React teams that want a marketing site in the same codebase as their app

PricingFree and open source (MIT); 143,251 GitHub stars; v16.4.0 (October 7, 2026).

+Set output to export and next build writes plain files to an out folder

+Server Components run at build time during a static export

+Deploys to any host that serves HTML, CSS and JavaScript, including Nginx

−Static export drops Incremental Static Regeneration, Server Actions, rewrites, redirects and headers

−The default image loader does not work in static export, so you bring Cloudinary or another loader

Visit Next.js (static export) →

5

### Docusaurus

Best for: Product documentation that needs versions, translations and search

PricingFree and open source (MIT); 66,430 GitHub stars; v3.10.2 (July 10, 2026).

+Document versioning keeps docs in sync with project releases

+MDX lets you embed interactive React components inside Markdown

+Translations and site search are built in rather than plugins you wire up

−Built on React, so even a simple docs page ships a client-side app

−Last release was July 10, so it moves slower than Astro or Next.js

Visit Docusaurus →

6

### VitePress

Best for: Vue and Vite developers writing documentation

PricingFree and open source (MIT); 18,385 GitHub stars; 2.0.0-alpha.20 (September 4, 2026).

+Powers the docs of Vite, Rollup, Pinia, VueUse, Vitest and Vue itself

+Vite dev server gives instant starts and edits under 100 ms by the project's own claim

+Static HTML first, then single-page navigation that the docs say feels instant

−The latest GitHub release is an alpha, so pin the version and read changelogs before upgrading

−Vue-centric: React teams get no benefit from its component model

Visit VitePress →

7

### Zola

Best for: Personal sites and small docs where a Rust binary beats a Node toolchain

PricingFree and open source (EUPL-1.2); 17,498 GitHub stars; v0.23.6 (September 12, 2026).

+Built-in Sass, syntax highlighting, search index and image processing

+Tera templates feel familiar if you know Jinja2, Django or Liquid

+Active project: v0.23.6 shipped September 12, 2026

−Smallest star count in this list at 17,498, so fewer themes and answers online

−Still 0.x, and the EUPL-1.2 license is less common than MIT, which some legal teams review

Visit Zola →

8

### Jekyll

Best for: GitHub Pages sites where you want zero build configuration

PricingFree and open source (MIT); 51,705 GitHub stars; v4.4.1 (January 29, 2025).

+GitHub says Pages is powered by Jekyll, so a push builds your site for free

+Files without front matter are treated as static assets, so plain HTML works

+51,705 stars and years of themes and tutorials

−Latest stable release is January 29, 2025, the oldest of the eight

−It needs a Ruby toolchain, which is a cost if your team writes JavaScript

Visit Jekyll →

## What it is

A static site generator is a build tool that turns templates and content, usually Markdown, into plain HTML, CSS and JavaScript files that any web server or CDN can host.

There is no application server to patch, so the attack surface is small and the hosting bill is close to nothing.

The trade-off is that anything personalized or real time has to come from client-side code or an external API.

That is why two of the eight below, [Next.js](https://nextjs.org/docs/app/guides/static-exports) and Astro, are really frameworks that can also emit static files, and why Next.js loses features such as Incremental Static Regeneration and Server Actions in static export mode.

## Why it matters

Three costs decide this choice, and none is the license. The first is builds: Cloudflare Pages allows 500 a month on its free plan, about 16 a day, and a monorepo that rebuilds on every push can burn through that. The second is the file cap.

A Pages site holds up to 20,000 files on the free plan, so a docs site with many locales and versions can reach it.

The third is bandwidth priced in credits. Netlify charges 15 credits per production deploy and 20 per GB of bandwidth, against 300 free credits a month.

Our arithmetic: a site that publishes 10 times a month spends 150 credits on deploys and has 150 left, which covers 7.5 GB of traffic before the free credits run out and the Personal plan becomes the next step.

Methodology. We read each project's repository through the GitHub API on October 8, 2026 for stars, license and latest release, and each project's own documentation for what it claims.

Hosting limits were read on Cloudflare's, Netlify's and GitHub's own pages the same day. We ran no build benchmarks, so speed is quoted only where a project publishes its own claim. We re-check this page when a major version or a hosting price changes.

## Key features to look for

Content modelEssential
Markdown files, MDX components or a CMS API. This decides how much your writers can ship without a developer.

JavaScript shipped to readersEssential
Astro and Eleventy send plain HTML by default. VitePress and Docusaurus hydrate a single-page app on top, which feels instant to navigate but adds weight.

Release freshness
Check the date of the last stable release. Jekyll's is January 29, 2025, while Astro shipped 7.3.7 on October 7, 2026.

Docs features
Versioning, search and translations. Docusaurus lists all three as built in, which saves a week of wiring on a docs site.

Host limits
Build counts, file caps and credit-based bandwidth on the free tier of the host you pick, not the generator.

## Pricing

Every generator here is free. What you pay for is where the files live, and prices below were read on the host's own pricing and limits pages on October 8, 2026. GitHub Pages documents a 1 GB published site size, a soft 100 GB monthly bandwidth limit and a soft limit of 10 builds an hour, and its limits page lists no price.

Cloudflare's Pages limits page lists 500 builds a month, 20,000 files per site and a 25 MiB cap on any single file on the free plan.

Netlify is the only one of the three with a published paid ladder: Free, Personal at $9 a month with 1,000 credits, and Pro at $20 a month with 3,000.

The cheapest way to stay free is Cloudflare Pages or GitHub Pages for a plain static site, and Netlify when you want deploy previews on every branch.

Sources: [Cloudflare Pages limits](https://developers.cloudflare.com/pages/platform/limits/), [Netlify pricing](https://www.netlify.com/pricing/) and [GitHub Pages limits](https://docs.github.com/en/pages/getting-started-with-github-pages/github-pages-limits).

Plan | Price | Best for |

All eight generators | Free | Open source projects, no per-seat or per-site fee for the software |

Cloudflare Pages Free | Free | 500 builds a month, 20,000 files per site, 25 MiB per file |

Cloudflare Pages Pro | Paid plan | 5,000 builds a month and up to 100,000 files per site |

Netlify Free | Free | 300 credits a month; production deploys cost 15 credits, bandwidth 20 credits per GB |

Netlify Personal | $9/month | 1,000 credits a month, secret detection and priority email support |

Netlify Pro | From $20/month | 3,000 credits a month, unlimited members, private organization repos |

GitHub Pages | Free | 1 GB site, soft 100 GB monthly bandwidth, soft 10 builds an hour |

Mistakes to avoid

×Picking by stars alone. Next.js has 143,251 stars, but a static export gives up Incremental Static Regeneration, Server Actions and redirects, which a pure generator never promised.

×Choosing Docusaurus or VitePress for a marketing site. They are built for documentation, so you will fight the sidebar layout and ship a single-page app for a landing page.

×Ignoring the host's build cap. A site that rebuilds on every commit can exceed Cloudflare's 500 free builds a month during a busy week.

Expert tips

→Run your real content through the generator before you commit, because build time depends on page count and images, not on a benchmark chart.

→Pin the generator version in your lockfile. Astro ships patch releases most weeks, and VitePress's latest release is still an alpha.

→Check the date of the latest stable release in each repository. A project with no stable release for 20 months, like Jekyll, can still be a fine choice, but you should choose it knowingly.

## The bottom line

Pick Astro for a blog or marketing site, Hugo when thousands of pages make build time the problem, and Docusaurus or VitePress when the site is documentation.

Use Eleventy if you want the least JavaScript and the most template choice, Zola if you prefer a Rust binary, and Jekyll only for GitHub Pages sites you already run.

Choose Next.js static export when the team already ships React.

Start on a free tier: Cloudflare Pages, GitHub Pages or Netlify Free cost nothing, and moving later is a config change rather than a rewrite.

Cite this: Devshot, "Best Static Site Generators in 2026", October 2026.

Related: best Git clients, API documentation tools, best code editors, Vercel review, Vercel vs Netlify and cheaper Vercel alternatives.

Tool pages: [Astro](https://toolradar.com/tools/astro), [Hugo](https://toolradar.com/tools/hugo), [Eleventy](https://toolradar.com/tools/eleventy), [Next.js](https://toolradar.com/tools/nextjs), [Docusaurus](https://toolradar.com/tools/docusaurus), [Jekyll](https://toolradar.com/tools/jekyll), [Zola](https://toolradar.com/tools/zola), [VitePress](https://toolradar.com/tools/vitepress).

Stars and releases from GitHub: [withastro/astro](https://github.com/withastro/astro), [gohugoio/hugo](https://github.com/gohugoio/hugo), [11ty/buildawesome](https://github.com/11ty/buildawesome), [vercel/next.js](https://github.com/vercel/next.js), [facebook/docusaurus](https://github.com/facebook/docusaurus), [vuejs/vitepress](https://github.com/vuejs/vitepress), [getzola/zola](https://github.com/getzola/zola), [jekyll/jekyll](https://github.com/jekyll/jekyll).

## Frequently asked questions

What is the best static site generator in 2026?

Astro is the best pick for most content sites in 2026: it ships HTML first, supports React, Vue, Svelte and other UI frameworks, and released version 7.3.7 on October 7, 2026. Choose Hugo for very large sites, or Docusaurus and VitePress for documentation.

How much does it cost to host a static site?

Hosting can be free. Cloudflare Pages allows 500 builds a month and 20,000 files on its free plan, GitHub Pages allows a 1 GB site, and Netlify Free gives 300 credits a month. Netlify Personal is $9 a month and Pro is from $20 a month, all read on October 8, 2026.

Are static site generators free?

Yes. Astro, Hugo, Eleventy, Next.js, Docusaurus, VitePress, Zola and Jekyll are open source projects with no license fee, and 25 of the 26 static site generators Toolradar evaluated have a free plan. The costs are hosting, a CMS if you want one, and developer time.

Is Eleventy the same as Build Awesome?

Yes. The Eleventy repository is now 11ty/buildawesome and its documentation says Build Awesome, but you can install either @awesome.me/buildawesome or the legacy @11ty/eleventy package. The latest stable release is v3.1.6 from June 2, 2026.

Can Next.js be a static site generator?

Yes, with output set to export in next.config.js, which writes HTML, CSS and JavaScript to an out folder. Incremental Static Regeneration, Server Actions, rewrites, redirects and the default image loader do not work in that mode.

Is Jekyll still worth using in 2026?

For GitHub Pages sites, yes, because GitHub builds Jekyll for you. Its latest stable release, v4.4.1, is from January 29, 2025, so for a new project with a JavaScript team, Astro or Eleventy fit better.

Related guides

Serverless PlatformsApi Documentation ToolsNetlify ReviewVercel ReviewGit ClientsCode EditorsRender ReviewDigitalocean ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Backend-as-a-Service Platforms in 2026

URL: https://devshot.co/reviews/best-backend-as-a-service
Type: review
Published: 2026-09-28
Updated: 2026-09-28
Summary: The backend-as-a-service platforms worth shipping on in 2026, compared on database model, free-tier pauses, and the USD price each vendor publishes.

Expert Guide

## The Best Backend-as-a-Service Platforms in 2026

The database and the meter decide the backend in 2026. An organization plan, a per-developer plan, and a peak-requests plan can share a sticker and still be different purchases.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 9 tools compared

TL;DR

Supabase is the backend most web teams should start on in 2026: Pro is $25/mo per organization, the database is Postgres, and auth, storage, and edge functions come with it. Firebase is the pick when the client is a phone and offline sync matters.

Appwrite Pro, Nhost Pro, and Convex Professional publish that same entry rate on different meters, so matching stickers still hide different bills. Xano Essential is $85/mo when you pay annually and want the API drawn instead of written. AWS Amplify has no platform minimum and fits an account already on AWS, while PocketBase publishes no hosted price because you run the binary.

Key facts9 tools compared: Supabase, Firebase, Appwrite, Convex, AWS Amplify, Nhost, Xano, Backendless…

- Updated: September 28, 2026
- Top pick: Supabase (best for: Web apps that want Postgres, auth, and files without running the server)
- Top pick price as of September 28, 2026: Supabase: From $25/mo (Pro, per organization); free plan pauses after 1 week

- 9 tools compared: Supabase, Firebase, Appwrite, Convex, AWS Amplify, Nhost, Xano, Backendless, PocketBase
- Firebase (best for: Mobile apps that need offline sync and the rest of Google Cloud): Spark is free; Blaze is usage with no monthly platform fee
- Appwrite (best for: Teams that want a Docker-hosted BaaS and a per-project cloud bill): From $25/mo per project (Pro); free plan pauses after 1 week
- Convex (best for: TypeScript apps that should update the client when data changes): From $25/mo per developer (Professional, up to 20); free plan

A backend-as-a-service is a hosted database, auth, file storage, and functions behind one API, so the app does not start with a server you patch. The short list in 2026 is not nine copies of [Firebase](https://toolradar.com/tools/firebase).

The database and the meter decide it, and a Postgres bill, a per-developer bill, and a peak-requests bill fail in different months.

Toolradar data: Toolradar, the software directory we run, evaluated 109 tools in its [September 2026 low-code backend ranking](https://toolradar.com/best/low-code-backend).

That is the nearest public category that actually includes [Supabase](https://toolradar.com/tools/supabase), Xano, Backendless, Convex, and PocketBase. This page is the developer cut of that market, not the enterprise low-code cut.

Prices below were verified on each vendor's pricing page in September 2026.

The Supabase vs Firebase write-up is the head-to-head if those two are the whole shortlist.

## Top Picks

Based on features, real-world fit, and value for money.

Best Backend-as-a-Service Platforms in 2026: 9 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Supabase](https://toolradar.com/tools/supabase) | From $25/mo (Pro, per organization); free plan pauses after 1 week | Web apps that want Postgres, auth, and files without running the server |

[Firebase](https://toolradar.com/tools/firebase) | Spark is free; Blaze is usage with no monthly platform fee | Mobile apps that need offline sync and the rest of Google Cloud |

[Appwrite](https://toolradar.com/tools/appwrite) | From $25/mo per project (Pro); free plan pauses after 1 week | Teams that want a Docker-hosted BaaS and a per-project cloud bill |

Convex | From $25/mo per developer (Professional, up to 20); free plan | TypeScript apps that should update the client when data changes |

AWS Amplify | Usage-based, no monthly minimum; free tier (1,000 build minutes a month) | Apps that already live in AWS and should not add another vendor |

Nhost | Pro per organization, same entry rate as Supabase; Starter pauses after 1 week | GraphQL APIs on Postgres when Hasura is the interface you want |

Xano | From $85/mo billed annually (Essential); free plan is rate limited | Teams who will draw the API and the business logic instead of writing them |

Backendless | From $15/mo per app (Scale); free plan for exploring | Visual backends where the bill should follow API peaks, not seats |

PocketBase | No public pricing; open-source binary you run yourself | A single-binary backend on a VPS you already operate |

Pricing read from each vendor's own published pricing page, checked Sep 2026. 1 of 9 does not publish one; those entries say so rather than estimating.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 9 do not publish a comparable monthly price and are left out rather than estimated.

1

### Supabase

Top Pick

Best for: Web apps that want Postgres, auth, and files without running the server

PricingFrom $25/mo (Pro, per organization); free plan pauses after 1 week

+Row-level security is SQL on the table, so every client hits the same rules

+Pro includes 100,000 monthly active users and 8 GB of disk before overage

+Spend caps are on by default on Pro, so usage does not run open-ended unless you allow it

−Free projects pause after a week idle, and only two may be active

−Team, where SOC 2 and ISO 27001 sit, is a separate plan from Pro

Visit Supabase →

2

### Firebase

Best for: Mobile apps that need offline sync and the rest of Google Cloud

PricingSpark is free; Blaze is usage with no monthly platform fee

+Spark includes Auth, 1 GiB of Firestore, 50,000 document reads a day, and Crashlytics

+Blaze Cloud Functions include 2 million invocations, then $0.40 per million

+SQL Connect adds Cloud SQL for PostgreSQL from $9.37/mo after a 3-month trial

−Spark cannot run Cloud Functions, so that meter starts when the project moves to Blaze

−Phone auth bills per SMS rather than a flat fee, so message volume sets that line

Visit Firebase →

3

### Appwrite

Best for: Teams that want a Docker-hosted BaaS and a per-project cloud bill

PricingFrom $25/mo per project (Pro); free plan pauses after 1 week

+Pro includes 2 TB of bandwidth, 150 GB of storage, 3.5 million executions, and 200,000 MAU

+Self-hosted 1.9.6 installs with Docker and needs 2 CPU cores and 4 GB of RAM

+Free allows two projects, 75,000 MAU, and 750,000 executions

−Self-hosted 1.9 defaults to MongoDB, and MariaDB was the default before 1.9.0, so an upgrade does not keep the old engine

−From 1 September 2026, function and site build artifacts count toward billed storage

Visit Appwrite →

4

### Convex

Best for: TypeScript apps that should update the client when data changes

PricingFrom $25/mo per developer (Professional, up to 20); free plan

+Professional includes 25 million function calls, then $2 per extra million

+The free allowance is 1 million calls, then $2.20 per extra million, and 0.5 GB of database

+US East is billed at the published 1.0x rate, so the list price holds in that region

−The seat is the developer, so a larger team pays more even when traffic stays flat

−Business and Enterprise start at a $2,500 monthly minimum, and the store is not Postgres

Visit Convex →

5

### AWS Amplify

Best for: Apps that already live in AWS and should not add another vendor

PricingUsage-based, no monthly minimum; free tier (1,000 build minutes a month)

+Team access is free, with no per-developer charge, so headcount does not move this line

+Standard builds are $0.01 per minute after 1,000 minutes, and large instances are $0.025 per minute

+AWS's published 5-developer example totals $8.08 for build minutes plus hosting

−Auth, APIs, and data are separate AWS services, so the Amplify page is not the whole invoice

−Web Application Firewall is a separate monthly line per app, before the firewall's own usage

Visit AWS Amplify →

6

### Nhost

Best for: GraphQL APIs on Postgres when Hasura is the interface you want

PricingPro per organization, same entry rate as Supabase; Starter pauses after 1 week

+Pro includes 10 GB of database and 50 GB of storage, plus Hasura subscriptions and event triggers

+Database past that allowance is $0.20/GB, storage past 50 GB is $0.05/GB

+Dedicated compute is $50 per vCPU per month, at $0.0012 per vCPU-minute, with 2 GB of RAM per vCPU

−Starter keeps a single project active, so extra free projects have to be paused

−Egress past 50 GB is $0.10/GB, and function time past 10 GB-hours is $0.18 per GB-hour

Visit Nhost →

7

### Xano

Best for: Teams who will draw the API and the business logic instead of writing them

PricingFrom $85/mo billed annually (Essential); free plan is rate limited

+Essential drops the free cap of 10 requests per 20 seconds and adds direct database access

+The paid Postgres database is not capped at the free plan's 100,000 records

+Essential lists SOC 2, ISO, and GDPR, with 5 seats and 3 workspaces

−The Essential and Pro rates are the annual-billing prices, and the month-to-month figure is a different number

−Free file storage is 1 GB and watermarked, and branching starts on Essential rather than Free

Visit Xano →

8

### Backendless

Best for: Visual backends where the bill should follow API peaks, not seats

PricingFrom $15/mo per app (Scale); free plan for exploring

+Developer seats are unlimited, and Scale Variable tier 1 is $0.50 per day at 0 to 5 requests per minute, so a quiet large team stays near the floor

+Scale raises realtime connections from 5 on Free to 1,000, and tables from 20 to 100

+File storage on Scale is 10 GB, and another 10 GB is a $20/mo pack

−Cloud Code still stops at 5 seconds on Scale, and the 20-second pack is $30/mo, so a longer job is an add-on

−A quiet app can sit near the published monthly floor, and a spiky day jumps tiers inside the same month

Visit Backendless →

9

### PocketBase

Best for: A single-binary backend on a VPS you already operate

PricingNo public pricing; open-source binary you run yourself

+Version 0.40.4 bundles the database, auth, files, and an admin UI in one binary

+Hooks in Go or JavaScript extend it without a second vendor

+Files can stay on local disk or move to S3

−You own uptime, backups, and disk, because the project publishes no managed plan and no SLA

−The database is embedded SQLite, so a second app server does not share it the way Postgres does

Visit PocketBase →

## What it is

You are buying a data store plus sign-in, files, code that must not ship in the client, and usually a generated API, and the purchase is not a language.

Supabase, Nhost, and Xano put Postgres under that API, so a team that already writes SQL can leave with a dump instead of a rewrite.

Firebase's default store is Firestore, which suits a phone client that needs offline sync and is the wrong store when the data is relational.

Convex stores documents and exposes TypeScript functions, so you are buying reactivity rather than portable SQL. Appwrite Cloud's default is TablesDB, while the self-hosted 1.9.6 install defaults to MongoDB, so the cloud project and the Docker box are not the same database.

Hosted and self-hosted are different products even when the logo matches.

Supabase's Docker setup and the CLI are free to run yourself, which is the exit when the pause or the organization bill is the objection. Appwrite ships 1.9.6 as a Docker installer and asks for 2 CPU cores and 4 GB of RAM, so self-hosting has a hardware floor.

Convex offers self-hosting if the team will operate the box to leave the per-developer meter.

PocketBase is the open-source binary, version 0.40.4, with no cloud SKU, so the cost is the machine and the person who patches it.

How we compared: we read nine vendors' own pricing and product pages, in USD, and ranked them for someone shipping an app rather than buying an enterprise low-code suite.

The 109-tool Toolradar category is the sample around that shortlist. This is not a load-test report.

## Why it matters

The invoice breaks in two places developers notice late. Free projects pause on Supabase, Appwrite, and Nhost after a week idle, so a demo is asleep when a teammate opens it, fine for a side project and fatal for shared staging.

A five-person [Convex](https://toolradar.com/tools/convex) team pays the Professional rate five times, while a five-person Supabase org pays the org rate once and then compute per project.

Lock-in is the schema, not the logo: a Supabase project can leave through SQL you already know, so the hosted plan stays a rental.

A Firestore project leaves through an export and a rewrite, which is what you pay when you change stores.

Amplify leaves through Cognito, Lambda, and whatever store you attached, a fair cost when the company already lives in that account and an expensive one when it does not.

If the frontend is still undecided, the serverless platforms guide covers where the UI is deployed, and the Vercel review is the common pair with a hosted Postgres backend.

When you want to drop the wrapper and keep the rows, the Postgres hosting guide is the next page, and GraphQL tools is the page when Hasura is the actual question. Stack changes like these show up in the Devshot brief.

## Key features to look for

The database you can leave
Postgres you can dump, a document store, or SQLite in one process: the exit cost is the feature that decides the buy.

What the free project does when idle
Several free tiers pause after a week, which a side project survives and a shared staging database does not.

The billing unit
Per organization, per project, per developer, or per peak minute of traffic, because headcount and traffic move different bills.

Auth and files in the same project
Row-level policies, GraphQL permissions, or security rules, depending on whether authorization lives next to the rows.

A place for server code
Edge functions, Cloud Functions, or a visual workflow, and Spark Firebase cannot run functions until you turn billing on.

Who operates the disk
Managed backups and a pause button, or a binary on a VPS you patch, which is the PocketBase choice.

## Pricing

Four hosted products cluster at one entry sticker, and then the meters diverge.

Supabase bills the organization and then compute per project: Micro is $10/mo, and Pro credits cover one of those, so an extra project opens a new line.

Appwrite bills Pro per project and prices a dedicated Micro database on that same compute line, with monthly credits beside it, so each app is its own subscription.

Past Pro's included 2 TB, extra bandwidth is billed per 100 GB, and storage past 150 GB is $2.8 per 100 GB, which is how file traffic leaves the flat subscription.

Firebase and Amplify publish no platform subscription, which keeps a quiet app cheap and gives a busy one no floor.

Blaze adds $0.40 per million function calls after the included 2 million, hosting transfer at $0.15/GB after 360 MB a day, and Realtime Database storage at $5/GB after 1 GB.

Amplify data transfer is $0.15/GB after 15 GB, CDN storage is $0.023/GB after 5 GB, and the published five-developer example is $8.08.

New AWS accounts can apply up to $200 in Free Tier credits, which delays that meter until the credits are gone.

Convex Professional is per developer, not per org, and the Business floor is $2,500/mo, so headcount moves the bill before traffic does.

Nhost Team and Supabase Team are each $599/mo when you need the compliance line. Xano Pro is $224/mo on annual billing, with a 99.99% SLA and 10 seats, the plan you reach when Essential runs out of seats.

Backendless Scale Variable adds daily tiers up to $230 for 7,001 to 10,000 requests per minute, so a spike bills that day rather than at renewal.

PocketBase has nothing to put on an invoice until you rent the machine, which is a VPS question, not a BaaS price.

Overage is where the entry plan stops being the bill. Supabase meters disk, egress, monthly active users, branch hours, and point-in-time recovery as separate lines.

Nhost's point-in-time recovery starts at $100 for 7 days, the same add-on shape as Supabase.

Appwrite bills extra executions per million past the Pro allowance, so a busy function can outrun the subscription. The rates are in the table, verified on the vendor pricing pages in September 2026.

Plan | Price | Best for |

Supabase Free | Free | 50,000 MAU, 500 MB database, pauses after 1 week, 2 active projects |

Supabase Micro compute | $10/mo | Shared CPU, 1 GB RAM; Pro credits cover one instance |

Supabase extra MAU | $0.00325 each | After 100,000 monthly active users on Pro |

Supabase disk | $0.125/GB | After 8 GB per project on Pro |

Supabase egress | $0.09/GB | After 250 GB on Pro |

Supabase branching | $0.01344 per branch-hour | Pro and above |

Supabase PITR | $100 per 7 days | Point-in-time recovery, per project |

Supabase Team | $599/mo | SOC 2, ISO 27001, 14-day backups |

Firebase Spark | Free | 1 GiB Firestore, 50,000 reads/day, no Cloud Functions |

Firebase Cloud Functions | $0.40 per million | After 2 million invocations on Blaze |

Firebase SQL Connect | From $9.37/mo | Cloud SQL for PostgreSQL after a 3-month trial |

Firebase Hosting transfer | $0.15/GB | After 360 MB/day on Blaze |

Firebase Realtime Database storage | $5/GB | After 1 GB stored on Blaze |

Firebase Remote Config | $0.06 per 10,000 | Above 100,000 fetches a day, in effect since 1 September 2026 |

Appwrite Free | Free | 5 GB bandwidth, 2 GB storage, 750,000 executions, 75,000 MAU |

Appwrite extra bandwidth | $15 per 100 GB | On Pro, past the included 2 TB |

Appwrite extra storage | $2.8 per 100 GB | Past 150 GB on Pro |

Appwrite extra executions | $2 per million | Past 3.5 million on Pro |

Appwrite dedicated Micro | $10/mo | 1 GB RAM, plus monthly database compute credits |

Convex free calls | $2.20 per extra million | After 1 million function calls |

Convex Professional calls | $2 per extra million | After 25 million function calls |

Convex database storage | $0.22/GB | Free plan, after 0.5 GB |

Convex Business | $2,500/mo minimum | 50+ developers, SAML, service SLAs |

Amplify standard build | $0.01 per minute | After 1,000 build minutes a month |

Amplify large build | $0.025 per minute | 16 GB memory, 8 vCPUs, no free minutes |

Amplify data transfer | $0.15/GB | After 15 GB a month |

Amplify CDN storage | $0.023/GB | After 5 GB stored |

Amplify SSR requests | $0.30 per million | After 500,000 requests a month |

Amplify WAF | $15/mo per app | Plus the firewall's own usage |

Amplify published example | $8.08/mo | AWS's 5-developer, 300-DAU build plus hosting example |

Nhost database overage | $0.20/GB | After 10 GB on Pro |

Nhost storage overage | $0.05/GB | After 50 GB on Pro |

Nhost egress | $0.10/GB | After 50 GB on Pro |

Nhost function time | $0.18 per GB-hour | After 10 GB-hours on Pro |

Nhost extra functions | $5 per 50 | After 50 functions on Pro |

Nhost dedicated vCPU | $50 per vCPU/mo | $0.0012 per vCPU-minute, 2 GB RAM per vCPU |

Nhost custom domain | $10 per project/mo | Add-on, including on paid plans |

Nhost point-in-time recovery | $100 for 7 days | Starts at that retention window |

Nhost Team | $599/mo | SOC 2 Type II, support SLA |

Xano Essential | $85/mo billed annually | 5 seats, 3 workspaces, no platform rate limit |

Xano Pro | $224/mo billed annually | 10 seats, 5 workspaces, 99.99% SLA |

Backendless Scale | From $15/mo per app | Variable tier 1 is $0.50/day at 0 to 5 requests per minute |

Backendless top variable tier | $230 per day | 7,001 to 10,000 requests per minute |

Backendless extra file storage | $20/mo | Another 10 GB |

Backendless longer Cloud Code | $30/mo | Raises the 5-second cap to 20 seconds |

Backendless security pack | $399/mo | Panic mode, audit log, HIPAA tools |

PocketBase | No hosted price | Open-source binary, v0.40.4, SQLite included |

Mistakes to avoid

×Treating a paused free project as staging means the database is down when a teammate opens the link, because Supabase, Appwrite, and Nhost pause idle free projects after a week.

×Comparing Convex's developer price with Supabase's organization price as if they were the same headcount undercounts one side: five developers on Professional is five seats, while five developers on Supabase Pro is one org and then a compute line per project.

×Turning on Firebase Blaze for a function and forgetting phone auth and Remote Config. SMS is metered, and Remote Config above 100,000 fetches a day has been billed since 1 September 2026.

×Assuming Appwrite Cloud and a self-hosted 1.9 box share a database. The install guide defaults new self-hosted instances to MongoDB. Dedicated cloud databases are a separate choice, including Postgres.

Expert tips

→Write the billing unit on the shortlist before the feature grid: org, project, developer, or peak requests per minute, because the feature lists overlap and the invoices do not.

→If the data has to remain SQL you can host elsewhere, start with Supabase or Nhost and keep policies in the database. Dropping the wrapper later is a hosting move, not a rewrite of the rows.

→Run PocketBase only where someone already patches a VM. Pair it with a host you operate, and keep the binary's data directory in the backup you already trust.

## The bottom line

Start on Supabase when the app is a web product and you want Postgres, auth, and files on one bill. Move to Firebase when the client is mobile and offline sync is the requirement, and accept that functions require Blaze.

Pick Appwrite when a Docker install on 4 GB of RAM is the exit you want written down, and Convex when the team will live in TypeScript and will pay per developer.

Amplify belongs to teams whose account and compliance review are already AWS.

Nhost belongs to teams that want Hasura in front of Postgres, and Xano belongs to teams that will maintain logic in a visual builder and will pay the annual Essential rate.

Backendless belongs where seats should be free and API peaks should set the bill, and PocketBase belongs on a machine you already run.

We re-read these pricing pages for the developer shortlist, separate from the 109-tool category ranking. Cite this: Devshot, "Best Backend-as-a-Service Platforms in 2026", September 2026.

## Frequently asked questions

What is the best backend-as-a-service in 2026?

Supabase, for a web app that should sit on Postgres. Pro is the organization plan, the free project pauses after a week, and you can self-host the Docker setup. Choose Firebase when the app is mobile and offline sync matters, Appwrite when you want the 1.9.6 Docker install as the exit, and Convex when the backend should be TypeScript functions billed per developer.

How much does a backend-as-a-service cost?

Hosted entry plans from Supabase, Appwrite, and Nhost are organization or project subscriptions, and Convex Professional is per developer, so the same kind of entry plan is not the same invoice. Xano Essential is $85/mo billed annually, and Xano Pro is $224/mo billed annually. Supabase Team and Nhost Team are $599/mo when you need the compliance line. Firebase Blaze and AWS Amplify have no platform subscription: Amplify's own five-developer example is $8.08, and Backendless Scale is priced per app rather than per seat. PocketBase publishes no hosted price because you run the binary. Figures were verified on vendor pricing pages in September 2026.

Is there a free backend-as-a-service?

Yes, Supabase, Firebase Spark, Appwrite, Convex, Nhost Starter, Xano, and Backendless each publish a free plan, and the limit is what that plan will actually run. Supabase, Appwrite, and Nhost pause an idle free project after a week, so the free tier is a demo and not shared staging. Firebase Spark includes Firestore and Auth but not Cloud Functions, so a server-side job forces a move to Blaze. Xano's free plan allows 10 requests per 20 seconds and watermarks images. PocketBase is free because you run it, not because a vendor hosts a free tier.

Supabase vs Firebase: which should a developer pick?

Supabase when the data is relational and you want to leave with SQL. Firebase when the clients are iOS, Android, or Flutter and offline sync is part of the product. Firebase SQL Connect can add Cloud SQL from $9.37/mo after the trial, which narrows the SQL gap without making Firestore a Postgres database. The longer comparison is the Supabase vs Firebase note, and each product has a Supabase review and a Firebase review.

Is Appwrite a better self-hosted backend than Supabase?

Appwrite's install is one Docker flow, and the current self-hosted release is 1.9.6, with a 2-core and 4 GB RAM floor. New installs default to MongoDB, so skip it when the rows must stay SQL. Supabase self-host is also Docker, and the database is Postgres, which is why most web teams still start on the hosted Pro plan instead. Pick Appwrite self-host for the smaller install, and Supabase when the rows need to stay SQL.

When is AWS Amplify the right backend?

When the account, the VPC, and the compliance review are already AWS. Amplify does not charge per seat, standard builds are $0.01 per minute after 1,000 minutes, and a Web Application Firewall is a separate monthly line per app. You still pay Cognito, Lambda, and the database on their own pages. If the team is not already in that account, Supabase or Appwrite is a smaller surface to operate.

Can I move off a BaaS later?

Postgres-backed projects move with a dump: Supabase, Nhost, and Xano are in that group, with the caveat that Hasura permissions and Xano workflows do not travel inside the dump, so the rows move and the logic often does not. PocketBase moves as a SQLite file plus your hooks, enough when one process owns the data. Firebase, Convex, and Amplify move as an export and a rewrite, which is the cost of leaving those stores. Appwrite self-hosted 1.9 is MongoDB or MariaDB, so the exit is that database's tools, not pg_dump, unless you chose a dedicated Postgres database on the cloud plan.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026
- [Firebase pricing](https://firebase.google.com/pricing)checked Sep 2026
- [Appwrite pricing](https://appwrite.io/pricing)checked Sep 2026

Related guides

Supabase ReviewFirebase ReviewPostgres HostingServerless PlatformsGraphql ToolsSelf Hosting ToolsSupabase vs FirebaseVercel ReviewDigitalocean ReviewTransactional Email ApisAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best VPS Hosting for Developers in 2026

URL: https://devshot.co/reviews/best-vps-hosting-for-developers
Type: review
Published: 2026-09-28
Updated: 2026-09-28
Summary: The best VPS hosting for developers in 2026, priced in USD: DigitalOcean, Vultr, Linode, Hetzner, UpCloud, OVHcloud, Lightsail, and Hostinger…

Expert Guide

## The Best VPS Hosting for Developers in 2026

The sticker on a 1 GB VM is the small number: IPv4, transfer, backups, and a box you forgot to destroy are the invoice.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

DigitalOcean is the developer VPS to buy in 2026. The 1 GiB Droplet is $6.00 a month, overage is $0.01 per GiB, and a powered-off Droplet bills until you destroy it.

Vultr's 1 GB plan is $5.00 a month for a city, Akamai shared CPU starts at $5 a month for egress, and Hetzner's US CPX11 is $20.49 a month before IPv4.

UpCloud's Starter plan is $3.5 a month with egress listed at zero under its Fair Transfer Policy, which suits a quiet project on a small disk. OVHcloud's VPS-1 starts at $4.54 a month on a 12-month upfront link, with 4 GB of RAM, so the low sticker is a year of commitment.

Key facts8 tools compared: DigitalOcean, Vultr, Linode, Hetzner, UpCloud, OVHcloud, Amazon Lightsail…

- Updated: September 28, 2026
- Top pick: DigitalOcean (best for: Developers who want a monthly cap and an API, and who will destroy idle boxes)
- Top pick price as of September 28, 2026: DigitalOcean: From $6.00/mo (1 GiB Basic, 1,000 GiB transfer); per-second, 672-hour cap

- 8 tools compared: DigitalOcean, Vultr, Linode, Hetzner, UpCloud, OVHcloud, Amazon Lightsail, Hostinger
- Vultr (best for: Apps that need a specific city and will skip the IPv6-only row): From $5.00/mo (1 vCPU, 1 GB, 1 TB); IPv6-only plan is $2.50/mo
- Linode (best for: Workloads that feel egress before they feel the VM price): From $5/mo (shared CPU); extra egress $0.005/GB in most regions
- Hetzner (best for: EU traffic budgets, after you price the US row and IPv4): US CPX11 $20.49/mo excl. IPv4; primary IPv4 is $0.60/mo

Buy [DigitalOcean](https://www.digitalocean.com/pricing/droplets) when you want a Linux VM, a monthly cap, and docs that match the dashboard.

The 1 GiB Basic plan is the first size that holds a small API, a reverse proxy, and a log, so it is the row to leave running.

The cheaper 512 MiB row is on the same page, and it is the wrong default for anything you intend to leave running.

Toolradar data: the [VPS and cloud infrastructure statistics](https://toolradar.com/statistics/vps-servers) track 5 published tools as of September 24, 2026, out of 10,680 tools in the directory.

This guide also prices [Vultr](https://toolradar.com/tools/vultr), [Linode](https://toolradar.com/tools/linode), [Hetzner](https://toolradar.com/tools/hetzner), and [AWS](https://toolradar.com/tools/aws) Lightsail, which have their own tool pages and sit outside that five.

A host that bills a stopped machine, sells IPv4 as a second line, or counts inbound traffic against the allowance will beat a lower sticker once the box has been up for a month.

That running meter is what decides the purchase for a box you leave up.

Prices below were verified on each vendor's pricing page in September 2026.

## Top Picks

Based on features, real-world fit, and value for money.

Best VPS Hosting for Developers in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[DigitalOcean](https://toolradar.com/tools/digitalocean) | From $6.00/mo (1 GiB Basic, 1,000 GiB transfer); per-second, 672-hour cap | Developers who want a monthly cap and an API, and who will destroy idle boxes |

[Vultr](https://toolradar.com/tools/vultr) | From $5.00/mo (1 vCPU, 1 GB, 1 TB); IPv6-only plan is $2.50/mo | Apps that need a specific city and will skip the IPv6-only row |

[Linode](https://toolradar.com/tools/linode) | From $5/mo (shared CPU); extra egress $0.005/GB in most regions | Workloads that feel egress before they feel the VM price |

[Hetzner](https://toolradar.com/tools/hetzner) | US CPX11 $20.49/mo excl. IPv4; primary IPv4 is $0.60/mo | EU traffic budgets, after you price the US row and IPv4 |

UpCloud | From $3.5/mo (Starter, 1 GB, 10 GB disk); Premium 1 GB is $6/mo | Side projects that want the bandwidth line at zero on a small disk |

OVHcloud | From $4.54/mo (VPS-1, 12-month upfront); 4 GB RAM, tax excluded | A 4 GB box when you will pay 12 months upfront |

Amazon Lightsail | From $3.50/mo (Linux Nano, IPv6-only); public IPv4 is a separate bundle | A VM that should sit in the same AWS account as the rest of the stack |

Hostinger | KVM 1 shows $6.49/mo, then renews at $11.99/mo for 2 years | A panel-managed KVM box if you budget the renewal ahead of the promo |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### DigitalOcean

Top Pick

Best for: Developers who want a monthly cap and an API, and who will destroy idle boxes

PricingFrom $6.00/mo (1 GiB Basic, 1,000 GiB transfer); per-second, 672-hour cap

+Bundled Basic plans cap at 672 hours, so a full month of the 1 GiB Droplet cannot exceed the card you can put in a budget

+Outbound transfer pools across the team, inbound transfer is free, and overage is a flat per-GiB rate, so uploads onto the box do not spend the shared pool

+Snapshots are $0.06/GB per month, and weekly backups are 20% of the Droplet price, so a restore habit is a standing line

−A powered-off Droplet keeps billing until you destroy it, because the reservation stays on the hypervisor

−The 512 MiB plan is $4.00 a month with 10 GiB of disk, which is tight once a container and a log share it

Visit DigitalOcean →

2

### Vultr

Best for: Apps that need a specific city and will skip the IPv6-only row

PricingFrom $5.00/mo (1 vCPU, 1 GB, 1 TB); IPv6-only plan is $2.50/mo

+The 1 GB regular plan includes 25 GB of disk and 1 TB of transfer, billed hourly up to 672 hours, so a full month stops at the card

+Automatic backups add 20% to the base fee, and stored snapshots are $0.05/GB per month, so the backup habit is a standing extra

+Bandwidth past the quota is $0.01 per GB, and the minimum charge for a server is one hour, so a short test still pays for that hour

−The $2.50 plan is IPv6 only, so a name that needs an A record starts on the next row

−Stopping an instance does not stop billing, so you destroy it, and some locations also price off the public grid

Visit Vultr →

3

### Linode

Best for: Workloads that feel egress before they feel the VM price

PricingFrom $5/mo (shared CPU); extra egress $0.005/GB in most regions

+The compute page starts shared CPU at the figure on the card, with 1 TB of transfer on the 1 GB size, enough for a small API and tight for files

+Extra egress is $0.005/GB in the majority of regions, below the flat overage DigitalOcean and Vultr print, so file hosting is the reason to pick it

+Dedicated CPU starts at $36/mo and high-memory plans at $60/mo, on the same compute page, a published step off shared cores

−Shared cores can contend with neighbors, and the docs say sustained use should stay under 80%, so move a CPU-bound job off shared cores

−The Europe price sheet and the Jakarta sheet do not match, so quote the region you will deploy or the budget fails at checkout

Visit Linode →

4

### Hetzner

Best for: EU traffic budgets, after you price the US row and IPv4

PricingUS CPX11 $20.49/mo excl. IPv4; primary IPv4 is $0.60/mo

+US CPX11 is 2 AMD vCPUs, 2 GB of RAM, and 40 GB NVMe, with at least 1 TB of included traffic, a smaller pipe than the EU rows

+EU locations include at least 20 TB, which is why a file host stays in Europe, and US CPX21, 3 vCPUs and 4 GB, is $37.49/mo excluding IPv4

+Primary IPv6 is free, while a primary IPv4 is its own line, and the server price excludes that line and VAT, so the VM number is not the invoice

−CX23 shows as not available, while the June 15, 2026 sheet still lists it at $6.49/mo excluding IPv4 in Germany and Finland, so that cheap row is not a new order

−A server bills until you delete it, including time you are not using it, so a paused project still needs a delete

Visit Hetzner →

5

### UpCloud

Best for: Side projects that want the bandwidth line at zero on a small disk

PricingFrom $3.5/mo (Starter, 1 GB, 10 GB disk); Premium 1 GB is $6/mo

+Starter includes one IPv4 address, bills hourly, and caps the month at 28 days, taxes excluded, so a short test stays hourly

+The 1 GB Starter allows 5 concurrent deployments, on previous-gen AMD CPUs, with a 99.99% SLA, which fits a handful of small apps

+Premium starts at the $6/mo 1 GB row with MaxIOPS storage, 1,000 Mbps, and a 99.999% SLA, so the jump is disk speed and uptime

−Starter and Premium bill while shut down, while Cloud Native, from $14/mo for 4 GB, bills only while powered on, which is the plan if the box sleeps

−The 1 GB Starter disk is 10 GB, which fills once an image and a log share it, and an extra public IPv4 is $3.85/mo

Visit UpCloud →

6

### OVHcloud

Best for: A 4 GB box when you will pay 12 months upfront

PricingFrom $4.54/mo (VPS-1, 12-month upfront); 4 GB RAM, tax excluded

+VPS-1 is 2 vCores, 4 GB RAM, 40 GB NVMe, a backup of the previous 24 hours, unlimited traffic, and 500 Mbps, so you buy memory and a one-day restore

+VPS-2 is $8.50/mo for 4 vCores, 8 GB, 75 GB NVMe, and 1 Gbps, on the same upfront link, the step when 4 GB runs out

+VPS-3 is $12.32/mo for 6 vCores, 12 GB, and 100 GB NVMe, and the card says installation fees are free, so the ladder is public

−Starting prices open a configurator with upfront12, so this is a 12-month upfront order and a poor fit for a one-month test

−The included backup is the previous 24 hours, prices are tax-excluded, and Asia-Pacific sites on the cloud VPS page cap monthly traffic then drop to 10 Mbps

Visit OVHcloud →

7

### Amazon Lightsail

Best for: A VM that should sit in the same AWS account as the rest of the stack

PricingFrom $3.50/mo (Linux Nano, IPv6-only); public IPv4 is a separate bundle

+Linux Nano with public IPv4 is 2 vCPUs, 0.5 GB RAM, 20 GB of disk, and 1 TB of transfer, a lot of CPU for a tight container

+The IPv6-only Nano is the lower bundle and will not serve an A record, while Micro with public IPv4 is $7.00/mo for 1 GB and 2 TB

+Billing is hourly up to the monthly plan cost, so a short test does not pay the whole month

−Inbound and outbound both count, and US East (N. Virginia) lists excess outbound at $0.09/GB, so uploads spend the same allowance you serve

−Mumbai, Sydney, Jakarta, Malaysia, Hong Kong, and Sao Paulo include half the transfer of other regions, so those cities run out sooner

Visit Amazon Lightsail →

8

### Hostinger

Best for: A panel-managed KVM box if you budget the renewal ahead of the promo

PricingKVM 1 shows $6.49/mo, then renews at $11.99/mo for 2 years

+KVM 1 is 1 vCPU, 4 GB RAM, 50 GB NVMe, and 4 TB of bandwidth, marked 67% off a $19.49 figure, generous only while the discount lasts

+KVM 2 is $8.99/mo on the card for 2 vCPUs, 8 GB, 100 GB, and 8 TB, renewing at $14.99/mo, which is the bill if you stay

+KVM 8 is $25.99/mo for 8 vCPUs, 32 GB, 400 GB, and 32 TB, so the ladder is printed on the card

−The low figure is a markdown: KVM 1 renews at $11.99/mo for 2 years, and KVM 4 renews at $28.99/mo, so copying the card leaves the budget short

−This is a panel VPS, and it is the wrong pick when the workflow is Terraform against a cloud API

Visit Hostinger →

## What it is

A VPS is a virtual machine with its own operating system, a public address, and a bill that does not scale to zero when the process is idle.

You SSH in, you patch it, and you pay for the reservation whether the CPU is busy, which suits a process that has to stay up.

That is a different product from the serverless platforms that bill per request, and from a managed database host.

If the only thing on the box is Postgres, the Postgres hosting comparison is the page that prices the database.

A VPS is the right shape when you want Docker, a queue, and a reverse proxy on one private network you control.

## Why it matters

Outbound transfer is where a quiet side project becomes an ugly month. DigitalOcean and Vultr both print a per-gigabyte overage once the included pool is gone, so a busy month costs more than the VM.

Lightsail's US East (N.

Virginia) rate on the same kind of excess is $0.09/GB, and both directions count toward the allowance before that excess starts, so a download onto the box spends the pool too.

Akamai's compute page puts extra egress at $0.005/GB in most regions, which is the line to read if the app serves files.

IPv4 is the second surprise: Hetzner bills a primary IPv4 separately from the server, so the sheet price is missing the address your DNS needs.

Vultr's cheapest regular row is IPv6 only, so a hostname that still needs an A record cannot use it.

Lightsail sells an IPv6-only Nano and a public-IPv4 Nano as two bundles, while UpCloud and OVHcloud include an IPv4 address on the VPS card.

Devshot data: 27,887 developers are on this list, with a 30% average open rate, from the audience file updated September 20, 2026.

Price-sheet changes land in the Devshot brief before the next pass on this page.

How we compared: eight hosts with a USD figure a developer can order, read on September 24, 2026, ranked on the VM, the address, the transfer, and whether a stopped machine still bills.

We did not provision servers for this piece.

A PaaS such as the one in the Vercel review is a different purchase, and so is a cluster from the Kubernetes tools guide.

## Key features to look for

Monthly cap versus hourly meter
DigitalOcean bundled Droplets, Vultr regular cloud compute, Lightsail, and Hetzner stop at a monthly cap, so a full month cannot run past the card. Vultr's VX1 and GPU plans do not use that hourly cap, and UpCloud caps the month at 28 days.

Whether powering off stops the bill
On DigitalOcean, Vultr, Hetzner, and UpCloud Starter or Premium, a stopped VM keeps the reservation and the charge, so you delete it to end the meter. Cloud Native on UpCloud is the exception that bills only while powered on.

IPv4 as its own line
Some entry plans are IPv6 only, and Hetzner prices a primary IPv4 apart from the server. If the app still publishes an A record, price the address before you compare RAM, or the cheapest row will not resolve the name.

Transfer pool and overage
Included outbound transfer is large on the EU Hetzner rows and metered on DigitalOcean, Vultr, and Lightsail, so a file host in Europe is a different bill. Inbound is free on DigitalOcean and counts on Lightsail, and that difference decides a box people upload into.

Backups and snapshots
Weekly backups are often a percentage of the VM, and snapshots are a per-gigabyte store, so a kept restore point is another invoice line. OVHcloud includes a backup of the previous 24 hours on the VPS card, a one-day window.

What you operate
These are unmanaged machines, so you own SSH keys, updates, and the firewall, and a panel does not remove that work or replace a pipeline.

## Pricing

Every dollar in the table was read on the vendor's own page on September 24, 2026.

DigitalOcean, Vultr, Akamai, UpCloud, OVHcloud, Lightsail, and Hostinger publish USD on those pages.

Hetzner's USD column is on the June 15, 2026 price adjustment, excluding VAT and excluding primary IPv4, so a US checkout still adds tax and the address.

The machine you can leave running is rarely the smallest row.

A 512 MiB Droplet and an IPv6-only Nano are real SKUs, and they are the wrong SKU for an app with an A record and a container, because the disk or the address runs out first.

Hetzner's orderable US entry is CPX11, and the cost-optimized CX line is marked not available for a new order.

Weekly backups sit on top of a Droplet or a Vultr instance as a percentage of that VM, so a restore habit changes the month.

A second UpCloud IPv4 is $3.85/mo, and Windows Server Standard on UpCloud is $14.40/mo per CPU core, which takes a Windows box out of a Linux budget.

Vultr DDoS protection is an extra $10/mo, and Scaleway's virtual-instance page publishes euro prices and no USD list, so it stays off this table.

Our addition: a US CPX11 at the sheet price plus one primary IPv4 is $21.09 before tax.

A 1 GiB Droplet with weekly backups is $7.20 for the month, before snapshots and before transfer overage.

Plan | Price | Best for |

DigitalOcean Basic 512 MiB | $4.00/mo | 1 shared vCPU, 10 GiB SSD, 500 GiB transfer |

DigitalOcean Basic 1 GiB | $6.00/mo | 1 vCPU, 25 GiB SSD, 1,000 GiB transfer, $0.00893/hr |

DigitalOcean Basic 2 GiB | $12.00/mo | 1 vCPU, 50 GiB SSD, 2,000 GiB transfer |

DigitalOcean Basic 2 GiB / 2 vCPU | $18.00/mo | 60 GiB SSD, 3,000 GiB transfer |

DigitalOcean Basic 4 GiB | $24.00/mo | 2 vCPUs, 80 GiB SSD, 4,000 GiB transfer |

DigitalOcean CPU-Optimized 4 GiB | $42.00/mo | 2 dedicated vCPUs, 25 GiB SSD |

DigitalOcean CPU-Optimized 16 GiB | $168.00/mo | 8 dedicated vCPUs, 100 GiB SSD |

DigitalOcean General Purpose 8 GiB | $63.00/mo | 2 dedicated vCPUs, 25 GiB SSD |

DigitalOcean Memory-Optimized 16 GiB | $84.00/mo | 2 dedicated vCPUs, 50 GiB NVMe |

DigitalOcean Storage-Optimized 16 GiB | $131.00/mo | 2 vCPUs, 300 GiB NVMe |

DigitalOcean snapshots | $0.06/GB/mo | Stored snapshots, separate from backup plans |

DigitalOcean weekly backups | 20% of Droplet | Daily backups are 30% of the Droplet price |

Vultr regular, IPv6 only | $2.50/mo | 1 vCPU, 0.5 GB, 10 GB disk, 0.50 TB, $0.004/hr |

Vultr regular 0.5 GB with IPv4 | $3.50/mo | 1 vCPU, 10 GB disk, 0.50 TB transfer |

Vultr regular 1 GB | $5.00/mo | 1 vCPU, 25 GB disk, 1 TB, $0.007/hr |

Vultr regular 2 GB | $10.00/mo | 1 vCPU, 55 GB disk, 2 TB, $0.015/hr |

Vultr regular 2 vCPU / 2 GB | $15.00/mo | 65 GB disk, 3 TB transfer |

Vultr regular 2 vCPU / 4 GB | $20.00/mo | 80 GB disk, 3 TB transfer |

Vultr regular 4 vCPU / 8 GB | $40.00/mo | 160 GB disk, 4 TB transfer |

Vultr high frequency 1 GB | $6.00/mo | 1 vCPU, 32 GB NVMe, 1 TB, $0.009/hr |

Vultr snapshots | $0.05/GB/mo | Stored snapshots. Automatic backups add 20% |

Vultr bandwidth overage | $0.01/GB | Past the included quota |

Vultr DDoS protection | $10/mo | Adds 10 Gbps of mitigation per instance |

Akamai shared CPU | $5/mo | Starting price on the compute page. 1 TB on the 1 GB size |

Akamai egress overage | $0.005/GB | Majority of regions, on the compute page |

Akamai dedicated CPU | $36/mo | Starting price on the compute page |

Akamai high memory | $60/mo | Starting price on the compute page |

Akamai Europe Nanode hourly | $0.0075/hr | 1 vCPU, 25 GB, 1 TB on the Europe sheet |

Akamai Europe Linode 2 GB | $12.00/mo | 1 vCPU, 50 GB, 2 TB. Europe sheet, not Jakarta |

Akamai Europe Linode 4 GB | $24.00/mo | 2 vCPUs, 80 GB, 4 TB, on the Europe sheet |

Hetzner US CPX11 | $20.49/mo | Excl. IPv4 and VAT. $0.0328/hr. 2 vCPU, 2 GB, 40 GB |

Hetzner primary IPv4 | $0.60/mo | Excl. VAT. Primary IPv6 is free |

Hetzner US CPX21 | $37.49/mo | Excl. IPv4 and VAT. 3 AMD vCPUs, 4 GB |

Hetzner EU CPX22 | $22.99/mo | Germany and Finland, excl. IPv4. 2 vCPU, 4 GB |

Hetzner CX23, listed | $6.49/mo | Excl. IPv4. Marked not available for a new order |

UpCloud Starter 1 GB | $3.5/mo | 1 vCPU, 10 GB disk, 5 deployments max |

UpCloud Starter 2 GB | $7/mo | 1 vCPU, 20 GB disk, 250 Mbit/s |

UpCloud Starter 2 GB / 2 vCPU | $10/mo | 30 GB disk |

UpCloud Starter 4 GB / 2 vCPU | $14/mo | 30 GB disk, 500 Mbit/s |

UpCloud Premium 1 GB | $6/mo | 25 GB MaxIOPS, 99.999% SLA |

UpCloud Premium 4 GB | $26/mo | 2 vCPUs, 50 GB MaxIOPS |

UpCloud Cloud Native 4 GB | $14/mo | 1 vCPU. Billed only while powered on |

UpCloud extra IPv4 | $3.85/mo | Additional public address or floating IP |

UpCloud Windows Standard | $14.40/mo per core | License, hourly, monthly figure is a 30-day estimate |

OVHcloud VPS-1 | $4.54/mo | 2 vCores, 4 GB, 40 GB NVMe, upfront12 link, tax excluded |

OVHcloud VPS-2 | $8.50/mo | 4 vCores, 8 GB, 75 GB NVMe, 1 Gbps |

OVHcloud VPS-3 | $12.32/mo | 6 vCores, 12 GB, 100 GB NVMe, 2 Gbps |

OVHcloud VPS-4 | $23.37/mo | 8 vCores, 24 GB, 200 GB NVMe |

Lightsail Linux Nano IPv6-only | $3.50/mo | 2 vCPUs, 0.5 GB, 20 GB, 1 TB |

Lightsail Linux Nano with IPv4 | $5.00/mo | Same shape as the IPv6 Nano, with a public IPv4 |

Lightsail Linux Micro with IPv4 | $7.00/mo | 2 vCPUs, 1 GB, 40 GB, 2 TB |

Lightsail Linux Small with IPv4 | $12.00/mo | 2 vCPUs, 2 GB, 60 GB, 3 TB |

Lightsail Linux Medium with IPv4 | $24.00/mo | 2 vCPUs, 4 GB, 80 GB, 4 TB |

Lightsail Linux Large with IPv4 | $44.00/mo | 2 vCPUs, 8 GB, 160 GB, 5 TB |

Lightsail 2Xlarge with IPv4 | $164.00/mo | 8 vCPUs, 32 GB, 640 GB, 7 TB |

Lightsail Windows Nano with IPv4 | $9.50/mo | 0.5 GB, 30 GB disk, 1 TB |

Lightsail Windows Nano IPv6-only | $8.00/mo | Windows without a public IPv4 |

Lightsail transfer overage, US East (N. Virginia) | $0.09/GB | Excess outbound only, after the allowance |

Hostinger KVM 1 promo | $6.49/mo | 1 vCPU, 4 GB, 50 GB NVMe, 4 TB. Struck price $19.49 |

Hostinger KVM 1 renewal | $11.99/mo | Renews for 2 years |

Hostinger KVM 2 promo | $8.99/mo | 2 vCPUs, 8 GB, 100 GB, 8 TB. Struck price $24.49 |

Hostinger KVM 2 renewal | $14.99/mo | Renews for 2 years |

Hostinger KVM 4 promo | $12.99/mo | 4 vCPUs, 16 GB, 200 GB, 16 TB. Struck price $42.99 |

Hostinger KVM 4 renewal | $28.99/mo | Renews for 2 years |

Hostinger KVM 8 promo | $25.99/mo | 8 vCPUs, 32 GB, 400 GB, 32 TB. Struck price $73.99 |

Hostinger KVM 8 renewal | $49.99/mo | Renews for 2 years |

Mistakes to avoid

×Ordering the IPv6-only row because it is the first price on the grid, then discovering the app has no A record, so the cheap box cannot serve the name.

×Leaving a stopped Droplet, Vultr instance, or Hetzner server in the account, because powering off does not end those bills.

×Comparing Hetzner's old EU bargain with a new US CPX11, because the cost-optimized plans are marked not available and the US regular row is a different price.

×Budgeting Hostinger at the marked-down card and ignoring the renewal printed under it, so the later term is missing from the launch budget.

×Putting production Postgres on the smallest VPS and skipping backups, because a 24-hour OVHcloud backup and a weekly backup billed as a percentage are different restore stories.

Expert tips

→Price one month of the size you will actually run, plus the backup percentage, plus one public IPv4, plus a transfer overage you can imagine. The ranking above flips once those lines are on the same sheet.

→Delete test machines the day the branch merges, because the hourly cap still charges every hour the VM exists, including nights it is powered off.

→If the workload is a static front end or a function, read the <a href="/reviews/best-serverless-platforms">serverless comparison</a> before you rent a VM. If you are about to self-host a stack of containers, the <a href="/reviews/best-self-hosting-tools">self-hosting tools guide</a> is the next list, and the <a href="/reviews/digitalocean-review">DigitalOcean review</a> is the longer note on Droplets, App Platform, and managed databases.

## The bottom line

DigitalOcean is the buy for a developer who will SSH, use an API, and want the month to stop at a printed cap.

Use the 1 GiB Basic plan, turn on the backup percentage you can explain, and destroy boxes you are finished with, because a powered-off machine is still a bill, and the longer write-up is the DigitalOcean review.

Choose Vultr when the city matters and you will skip the IPv6-only row, because that row cannot serve an A record.

Choose Linode on Akamai when egress is the risk.

Choose UpCloud when a zero egress line and a tiny disk fit the job. Choose OVHcloud when 4 GB and unlimited traffic are worth a 12-month upfront order.

Choose Lightsail when the account is already AWS and transfer that counts both ways is acceptable.

Choose Hostinger when you want a panel and you have written the renewal into the budget. Choose Hetzner when the EU traffic allowance is the point and the US CPX11 price still fits.

Cite this: Devshot, "Best VPS Hosting for Developers in 2026", September 2026.

## Frequently asked questions

What is the best VPS hosting for developers in 2026?

DigitalOcean, for a Linux VM with a monthly cap and an API you can script, if you will destroy idle boxes. The 1 GiB Basic Droplet is the size to start on, because the smaller row runs out of disk once a container and a log share it. Vultr is the pick when you care which city the box is in. Linode, now Akamai Cloud, is the pick when egress overage is the expensive line. Hetzner is the pick for a large EU traffic allowance after you read the US CPX11 price, which is no longer the cheap default.

How much does a developer VPS cost in 2026?

Checked in September 2026, UpCloud Starter is $3.5/mo and OVHcloud VPS-1 is $4.54/mo on a 12-month upfront link, so the lowest stickers are a small disk or a year paid ahead. Vultr's 1 GB plan and the Lightsail IPv4 Nano sit in the middle of the table. The Hetzner US row is the CPX11 cap before IPv4. Hostinger prints a promo and a higher renewal on the same card. Weekly backups as a percentage of the VM, and snapshots by the gigabyte, sit on top of several of those stickers.

Is Hetzner still the cheap VPS?

A new US order is no longer the cheap VPS. The June 15, 2026 price sheet lists US CPX11 at $20.49/mo excluding IPv4, and a primary IPv4 at $0.60/mo excluding VAT. Germany and Finland CPX22 is $22.99/mo excluding IPv4, the regular EU row when you need 4 GB there. The cost-optimized entry remains on that sheet and is marked not available for a new order. EU locations include at least 20 TB of traffic, and US locations include at least 1 TB, which is why the EU box is still the traffic play.

Does a stopped VPS still cost money?

Yes on the hosts most developers compare, so a paused project is still an invoice. DigitalOcean bills a powered-off Droplet until you destroy it, and Vultr bills a stopped server until you destroy it, with a one-hour minimum. Hetzner bills a cloud server until you delete it. UpCloud Starter and Premium bill while shut down, while UpCloud Cloud Native bills only while powered on. Lightsail bills the hourly rate up to the monthly cap for as long as the instance exists.

Is there a free developer VPS?

None of these eight publishes a permanent free server, so a side project still needs a card on file. DigitalOcean describes promotional credits as an example, and the other seven show a paid entry on the pages checked in September 2026. A free function tier lives on the serverless platforms page.

DigitalOcean vs Vultr vs Lightsail: which invoice is easier to predict?

DigitalOcean and Vultr cap ordinary VMs at 672 hours and charge a flat per-gigabyte overage after a transfer pool, so the surprise on a full month is the traffic. Lightsail also caps at the monthly bundle, then charges excess outbound, with US East (N. Virginia) at $0.09/GB, and it counts inbound toward the allowance. Vultr's lowest regular row is IPv6 only, and Lightsail splits IPv6-only and public IPv4 into two Nano prices. For a single always-on API, the DigitalOcean cap is the number that is easiest to defend in a budget.

Should I self-host on a VPS or use a platform?

Use a VPS when you want root and a process that stays up. Use a platform when you want previews and someone else to patch the runtime. Software you would install is on the self-hosting tools list. A cluster is the Kubernetes tools decision, a different bill from one VM. Alerts and the pipeline are the monitoring tools and CI/CD tools lists. Managed Postgres is priced on the Postgres hosting page.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [DigitalOcean pricing](https://digitalocean.com/pricing)checked Sep 2026
- [Vultr pricing](https://vultr.com/pricing)checked Sep 2026
- [Linode pricing](https://linode.com)checked Sep 2026
- [Hetzner pricing](https://hetzner.com)checked Sep 2026

Related guides

Digitalocean ReviewPostgres HostingServerless PlatformsSelf Hosting ToolsKubernetes ToolsMonitoring ToolsVercel ReviewCi Cd ToolsCloud Free Tiers Compared (2026): AWS, Google Cloud, Azure, OracleCheap VpsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Best AI App Development Platforms

URL: https://devshot.co/reviews/best-ai-app-development-platforms
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: We ranked Dify, Emergent, and Botpress for building AI apps in 2026, with honest pros, cons, and pricing for developers and teams.

Expert Guide

## Best AI App Development Platforms

For developers choosing a platform to build, ship, and scale AI apps and agents, here are four honest options ranked by real fit.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 tools compared

TL;DR

Dify is the best pick for most developers: it is open source, self-hostable, model-agnostic, and free to start, so you keep full control of your AI apps and agent workflows. If you want to go from a plain-language prompt to a deployed full-stack app fast, Emergent is the stronger choice. Teams building conversational agents that live in WhatsApp, Slack, or a web widget should look hard at Botpress. Snowfire AI fits enterprise business teams building internal decision tools.

Key facts4 tools compared: Dify, Emergent, Botpress, Snowfire AI

- Updated: September 25, 2026
- Top pick: Dify (best for: Developers who want an open-source, self-hostable LLM app platform)
- Top pick price as of September 25, 2026: Dify: From $59/mo (Professional) or $590/yr; free Sandbox, Community Edition free to self-host

- 4 tools compared: Dify, Emergent, Botpress, Snowfire AI
- Emergent (best for: Shipping a full-stack MVP from a prompt, fast): From $17/mo (Standard, billed annually); free tier with 10 credits/mo
- Botpress (best for: Building conversational AI agents across messaging channels): Free (25 conversations, 3 seats); Plus $150/mo and Team $750/mo, billed annually
- Snowfire AI (best for: Enterprise business teams building internal decision tools): Custom quote; no published price list or free plan

AI app development platforms now split into three camps, and picking the wrong one wastes weeks. Some, like Dify, give developers an open framework to build LLM apps, RAG pipelines, and agents with full control. Others, like Emergent, turn a written prompt into deployed code so you ship an MVP in an afternoon.

A third group focuses on one job, such as Botpress for chat agents. The real question is not which tool is best overall, but which matches how much control you want versus how fast you need to move.

## Top Picks

Based on features, real-world fit, and value for money.

Best AI App Development Platforms in 2026: 4 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Dify](https://toolradar.com/tools/dify) | From $59/mo (Professional) or $590/yr; free Sandbox, Community Edition free to self-host | Developers who want an open-source, self-hostable LLM app platform |

[Emergent](https://toolradar.com/tools/emergent) | From $17/mo (Standard, billed annually); free tier with 10 credits/mo | Shipping a full-stack MVP from a prompt, fast |

[Botpress](https://toolradar.com/tools/botpress) | Free (25 conversations, 3 seats); Plus $150/mo and Team $750/mo, billed annually | Building conversational AI agents across messaging channels |

[Snowfire AI](https://toolradar.com/tools/snowfire-ai) | Custom quote; no published price list or free plan | Enterprise business teams building internal decision tools |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Dify

Top Pick

Best for: Developers who want an open-source, self-hostable LLM app platform

PricingFrom $59/mo (Professional) or $590/yr; free Sandbox, Community Edition free to self-host

+Open source and self-hostable, so you keep full control of code, data, and models

+Model-agnostic: swap between OpenAI, Anthropic, and open models freely

+Visual workflow builder plus a strong built-in RAG pipeline

−Credit-based cloud pricing climbs fast once apps get real traffic

−Self-hosting needs real DevOps effort to run and maintain

Visit Dify →

2

### Emergent

Best for: Shipping a full-stack MVP from a prompt, fast

PricingFrom $17/mo (Standard, billed annually); free tier with 10 credits/mo

+Goes from idea to a working, deployed app in minutes, not weeks

+You own and can export the generated code, so no lock-in

+Handles frontend, backend, and deployment in one flow

−Credits burn quickly on complex apps, and the $17 to $167 jump is steep

−Generated code quality varies and needs a developer to review it

Visit Emergent →

3

### Botpress

Best for: Building conversational AI agents across messaging channels

PricingFree (25 conversations, 3 seats); Plus $150/mo and Team $750/mo, billed annually

+Strong mix of visual flow builder and real code and SDK control

+Deploys the same agent across many channels out of the box

+Free plan with 25 conversations and 3 agents to prototype before you pay

−Built for chat agents, not general app or dashboard building

−Extra conversations cost $65 per 100 on Plus, which adds up at scale

Visit Botpress →

4

### Snowfire AI

Best for: Enterprise business teams building internal decision tools

PricingCustom quote; no published price list or free plan

+Focused on business decision-making and executive use cases

+Enterprise-oriented, with an emphasis on adaptive AI

+Low-code approach lets non-engineers build internal tools

−Opaque enterprise pricing, with no published price list or free plan

−Aimed at business users, so developers get less code-level control

Visit Snowfire AI →

## What it is

An AI app development platform gives you the building blocks to create software powered by large language models without wiring every piece from scratch.

That usually means a visual or code-based builder, connectors to models like GPT and Claude, a way to add your own data through retrieval (RAG), and one-click deployment to a web app, API, or messaging channel. Some are full open-source frameworks, others generate the whole app for you from a description.

## Why it matters

The platform you choose decides how far a project goes before it hits a wall. A closed, business-user tool means your engineers lose control of the code, the models, and the hosting. Something too raw means months spent building plumbing instead of features.

Pricing matters too: credit and usage-based models look cheap at demo scale and get expensive in production. Choose well and you ship faster now, without a painful rebuild on another stack six months later.

## Key features to look for

Model flexibilityEssential
Support for multiple LLM providers (OpenAI, Anthropic, open models) so you are not locked to one vendor's pricing or roadmap.

Deployment and hosting optionsEssential
Whether you can self-host, use a managed cloud, or export the code. This decides who controls uptime, data, and long-term costs.

Retrieval and knowledge (RAG)Essential
Built-in ways to feed your own documents and data into the app so answers are grounded in your content, not just the base model.

Visual builder vs code control
A drag-and-drop flow builder speeds up non-experts, but developers need code access and an SDK for custom logic.

Pricing model transparency
Flat, credit, or usage-based billing changes the math at scale. Clear limits beat surprise overages in production.

Integrations and channels
Prebuilt connectors to Slack, WhatsApp, web widgets, and APIs so your app reaches users where they already are.

## Pricing

Dify, Emergent, and Botpress publish paid prices, and Snowfire AI is quote-only. Botpress starts on a free plan capped at 25 conversations, then prices Plus at $150/mo and Team at $750/mo, billed annually, plus a custom Enterprise plan.

The cheapest published paid entry is Emergent Standard at $17/mo, billed annually, above a free tier of 10 credits a month. Dify also offers a free Sandbox and a free self-hosted Community Edition, and its Professional and Team plans cost $59 and $159 a month, or $590 and $1,590 a year.

Costs jump at Emergent Pro, $167/mo billed annually, at Botpress Team, and on the custom Enterprise quotes from Dify, Emergent, and Botpress.

Plan | Price | Best for |

Dify Professional | $59/mo ($590/yr) | Managed cloud for 3 members, 5,000 message credits a month |

Dify Team | $159/mo ($1,590/yr) | Up to 50 members and 10,000 message credits a month |

Dify Enterprise | Custom quote | Custom enterprise agreement |

Emergent Standard | $17/mo, billed annually | Paid plan above the free tier |

Emergent Pro | $167/mo, billed annually | Higher plan, billed annually |

Emergent Enterprise | Custom quote | Custom enterprise agreement |

Botpress Plus | $150/mo, billed annually | 250 conversations a month with $25 of AI usage included |

Botpress Team | $750/mo, billed annually | 1,500 conversations a month, unlimited seats and routing |

Botpress Enterprise | Custom quote | Custom enterprise agreement |

Snowfire AI | Custom quote | Contact sales for enterprise pricing |

Mistakes to avoid

×Picking a closed, business-user platform for a project your engineers will own. You lose control of the code, models, and hosting right when you most need to customize.

×Judging tools by demo-scale pricing. Credit and usage-based plans look cheap with ten test users and get painful once real production traffic arrives.

×Choosing a prompt-to-app generator for a product with heavy custom logic. It ships an MVP fast, then stalls the moment your requirements get non-standard.

Expert tips

→Prototype the same feature in your top two picks before committing. A weekend of real building tells you more than any feature comparison table.

→If data control or vendor lock-in matters to you, favor an open-source, self-hostable option like Dify over a closed, hosted-only platform.

→Model your costs at production scale, not demo scale, before signing. Ask exactly what one credit or AI-usage unit actually buys you.

## The bottom line

For most developers and engineering teams, Dify is the pick: open source, self-hostable, model-agnostic, and free to start, so you keep control as the project grows. If your goal is speed and you want a full-stack app deployed from a written prompt, Emergent gets you there fastest, just budget for code review.

Building a chat agent that lives across WhatsApp, Slack, and the web? Botpress is purpose-built for it. Snowfire AI fits enterprise business teams building internal decision tools, where low-code and a sales-led rollout are fine.

Match the tool to how much control you need versus how fast you must move.

## Frequently asked questions

What is the difference between Dify and a prompt-to-app tool like Emergent?

Dify is a platform you build on: you assemble LLM apps, RAG pipelines, and agents with a visual editor and keep full control, including self-hosting. Emergent generates the whole app from a written description and deploys it for you. Dify suits teams who want ongoing control; Emergent suits shipping an MVP fast. Well-known prompt-to-app alternatives include Lovable and Bolt.new.

Which platform is best for building an AI chatbot or agent?

Botpress is the most focused choice for conversational agents, with a visual flow builder, code access, and one-click deploy to WhatsApp, Slack, and web widgets. Dify can also build agents and chat apps if you want more general flexibility or self-hosting. If you specifically need voice or IVR, Voiceflow is worth a look too.

Are these platforms free, and how does pricing usually work?

Dify, Emergent, and Botpress all have free tiers to start, and Dify's Community Edition is fully open source and free to self-host. Paid plans use credits (Dify, Emergent) or conversation bundles (Botpress), which stay cheap at demo scale but grow with production traffic. Snowfire AI is enterprise-only with custom, sales-led pricing.

Which should I choose if I want to avoid vendor lock-in?

Dify, because the Community Edition is open source and self-hostable, so you control the code, data, and models. Emergent is next best since you own and can export the generated code. Botpress is hosted-first, and Snowfire AI is a closed enterprise platform, so both tie you more closely to the vendor.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Dify pricing](https://dify.ai/pricing)checked Sep 2026
- [Emergent pricing](https://emergent.sh/pricing)checked Sep 2026
- [Botpress pricing](https://botpress.com/pricing)checked Sep 2026
- [Snowfire AI pricing](https://snowfire.ai/pricing)

Related guides

Openrouter AlternativesWeb Scraping ToolsDocker Desktop AlternativesN8n ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best AI Coding Assistants in 2026

URL: https://devshot.co/reviews/best-ai-coding-assistants
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: Claude Code, Cursor, Copilot, Codex, Devin Desktop, Cline, Zed, Amp, and Aider compared on verified 2026 pricing, real dev workflows, and honest tradeoffs.

Expert Guide

## The Best AI Coding Assistants in 2026

For developers weighing autocomplete, agentic, and terminal-native coding tools, ranked on verified 2026 pricing and real dev workflows.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 9 tools compared

TL;DR

Claude Code tested strongest for large cross-file refactors from the terminal, while Cursor is the best all-in-one AI IDE combining autocomplete and agent mode. Teams standardized on GitHub get the deepest fit from GitHub Copilot. For the best free, no-lock-in value, Cline and Aider are fully open source and cost only your model provider's API rate. Pick based on where you actually work: the terminal, an IDE, or a GitHub-native workflow, not on which tool had the loudest launch.

Key facts9 tools compared: Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Devin Desktop, Cline…

- Updated: September 25, 2026
- Top pick: Claude Code (best for: Large cross-file refactors and migrations from the terminal)
- Top pick price as of September 25, 2026: Claude Code: Free (chat only) / Pro $17-20/mo / Max from $100/mo

- 9 tools compared: Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Devin Desktop, Cline, Zed, Amp, Aider
- Cursor (best for: An AI-native IDE combining fast autocomplete and agent mode): Free (Hobby) / Pro $20/mo / Pro+ $60/mo / Teams $40/user/mo
- GitHub Copilot (best for: Teams already standardized on GitHub): Free / Pro $10/mo / Pro+ $39/mo / Business $19/user/mo
- OpenAI Codex (best for: Async tasks you hand off and check on later): Limited on ChatGPT Free and Go ($8/mo); Plus $20/mo; Pro from $100/mo

The label "AI coding assistant" now covers three different tools wearing the same name.

Autocomplete tools predict your next line as you type, agentic tools take a task and edit across many files on their own, and terminal-native tools live in your shell instead of an editor. Most vendors blur these lines in their marketing, so we sorted nine real tools by what they actually do rather than what they claim.

We ran each against a normal week's work: a multi-file rename across a mid-size repo, adding an endpoint and wiring it to the frontend, fixing a flaky test without breaking two others. Pricing below is verified against each vendor's pricing page as of mid-2026, and where a number is not public we say so.

One thing to flag up front: Windsurf no longer exists as a standalone product, since Cognition folded it into Devin Desktop in 2025.

## Top Picks

Based on features, real-world fit, and value for money.

Best AI Coding Assistants in 2026: 9 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Claude Code](https://toolradar.com/tools/claude-code) | Free (chat only) / Pro $17-20/mo / Max from $100/mo | Large cross-file refactors and migrations from the terminal |

[Cursor](https://toolradar.com/tools/cursor) | Free (Hobby) / Pro $20/mo / Pro+ $60/mo / Teams $40/user/mo | An AI-native IDE combining fast autocomplete and agent mode |

[GitHub Copilot](https://toolradar.com/tools/github-copilot) | Free / Pro $10/mo / Pro+ $39/mo / Business $19/user/mo | Teams already standardized on GitHub |

[OpenAI Codex](https://toolradar.com/tools/openai-codex) | Limited on ChatGPT Free and Go ($8/mo); Plus $20/mo; Pro from $100/mo | Async tasks you hand off and check on later |

[Devin Desktop](https://toolradar.com/tools/devin) | Free / Pro $20/mo / Max $200/mo / Teams $40/seat ($80/mo minimum) | Running several agent threads from one IDE |

[Cline](https://toolradar.com/tools/cline) | Free (open source) + your model API costs | Cost control with your own model and no lock-in |

[Zed](https://toolradar.com/tools/zed) | Free / Pro $10/mo / Business $30/seat/mo | Developers who want the fastest editor with light AI |

Amp | Free Hobby tier (bring your own keys); Individual $20/mo; extra credits at provider API rates | Developers who want the tool to pick the best model per task |

[Aider](https://toolradar.com/tools/aider) | Free (open source) + your model API costs | Terminal-first developers who treat git history as truth |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Claude Code

Top Pick

Best for: Large cross-file refactors and migrations from the terminal

PricingFree (chat only) / Pro $17-20/mo / Max from $100/mo

+Strongest tested for large cross-file refactors and migrations

+Subagents delegate subtasks without polluting the main context

+CLI-first, with VS Code and JetBrains extensions wrapping the same agent

−No meaningful free agent tier, chat only on the free plan

−A long migration session hits Pro's usage limits fast and pushes you to Max

Visit Claude Code →

2

### Cursor

Best for: An AI-native IDE combining fast autocomplete and agent mode

PricingFree (Hobby) / Pro $20/mo / Pro+ $60/mo / Teams $40/user/mo

+Tab predicts multi-line edits, not just the next token

+Composer/Agent mode executes a change across a whole feature

+Bugbot reviews pull requests on GitHub before a human does

−You maintain a fork, so extension compatibility occasionally lags

−Heavy agent use in a large monorepo can get slow

Visit Cursor →

3

### GitHub Copilot

Best for: Teams already standardized on GitHub

PricingFree / Pro $10/mo / Pro+ $39/mo / Business $19/user/mo

+Deepest GitHub integration, agent works on issues and opens PRs

+Runs in VS Code, JetBrains, Visual Studio, and Neovim

+Free tier gives 2,000 completions a month

−Agent mode for large refactors has trailed Claude Code and Cursor

−Premium model quotas on Pro run out fast with regular Opus use

Visit GitHub Copilot →

4

### OpenAI Codex

Best for: Async tasks you hand off and check on later

PricingLimited on ChatGPT Free and Go ($8/mo); Plus $20/mo; Pro from $100/mo

+Included with a ChatGPT subscription, not billed separately

+Cloud agent runs async in a sandboxed container and opens a PR

+Codex CLI also runs locally against a raw API key

−Sandbox isolation adds latency and disconnects from local terminal state

−No standalone pricing, tied to a ChatGPT plan

Visit OpenAI Codex →

5

### Devin Desktop

Best for: Running several agent threads from one IDE

PricingFree / Pro $20/mo / Max $200/mo / Teams $40/seat ($80/mo minimum)

+Manage several agent threads from one command-center IDE

+Existing Windsurf accounts carried over with settings intact

+Free tier plus an individual Pro plan at the same price as Cursor Pro

−A mid-adoption rebrand disrupts muscle memory

−Docs and forum answers still say Windsurf

Visit Devin Desktop →

6

### Cline

Best for: Cost control with your own model and no lock-in

PricingFree (open source) + your model API costs

+Fully open source under Apache 2.0, no vendor lock-in

+Bring-your-own-key with full cost transparency

+Plan and Act modes plus Model Context Protocol tool support

−No bundled subscription cushioning a long session's API bill

−You are responsible for picking a model good at the task

Visit Cline →

7

### Zed

Best for: Developers who want the fastest editor with light AI

PricingFree / Pro $10/mo / Business $30/seat/mo

+Rust-built, lower latency than Electron-based competitors

+Personal use free forever with 2,000 edit predictions included

+Pro is $10/month with unlimited predictions and $5 token credit

−Agent panel is newer and the ecosystem is still catching up

−Less mature agent tooling than Cursor or Claude Code

Visit Zed →

8

### Amp

Best for: Developers who want the tool to pick the best model per task

PricingFree Hobby tier (bring your own keys); Individual $20/mo; extra credits at provider API rates

+Auto-routes to the best model per task, no manual choice

+No markup on provider API rates for individuals

+Shareable thread links for reviewing an agent's reasoning

−Included usage on the paid tier does not roll over, extra usage is billed as credits

−The lack of model choice is deliberate but won't suit everyone

Visit Amp →

9

### Aider

Best for: Terminal-first developers who treat git history as truth

PricingFree (open source) + your model API costs

+Auto-commits each change, so every edit is reviewable and revertible

+Builds a repo map to send the right context on large codebases

+Free and open source, bring-your-own-key billing

−No GUI at all, no inline diff viewer or visual file tree

−You pick and pay for the model yourself

Visit Aider →

## What it is

An AI coding assistant reads your code and helps you write or change it, but the three types work very differently. Autocomplete predicts what you are about to type, one line or a few lines at a time, and you accept or reject every suggestion with a keystroke.

Cursor's Tab, Copilot's completions, and Zed's edit predictions all work this way, keeping you in control of each change.

Agentic tools operate at a higher level. You hand over a task description and the agent plans a sequence of edits, executes them across multiple files, and runs your tests to check its own work before you review a single line.

Claude Code, Aider, Cline, Amp, and Devin Desktop's agents fit here. A third group, terminal-native tools like Claude Code, Aider, and Codex CLI, skips the editor entirely and runs from your shell, which matters if your workflow already runs through SSH, scripts, or CI.

## Why it matters

The tool you pick shapes your daily cost, your lock-in, and how well the assistant fits the way you already work.

Subscription tools like Cursor or Copilot give a fixed monthly bill, while bring-your-own-key tools like Cline and Aider bill you directly for every API call, which can be cheaper for light use but shows up as a real itemized cost during a long agent session.

Lock-in matters too. Open-source, BYOK tools let you swap models and avoid depending on one vendor's roadmap, while a fork like Cursor or a rebranded product like Devin Desktop ties you to that team's decisions. And a terminal tool is wasted on someone who lives in an IDE, just as an editor-bound agent frustrates a shell-heavy engineer.

## Key features to look for

Autocomplete versus agentic modeEssential
Autocomplete predicts your next line as you type and you approve each keystroke. Agentic mode takes a whole task, edits many files, and runs your tests on its own. Know which one a tool actually gives you before you pay.

Cross-file refactor capabilityEssential
The real test of an agent is whether it can rename a data model across forty files or migrate a test framework, tracing changes through a service boundary and iterating against your tests rather than editing one file at a time.

Pricing model: subscription, BYOK, or pay-as-you-goEssential
Subscriptions like Cursor or Copilot give a fixed monthly bill. Bring-your-own-key tools like Cline and Aider bill your API usage directly, cheaper for light work but a real itemized cost during a long agent session.

Terminal-native versus IDE-based
Terminal tools like Claude Code, Aider, and Codex CLI fit SSH, scripting, and CI. IDE tools like Cursor, Copilot, and Zed give inline diffs and a visual review step. A few, Claude Code and Amp among them, ship both.

Model choice and bring-your-own-key
Open-source BYOK tools let you swap between Anthropic, OpenAI, Google, or local models and avoid depending on one vendor. Others pick the model for you, which is simpler but ties you to that team's roadmap.

Git integration and reviewable diffs
Aider auto-commits every AI edit as a revertible git commit, and Amp shares threads by link for review. A visible, reviewable trail matters when you are trusting an agent to change code across your repo.

Mistakes to avoid

×Treating all three types as interchangeable. Buying an autocomplete tool when you need an agent for cross-file refactors, or the reverse, wastes money on the wrong workflow.

×Ignoring the real cost of BYOK tools. 'Free and open source' still means a long agent session against a big model shows up as a real API bill, not $0.

×Pointing an individual tier at code under an NDA. Individual plans don't guarantee the data commitments, no training, audit logs, and SSO, that Business and Enterprise tiers add.

Expert tips

→Pick based on where you actually work. A terminal-native tool is wasted in an IDE-heavy workflow, and an editor-bound agent frustrates a shell-first engineer.

→Start free before you commit. Cline, Aider, Copilot Free, Cursor Hobby, Zed Personal, and Devin Desktop Free all give real, if limited, access to test the fit.

→Read the actual data processing agreement, not the marketing page, before pointing any assistant at proprietary code.

## The bottom line

There is no single best AI coding assistant, because the category splits into three real jobs.

For agentic, cross-file refactors from the terminal, Claude Code tested strongest for us. If you want one AI-native IDE that does autocomplete and agent mode well, Cursor is the pick, and teams living inside GitHub get more from Copilot's issue-and-PR integration than anything else here.

For the best free, no-lock-in value, Cline and Aider are fully open source and cost only your model provider's rate. Budget $10-20/month for daily use of one serious tool, and $100-200/month only if you are running heavy agent sessions all day. Pick based on where you work, not which tool had the loudest launch.

## Frequently asked questions

What is the best AI coding assistant in 2026?

There is no single winner, since the category splits into three types. For agentic multi-file refactors from the terminal, Claude Code tested strongest. For an AI-native IDE with autocomplete and agent mode, Cursor. For teams standardized on GitHub, Copilot's integration is hard to match. Pick based on where you actually work.

Are there any free AI coding assistants?

Yes, and some are genuinely free rather than trials. Cline and Aider are fully open source, so you pay only your model provider's API rate, a few dollars a month for light use. Copilot Free gives 2,000 completions monthly, and Cursor Hobby, Zed Personal, and Devin Desktop Free all offer real, if limited, access.

How much should a solo developer expect to pay per month?

For light use, $0 is realistic on Cline or Aider with a cheap model. For daily use of one serious tool, budget $10-20/month, roughly Cursor Individual, Copilot Pro, Zed Pro, or Claude Pro. Heavy all-day agent use pushes toward $100-200/month like Claude Max, Devin Desktop Max, or ChatGPT Pro.

Is it safe to use these on proprietary or enterprise code?

Stick to the plans built for it. Copilot Business/Enterprise, Cursor Teams/Enterprise, Claude's Team/Enterprise, and Devin Desktop's Teams add data commitments, no training on your code, audit logs, and SSO, that individual tiers don't guarantee. Read the actual data processing agreement before pointing any of these at code under an NDA.

What is the difference between autocomplete and agentic tools?

Autocomplete (Copilot's completions, Zed's predictions, Cursor's Tab) predicts what you are about to type, one line at a time, with you in control of every keystroke. Agentic tools (Claude Code, Aider, Cline, Amp) take a task, plan a sequence of edits, execute them across many files, and check the result against your tests before you review.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [OpenAI Codex pricing](https://openai.com/pricing)
- [Devin Desktop pricing](https://cognition-labs.com)checked Sep 2026
- [Cline pricing](https://github.com/cline/cline)checked Sep 2026
- [Zed pricing](https://zed.dev/pricing)checked Sep 2026
- [Aider pricing](https://aider.chat)checked Sep 2026

Related guides

Ai For Code ReviewAi For DebuggingFree Ai Coding ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The 8 Best AI Code Review Tools in 2026

URL: https://devshot.co/reviews/best-ai-for-code-review
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: CodeRabbit, GitHub Copilot, Graphite, Greptile, and Qodo compared on real 2026 pricing, false-positive rates, and where each one actually catches bugs.

Expert Guide

## The 8 Best AI Code Review Tools in 2026

For developers and engineering teams: 8 AI code reviewers ranked on real 2026 pricing, false-positive rates, and where each one actually catches bugs.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

CodeRabbit and Greptile are the strongest general-purpose reviewers if you are choosing from scratch: CodeRabbit for a tunable drop-in on every PR, Greptile for the deepest whole-codebase context. GitHub Copilot is the obvious pick if you already pay for completions, since PR review is effectively free at that point. For the best value across GitHub, GitLab, and Bitbucket, Bito is the cheapest per seat at $12/month. Graphite only makes sense if you already stack PRs, and Qodo Merge is built for the narrow case of PRs that do not implement the ticket.

Key facts8 tools compared: CodeRabbit, GitHub Copilot, Graphite, Greptile, Qodo Merge, Bito, Sourcery…

- Updated: September 25, 2026
- Top pick: CodeRabbit (best for: A drop-in reviewer on every PR)
- Top pick price as of September 25, 2026: CodeRabbit: Free on public repos; Essentials $24/dev/mo (annual); Team $48/dev/mo (annual); 14-day trial

- 8 tools compared: CodeRabbit, GitHub Copilot, Graphite, Greptile, Qodo Merge, Bito, Sourcery, Codacy
- GitHub Copilot (best for: Teams already all-in on GitHub): Free (no PR review); Pro $10/mo; Pro+ $39/mo; Max $100/mo
- Graphite (best for: Stacked-diff, trunk-based teams): Free (limited); Starter $20/user/mo; Team $40/user/mo (annual) for unlimited
- Greptile (best for: Deep cross-file, whole-codebase context): Free (50 credits/mo, 1 dev); Pro $30/seat/mo; Enterprise custom

Every AI code reviewer will summarize a diff in ten seconds and leave a confident comment. The question that matters is whether it catches the null check you missed or the auth bypass hidden in a 40-file refactor, without burying that finding under thirty nitpicks about variable naming.

We looked at eight tools that show up in real pull-request queues: general-purpose reviewers, a native GitHub feature, a codebase-indexing tool built for cross-file context, and platforms where AI sits on top of existing static analysis.

We checked pricing against vendor pages and flagged usage-based billing, since a per-seat number that turns metered at scale changes the math. None of these replace a reviewer who understands why a change exists. They are fast at the mechanical part: syntax, security patterns, missing tests, and style drift. Architecture judgment calls still need a human who has read the ticket.

## Top Picks

Based on features, real-world fit, and value for money.

8 Best AI Code Review Tools in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[CodeRabbit](https://toolradar.com/tools/coderabbit) | Free on public repos; Essentials $24/dev/mo (annual); Team $48/dev/mo (annual); 14-day trial | A drop-in reviewer on every PR |

[GitHub Copilot](https://toolradar.com/tools/github-copilot) | Free (no PR review); Pro $10/mo; Pro+ $39/mo; Max $100/mo | Teams already all-in on GitHub |

[Graphite](https://toolradar.com/tools/graphite) | Free (limited); Starter $20/user/mo; Team $40/user/mo (annual) for unlimited | Stacked-diff, trunk-based teams |

[Greptile](https://toolradar.com/tools/greptile) | Free (50 credits/mo, 1 dev); Pro $30/seat/mo; Enterprise custom | Deep cross-file, whole-codebase context |

[Qodo Merge](https://toolradar.com/tools/qodo) | 14-day trial; Pro Team from $30/mo (up to 30 users) + credits | Validating PRs against Jira or Linear tickets |

[Bito](https://toolradar.com/tools/bito) | Team $12/seat/mo (annual); Professional $20/seat/mo (annual); Enterprise custom | Teams on GitHub, GitLab, or Bitbucket |

[Sourcery](https://toolradar.com/tools/sourcery) | Free for open source; Pro $12/seat/mo; Team $24/seat/mo | Review plus security scanning |

[Codacy](https://toolradar.com/tools/codacy) | Free (individual); Team $18/dev/mo (annual); Business custom | Quality gates across 40+ linters |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 8 does not publish a comparable monthly price and is left out rather than estimated.

1

### CodeRabbit

Top Pick

Best for: A drop-in reviewer on every PR

PricingFree on public repos; Essentials $24/dev/mo (annual); Team $48/dev/mo (annual); 14-day trial

+Plain-English summaries plus line-by-line comments with one-click fixes

+Generates sequence diagrams for non-trivial changes and runs SAST tools and linters under the LLM

+Free forever on public repositories, and every plan starts with a 14-day trial, no card

−Can get chatty on large or messy PRs, restating what the diff already shows

−Private repositories need a paid plan once the 14-day trial ends

Visit CodeRabbit →

2

### GitHub Copilot

Best for: Teams already all-in on GitHub

PricingFree (no PR review); Pro $10/mo; Pro+ $39/mo; Max $100/mo

+Review lives inside GitHub with no separate app or second place to check comments

+Custom review instructions via an instructions.md file on Pro+ and Max

+Strong at typos, obvious null-pointer risks, and missing error handling

−Review depth is shallower than a purpose-built reviewer

−No cross-file context to catch an architectural problem three services away

Visit GitHub Copilot →

3

### Graphite

Best for: Stacked-diff, trunk-based teams

PricingFree (limited); Starter $20/user/mo; Team $40/user/mo (annual) for unlimited

+AI reviewer reviews each diff in a stack aware of the diffs underneath it

+Free Hobby tier includes CLI stacking, MCP, and a VS Code extension

+Team plan unlocks unlimited reviews plus customization of what gets flagged

−The AI review worth having is throttled on the cheaper plans, so budget for Team

−Wasted if you do not stack PRs

Visit Graphite →

4

### Greptile

Best for: Deep cross-file, whole-codebase context

PricingFree (50 credits/mo, 1 dev); Pro $30/seat/mo; Enterprise custom

+Indexes the entire codebase as a graph, so it knows every caller of a function

+Catches the caller-side break, not just 'this variable could be null'

+Free for qualified open source, and 50% off for startups under $2M revenue

−Credit-based: TREX reviews and high PR volume burn the allotment fast

−Extra reviews cost $1 per credit once the pool runs out

Visit Greptile →

5

### Qodo Merge

Best for: Validating PRs against Jira or Linear tickets

Pricing14-day trial; Pro Team from $30/mo (up to 30 users) + credits

+Pulls the linked Jira or Linear ticket and flags PRs that do not implement what was scoped

+Standard review, describe, and improve commands alongside ticket validation

+Beta feature catches breaking changes in shared packages before they hit a dependent repo

−No permanent free tier, only a 14-day trial

−Credit-based shared pool is harder to forecast than a flat per-seat number

Visit Qodo Merge →

6

### Bito

Best for: Teams on GitHub, GitLab, or Bitbucket

PricingTeam $12/seat/mo (annual); Professional $20/seat/mo (annual); Enterprise custom

+Reviews GitHub, GitLab, and Bitbucket from one plan

+Inline, codebase-aware feedback with one-click fixes and Jira integration

+Ties Sourcery Pro for the lowest per-seat price here, $12/seat/mo annually

−Lines-of-code metering (5,000/seat, then $5 per extra 1,000) triggers overage on big refactors

−Newer than CodeRabbit or Copilot with a smaller integration ecosystem

Visit Bito →

7

### Sourcery

Best for: Review plus security scanning

PricingFree for open source; Pro $12/seat/mo; Team $24/seat/mo

+Free Open Source tier gives full Pro features on public repos

+Full PR review plus security scanning positioned as high signal, low noise

+Custom review rules, diagrams, and repo analytics; Team adds bring-your-own-LLM

−'Low noise' is Sourcery's own framing, not an independent benchmark

−Deepest strength is still Python; JVM or Rust-heavy stacks should pilot first

Visit Sourcery →

8

### Codacy

Best for: Quality gates across 40+ linters

PricingFree (individual); Team $18/dev/mo (annual); Business custom

+Static analysis across 40-plus linters (SAST, SCA, secrets) with AI layered on top

+Strongest at org-wide policy enforcement across many repos

+Free Developer tier with an IDE plugin, scan-as-you-type, and AI auto-fix

−Weaker at logic bugs that require understanding intent

−AI rides on rules and linters rather than reading the diff from scratch

Visit Codacy →

## What it is

AI code review tools plug into your Git host and read every pull request before a human does. They post a plain-English summary of the change, then drop inline comments on specific lines, often with a one-click fix you can commit straight from the PR.

Under the LLM layer, most also run traditional linters and static-analysis scanners, so a single check covers style, security patterns, and obvious logic slips.

The difference between a shallow reviewer and a useful one is context. A basic tool reads only the diff; a stronger one indexes your whole codebase as a graph, so when it comments on a function it knows every place that function is called.

That is the gap between "this variable could be null" and "this breaks the caller in billing/invoice.go that assumes the field is always set." Some go further and validate the PR against the linked Jira or Linear ticket, flagging changes that look complete but do not implement what was scoped.

## Why it matters

The choice comes down to three things: how the tool bills, whether it locks you into a workflow, and how much noise it adds to every PR. Per-seat pricing is easy to forecast; credit pools, lines-of-code caps, and per-review overages are not, and a couple of large refactor sprints can blow through an allotment that looked generous on paper.

Workflow fit matters just as much. A stacked-diff reviewer is wasted if your team does not stack PRs, and a GitHub-only tool is a dead end if your org also runs GitLab or Bitbucket.

And every vendor claims "low noise" while none benchmarks it independently, so the false-positive rate you actually get is the one you measure on your own pull requests during the trial.

## Key features to look for

Whole-codebase contextEssential
The best reviewers index your entire repo as a graph, so a comment on one function knows every caller. Diff-only tools miss the cross-file break; context-aware ones catch it three services away.

Signal-to-noise controlEssential
False positives are the top complaint across every tool here. Look for one that lets you mute rule categories and tune sensitivity, then test the real noise level on your own PRs, not the vendor's 'low noise' claim.

Git host coverage
Most tools support GitHub, several add GitLab, and only a few cover Bitbucket. If your org spans hosts this narrows the list fast, because a GitHub-only reviewer is a dead end when half your repos live elsewhere.

Pricing model
Flat per-seat plans are easy to forecast. Credit pools, lines-of-code caps, and per-review overages are not, and a few large refactor sprints can trigger charges a flat-rate tool would never bill.

Security scanning
Tools with SAST and SCA underneath reliably catch known vulnerability patterns and vulnerable dependencies. None reliably catch a business-logic flaw like an auth check that runs but tests the wrong condition.

Workflow and ticket fit
Some reviewers customize what gets flagged via a config file or custom rules; a few validate the PR against its Jira or Linear ticket and flag work that looks done but skips part of the scope.

Mistakes to avoid

×Judging a tool too noisy before spending twenty minutes tuning its config and muting rule categories you do not care about.

×Buying a credit, lines-of-code, or per-review plan without doing the math on your real PR volume, then getting hit with overage after one heavy refactor sprint.

×Adopting a workflow-specific tool like stacked-diff Graphite for the reviewer alone, or picking a GitHub-only tool when your org also runs GitLab or Bitbucket.

Expert tips

→Run the trial on real pull requests for two full sprints before rolling anything out org-wide.

→If architecture-level bugs matter more than per-review cost, favor a tool that indexes the whole codebase, and check the credit math first.

→Keep a human as final approver on anything touching auth, payments, or data migrations; AI handles the mechanical first pass, not intent.

## The bottom line

There is no single best AI code reviewer; the right one depends on your git host, your workflow, and what you already pay for. If you are evaluating from scratch, start with CodeRabbit for a tunable drop-in reviewer or Greptile when architecture-level bugs matter more than per-review cost.

Already on GitHub and paying for Copilot? Turn on its PR review before adding a second tool. Split across GitHub, GitLab, and Bitbucket?

Bito covers all three from one plan at the lowest per-seat price here. Whatever you shortlist, run it on real PRs for two sprints and keep a human as final approver on anything touching auth, payments, or data migrations.

## Frequently asked questions

What is the best AI for code review in 2026?

There is no single best one; it depends on what you already run. CodeRabbit and Greptile are the strongest general-purpose reviewers from scratch. GitHub Copilot wins if you already pay for it. Graphite only fits stacked-PR teams. Pick on your git workflow, not the loudest launch post.

How much should a small team budget for AI code review?

For a five-person team, expect $60 to $150 a month at the low end with CodeRabbit or Bito's cheaper tiers, up to $150 to $200 for Greptile or Qodo Merge at typical usage. Run the trial for two sprints before deciding if the catch rate justifies an org-wide rollout.

Is there a free or open-source AI code review tool?

Yes. Qodo's PR-Agent is open source and self-hostable with your own model key. Sourcery is free for open-source repos, Codacy for individual developers, and Greptile's Starter tier gives 50 credits a month. CodeRabbit reviews public repositories free, but private repos need a paid plan after the trial.

Does AI code review replace human review?

No. It replaces the mechanical first pass: syntax errors, missing null checks, obvious security patterns, and style drift. It does not judge whether a change is right for your product. Keep a human as final approver on anything touching auth, payments, or data migrations.

Which of these works with GitLab or Bitbucket, not just GitHub?

Bito supports all three from one plan, rare on this list. CodeRabbit, Greptile, Qodo Merge, and Sourcery also cover GitHub and GitLab (verify Bitbucket first). Copilot's review is GitHub-only and Graphite is built around GitHub's PR model. Spanning hosts, narrow to Bito, CodeRabbit, or Greptile.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [CodeRabbit pricing](https://coderabbit.ai/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Graphite pricing](https://graphite.com/pricing)checked Sep 2026
- [Greptile pricing](https://www.greptile.com/pricing)checked Sep 2026
- [Qodo Merge pricing](https://www.qodo.ai/pricing)checked Sep 2026
- [Bito pricing](https://bito.ai/pricing)checked Sep 2026
- [Sourcery pricing](https://sourcery.ai/pricing)checked Sep 2026
- [Codacy pricing](https://codacy.com/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsAi For DebuggingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best AI for Debugging in 2026

URL: https://devshot.co/reviews/best-ai-for-debugging
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: 8 AI debugging tools tested and compared for 2026: Claude Code, Cursor, Sentry Seer, Datadog Bits AI, Rollbar, Honeycomb, and more, with real pricing.

Expert Guide

## The Best AI for Debugging in 2026

For developers and engineering teams, eight AI debugging tools ranked on how they trace code bugs and correlate incidents across live production systems.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

Debugging splits into two jobs. For tracing bugs in your own codebase, Claude Code and Cursor are the strongest coding agents, reading full-repo context and running your tests to verify a fix. For production incidents across distributed systems, Datadog's Bits AI and Sentry's Seer lead on cross-service correlation. The best free starting point is GitHub Copilot's free tier, alongside the free platform tiers on Rollbar and Honeycomb. Most teams run one coding agent and one observability layer, not one or the other.

Key facts8 tools compared: Claude Code, Cursor, GitHub Copilot, Sentry, Datadog, Rollbar, Honeycomb…

- Updated: September 25, 2026
- Top pick: Claude Code (best for: Tracing bugs that span more than one file)
- Top pick price as of September 25, 2026: Claude Code: From $17/mo (Pro, billed yearly); $20 if billed monthly

- 8 tools compared: Claude Code, Cursor, GitHub Copilot, Sentry, Datadog, Rollbar, Honeycomb, New Relic
- Cursor (best for: Debugging without leaving your code editor): From $20/mo (Pro); free Hobby tier
- GitHub Copilot (best for: Teams that already work in GitHub): From $10/mo (Pro); free tier available
- Sentry (best for: A pull request that fixes a narrow error): From $26/mo (Team, billed yearly); Developer plan free

AI changed how bugs get hunted, but not uniformly. A coding assistant like Claude Code or Cursor handles the local problem: read the trace, read the code around it, propose a fix, and run the tests.

An observability platform's AI layer, like Sentry's Seer or Datadog's Bits AI, handles a different problem, correlating a latency spike with a deploy and a noisy pod across systems no single engineer holds full context on. Treating these as interchangeable is how teams end up disappointed.

This guide covers both categories and stays explicit about which is which. Pricing was checked against vendor pages directly, and where a vendor doesn't publish a number, that's flagged instead of guessed. Devshot covers AI and dev tools daily, so the picks below are grouped by where the bug actually lives, on your machine or in production.

## Top Picks

Based on features, real-world fit, and value for money.

Best AI for Debugging in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Claude Code](https://toolradar.com/tools/claude-code) | From $17/mo (Pro, billed yearly); $20 if billed monthly | Tracing bugs that span more than one file |

[Cursor](https://toolradar.com/tools/cursor) | From $20/mo (Pro); free Hobby tier | Debugging without leaving your code editor |

[GitHub Copilot](https://toolradar.com/tools/github-copilot) | From $10/mo (Pro); free tier available | Teams that already work in GitHub |

[Sentry](https://toolradar.com/tools/sentry) | From $26/mo (Team, billed yearly); Developer plan free | A pull request that fixes a narrow error |

[Datadog](https://toolradar.com/tools/datadog) | From $500/mo for 500 credits (Bits AI, billed annually) or $1.30/credit | Production incidents that span more than one service |

[Rollbar](https://toolradar.com/tools/rollbar) | Free at $0/mo (5,000 occurrences/month); paid plans unpublished | Smaller teams that want related errors grouped |

[Honeycomb](https://toolradar.com/tools/honeycomb) | From $150/mo (Pro); free up to 20M events/month, copilot included | Teams on OpenTelemetry querying traces with many fields |

[New Relic](https://toolradar.com/tools/new-relic) | Free (100GB ingest/month); overage $0.40/GB ($0.60/GB Data Plus) | Teams that already run New Relic |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Claude Code

Top Pick

Best for: Tracing bugs that span more than one file

PricingFrom $17/mo (Pro, billed yearly); $20 if billed monthly

+Full-repo context without manual copy-pasting

+Uses git blame and log history as tools while tracing

+Actually executes tests instead of asserting a fix

−No built-in access to production logs; you wire it up via MCP

−Pro-tier usage limits get hit fast on long loops in a large monorepo

Visit Claude Code →

2

### Cursor

Best for: Debugging without leaving your code editor

PricingFrom $20/mo (Pro); free Hobby tier

+Fewer round trips than a separate chat tool

+Bugbot catches real bugs at PR review time

+Multi-model access when one model stalls

−Still fundamentally local; won't pull production telemetry on its own

−Usage-based billing on heavier tasks adds up for high-volume teams

Visit Cursor →

3

### GitHub Copilot

Best for: Teams that already work in GitHub

PricingFrom $10/mo (Pro); free tier available

+Deep Issues, PR, and Actions integration

+A real free tier to start on

+Model choice for harder bugs

−Inconsistent quality across models

−Free tier's 2,000 completions/month disappears fast in a real session

Visit GitHub Copilot →

4

### Sentry

Best for: A pull request that fixes a narrow error

PricingFrom $26/mo (Team, billed yearly); Developer plan free

+Stack trace and code context arrive automatically

+Strong on bounded exceptions like duplicate keys

+Suggestions land as real PRs to review

−Weaker on timing bugs and race conditions the trace doesn't explain

−Seer isn't in any base plan; it costs $40 per active contributor a month on top

Visit Sentry →

5

### Datadog

Best for: Production incidents that span more than one service

PricingFrom $500/mo for 500 credits (Bits AI, billed annually) or $1.30/credit

+Genuinely strong at cross-service correlation during live incidents

+Turns a 45-minute log hunt into a five-minute review

+Ties latency spikes to specific deploy markers

−Surfaces correlation, not causation; you confirm before rolling back

−AI features run on a separate credits system from base pricing

Visit Datadog →

6

### Rollbar

Best for: Smaller teams that want related errors grouped

PricingFree at $0/mo (5,000 occurrences/month); paid plans unpublished

+Breaks down the usual frontend/backend silo

+Free tier (5,000 occurrences/month) is a real option

+Correlates mobile crashes with backend schema changes

−Rollbar Resolve auto-PR fixes are still coming soon; don't rely on it yet

−Essentials and Advanced pricing isn't published

Visit Rollbar →

7

### Honeycomb

Best for: Teams on OpenTelemetry querying traces with many fields

PricingFrom $150/mo (Pro); free up to 20M events/month, copilot included

+Copilot lowers the barrier to asking sharp questions for OpenTelemetry teams

+Answers high-cardinality questions dashboards can't

+Copilot included on both the free and Pro tiers

−Doesn't fix code or open PRs; it only generates queries

−Only as good as your instrumentation

Visit Honeycomb →

8

### New Relic

Best for: Teams that already run New Relic

PricingFree (100GB ingest/month); overage $0.40/GB ($0.60/GB Data Plus)

+AI layer is there without adopting a new platform

+Compresses alert noise into one narrative

+Groups multi-service alerts back to a single rollout

−No separate AI pricing published; bundled into Advanced Compute

−Ingest overage ($0.40/GB beyond 100GB free) gets expensive for chatty services

Visit New Relic →

## What it is

AI debugging tools fall into two concrete groups. Coding agents such as Claude Code, Cursor, and GitHub Copilot work with your source code, git history, and test suite. They read a stack trace, walk the call chain to the function upstream of the one that threw, patch it, and rerun the tests in a loop until they pass. Their output is a diff or a pull request you review.

Observability platforms with an AI layer, Sentry, Datadog, Rollbar, Honeycomb, and New Relic, work with telemetry from live systems: APM traces, infra metrics, logs, and deploy markers. Their AI correlates what already happened, turning a 45-minute log-grepping session into a five-minute review of one timeline.

It surfaces a strong hypothesis, not code, so an engineer still confirms causation before rolling back.

## Why it matters

Picking the wrong category wastes money and trust. A coding agent won't pull production telemetry on its own, and an observability AI layer won't fix code, so a mismatch leaves you paying for a tool that only summarizes an incomplete picture confidently.

Cost is rarely the sticker price. Sentry's Seer and Datadog's Bits AI are add-ons billed on top of the base plan, per contributor or per credit, so get a real quote or run a pilot before committing.

Lock-in matters too: the AI layer is only worth adopting if your instrumentation already flows in, and connecting an agent to your full repo or logs touching customer data means checking retention and model-training policy first.

## Key features to look for

Full-repo contextEssential
Coding agents that read the whole repository instead of only the open file catch bugs upstream of the function that threw, which is where most real bugs actually live.

Executes tests to verify fixesEssential
A tool that reruns your test suite in a loop until it passes verifies its own fix, rather than asserting a change works and leaving you to find out in production.

Cross-service correlationEssential
For production incidents, the AI ties APM traces, infra metrics, logs, and deploy markers into one timeline, the hardest part of debugging to do by hand under pressure.

Real PRs, not just suggestions
Output that lands as a reviewable pull request, not just a chat suggestion, fits into an existing Git workflow and keeps a human in the loop before anything ships.

Transparent pricing
Many AI features are add-ons billed per active contributor or on a credits system, separate from the base plan, so the headline price rarely reflects the real cost.

Instrumentation fit
An observability AI is only as sharp as the telemetry feeding it. Poor OpenTelemetry instrumentation means the AI summarizes an incomplete picture with false confidence.

## Pricing

Coding agents publish list prices. Sentry publishes Team at $26/mo and Business at $80/mo, both billed yearly, and Seer is a separate add-on billed per active contributor. Rollbar is free at $0/mo, then Essentials and Advanced are usage-based and unpublished, and Enterprise is custom, as are Cursor Enterprise and Honeycomb Enterprise.

New Relic is free for 100GB a month, then $0.40/GB overage ($0.60/GB for Data Plus), while Standard, Pro, and Enterprise scale up with no list price and the AI is bundled into Advanced Compute.

The cheapest paid entry is GitHub Copilot Pro at $10/mo, with a free tier, then Claude Code at $17/mo on Pro billed yearly, Cursor Pro just above it, and Claude Max from ~$100/mo.

Costs jump at Honeycomb Pro from $150/mo and at Datadog Bits AI, $500/mo for 500 monthly credits billed annually, or $1.30 a credit on demand, on top of APM from $31/host/mo and Infrastructure Pro from $15/host/mo.

Plan | Price | Best for |

Claude Code Pro | $17/mo yearly ($20 monthly) | Included with a Claude Pro subscription |

Claude Code API | Pay as you go | Pay-as-you-go API, also available |

Claude Code Max | From ~$100/mo | Max plan above Pro |

Cursor Pro | $20/mo | Individual plan for one person |

Cursor Teams | $40/user/mo | Per-user plan for teams |

Cursor Enterprise | Custom quote | Enterprise plan priced by quote |

GitHub Copilot Pro | $10/mo | Paid plan above the free tier |

GitHub Copilot Pro+ | $39/mo | Higher individual plan |

GitHub Copilot Business | $19/user/mo | Per-user plan for a business |

GitHub Copilot Enterprise | $39/user/mo | Per-user plan for an enterprise |

Sentry Team | $26/mo, billed yearly | Yearly team plan above free Developer |

Sentry Business | $80/mo, billed yearly | Higher yearly base plan |

Sentry Seer | $40/active contributor/mo | AI debugging add-on for paid Team, Business or Enterprise plans |

Datadog Bits AI | From $500/mo, billed annually | Bundle of 500 AI credits a month, unused credits do not roll over |

Datadog Bits AI on-demand | $1.30/credit | On-demand price per credit |

Datadog APM | From $31/host/mo | Base APM, priced per host |

Datadog Infrastructure Pro | From $15/host/mo | Infrastructure Pro, priced per host |

Rollbar Essentials | Custom quote | Usage-based, price not published |

Rollbar Advanced | Custom quote | Usage-based, price not published |

Rollbar Enterprise | Custom quote | Enterprise plan priced by quote |

Honeycomb Pro | From $150/mo | Up to 750M events a month |

Honeycomb Enterprise | Custom quote | Enterprise plan priced by quote |

New Relic Standard | Scales up, no list price | No published price, scales up from free |

New Relic Pro | Scales up, no list price | Higher tier, no published price |

New Relic Enterprise | Scales up, no list price | Top tier, no published price |

New Relic overage | $0.40/GB | Ingest past the free 100GB a month |

New Relic Data Plus | $0.60/GB | Data Plus ingest overage |

Mistakes to avoid

×Treating coding agents and observability AI as interchangeable. One reads code and runs tests; the other correlates live telemetry. Buy the category that matches where your bug actually lives.

×Judging by sticker price. Seer and Bits AI are add-ons billed on top of the platform, so the base plan's number understates what you'll actually pay.

×Enabling an observability AI on top of weak instrumentation. If your traces and metrics are incomplete, the AI just summarizes a partial picture with confidence you shouldn't trust.

Expert tips

→Run one of each. Most teams pair a coding agent (Claude Code or Cursor) for local bugs with one observability AI layer for production, not one or the other.

→Match the observability tool to your instrumentation. Honeycomb's copilot rewards OpenTelemetry shops; Datadog and New Relic assume metrics, logs, and traces already flow in.

→Check retention and model-training policy before connecting an agent to your full repo or an observability AI to logs touching customer PII.

## The bottom line

There's no single best AI for debugging, because debugging is two jobs. For bugs you can reproduce on your machine, Claude Code is the strongest pick, with Cursor close behind if you want the fix inside your editor and GitHub Copilot the best free starting point for GitHub-native teams.

For production incidents across distributed systems, Datadog's Bits AI leads on cross-service correlation, Sentry's Seer is the most mature at autofixing scoped exceptions, and Honeycomb wins for OpenTelemetry-heavy shops. Rollbar and New Relic fit teams already on those platforms.

Run a pilot on the paid add-ons before committing, and treat every AI root cause as a strong hypothesis, not a verdict.

## Frequently asked questions

What is the best AI for debugging in 2026?

There's no single winner, because debugging splits into two jobs. For tracing bugs in your own codebase, Claude Code and Cursor are the strongest coding agents. For production incidents across distributed systems, Datadog's Bits AI and Sentry's Seer are the most mature. Pick based on which job you're doing.

Are there free AI debugging tools?

Yes. GitHub Copilot's free tier gives 2,000 completions a month, and Cursor's Hobby tier is free with limited agent requests. Sentry, Rollbar, Honeycomb, and New Relic all have free platform tiers, though the AI features like Seer and Bits AI are usually paid add-ons even when the base platform is free.

How much do AI debugging tools cost?

Coding agents run cheap: Claude Code from $17/month, Cursor Pro at the same monthly rate as Claude Pro, Copilot Pro at $10/month. Observability AI costs more and hides in add-ons: Datadog's Bits AI starts at $500/month for 500 credits, and Sentry's Seer adds $40 per active contributor a month.

How reliable is AI-generated root cause analysis?

For well-scoped, single-cause bugs, a null check, a duplicate key, a missing field, it's reliable enough to ship with review, not without it. For multi-cause incidents, especially timing and concurrency bugs, AI narrows the search well but the final causal claim still needs a human. Treat it as a hypothesis, not a verdict.

Do these tools replace observability platforms?

No. Coding agents need telemetry fed to them; they don't collect it. Observability platforms with AI layers still need proper instrumentation to have anything to correlate. The AI sits on top of existing logs, metrics, and traces rather than replacing them.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Sentry pricing](https://sentry.io/pricing)checked Sep 2026
- [Datadog pricing](https://datadoghq.com/pricing)checked Sep 2026
- [Rollbar pricing](https://rollbar.com/pricing)checked Sep 2026
- [Honeycomb pricing](https://honeycomb.io/pricing)checked Sep 2026
- [New Relic pricing](https://newrelic.com/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsAi For Code ReviewSentry ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best AI for DevOps in 2026

URL: https://devshot.co/reviews/best-ai-for-devops
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: We compared Pulumi AI, Cursor, Claude Code, Datadog Bits AI, PagerDuty, k8sGPT, Harness and BigPanda for real DevOps and SRE workflows in 2026.

Expert Guide

## The Best AI for DevOps in 2026

For DevOps and SRE teams: nine AI tools for infrastructure code, Kubernetes diagnostics, and incident response, ranked on price and workflow fit.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 9 tools compared

TL;DR

"AI for DevOps" covers three different jobs, so there's no single winner. For infrastructure code, Pulumi AI drafts reviewable IaC diffs, while Cursor and Claude Code handle Terraform and Helm inside your repo. For Kubernetes diagnostics, k8sGPT is the best value: free, open source, and purpose-built. For 3 a.m. pages and alert fatigue, PagerDuty Advance and BigPanda correlate the noise. Match the tool to the job, and start with the AI already sitting inside tools you already pay for.

Key facts9 tools compared: Pulumi AI, Cursor, Claude Code, Datadog Bits AI, PagerDuty AI, k8sGPT…

- Updated: September 25, 2026
- Top pick: Pulumi AI (best for: Pulumi teams wanting AI-drafted stacks and PR reviews)
- Top pick price as of September 25, 2026: Pulumi AI: Free (1 user); Essentials $40/mo with 40 credits, unlimited users; Neo at $3 per million tokens

- 9 tools compared: Pulumi AI, Cursor, Claude Code, Datadog Bits AI, PagerDuty AI, k8sGPT, Harness AI, Firefly, BigPanda
- Cursor (best for: Editing Terraform, Helm, and app code in one IDE): Free (Hobby); Pro from $20/mo; Teams $40/user/mo
- Claude Code (best for: CLI-driven repo-wide IaC refactors and CI automation): In Claude Pro $17/mo (annual); Max from $100/mo
- Datadog Bits AI (best for: Correlating traces, logs, and metrics during incidents): $500 per 500 credits/mo (annual); $1.30/credit on-demand

"AI for DevOps" gets used as a catch-all, and that's the trap. An AI that writes a Terraform module isn't the same as one that triages a 3 a.m. page, and neither reads kubectl describe output to explain why a pod is stuck in CrashLoopBackOff.

Vendors blur these categories because it sells more seats. We split the space into three real jobs: generating and reviewing infrastructure-as-code, diagnosing Kubernetes clusters, and handling incidents and alert noise (AIOps).

Then we compared the tools people actually run for each job and checked current pricing directly against each vendor's page. None of these replace an on-call engineer's judgment. The good ones cut the time between "something's wrong" and "here's what's wrong," or between "I need a VPC with three subnets" and a reviewable pull request.

This guide ranks nine of them so you can pick by the job in front of you, not the marketing category.

## Top Picks

Based on features, real-world fit, and value for money.

Best AI for DevOps in 2026: 9 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Pulumi AI](https://toolradar.com/tools/pulumi) | Free (1 user); Essentials $40/mo with 40 credits, unlimited users; Neo at $3 per million tokens | Pulumi teams wanting AI-drafted stacks and PR reviews |

[Cursor](https://toolradar.com/tools/cursor) | Free (Hobby); Pro from $20/mo; Teams $40/user/mo | Editing Terraform, Helm, and app code in one IDE |

[Claude Code](https://toolradar.com/tools/claude-code) | In Claude Pro $17/mo (annual); Max from $100/mo | CLI-driven repo-wide IaC refactors and CI automation |

[Datadog Bits AI](https://toolradar.com/tools/datadog) | $500 per 500 credits/mo (annual); $1.30/credit on-demand | Correlating traces, logs, and metrics during incidents |

[PagerDuty AI](https://toolradar.com/tools/pagerduty) | From $21/user/mo (Professional, annual); Business $41/user/mo | Auto-triaging and summarizing incidents on-call |

[k8sGPT](https://toolradar.com/tools/k8sgpt) | Free, open source (Apache-2.0) | Diagnosing cluster issues without a subscription |

[Harness AI](https://toolradar.com/tools/harness) | Free tier; Essentials and Enterprise quote-only | Teams already standardized on Harness pipelines |

[Firefly](https://toolradar.com/tools/firefly) | 14-day free trial; Essential $2,499/mo (annual, up to 20K assets) | Multi-cloud shops that need IaC drift detection |

[BigPanda](https://toolradar.com/tools/bigpanda) | Custom quote; no public pricing | Enterprise NOC teams drowning in alert noise |

Pricing read from each vendor's own published pricing page, checked Sep 2026. 1 of 9 does not publish one; those entries say so rather than estimating.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 9 do not publish a comparable monthly price and are left out rather than estimated.

1

### Pulumi AI

Top Pick

Best for: Pulumi teams wanting AI-drafted stacks and PR reviews

PricingFree (1 user); Essentials $40/mo with 40 credits, unlimited users; Neo at $3 per million tokens

+Drafts actual TypeScript, Python, or Go as a reviewable diff against your existing stack

+Runs as an automated PR reviewer, flagging public S3 buckets or missing tags

+Free tier for one user, no credit card required

−Only knows Pulumi's model; no benefit for raw Terraform or CloudFormation until you migrate

−Token-based overage pricing gets unpredictable on autonomous multi-step tasks over a large stack

Visit Pulumi AI →

2

### Cursor

Best for: Editing Terraform, Helm, and app code in one IDE

PricingFree (Hobby); Pro from $20/mo; Teams $40/user/mo

+Edits .tf files in place with full repo context

+Free Hobby tier covers light IaC editing

+Handles Terraform, Helm charts, and application code in one place

−No awareness of live cloud state unless wired to a state-reading MCP server or given shell access

−Can generate a block for a resource that already exists or a change that forces a destructive replace

Visit Cursor →

3

### Claude Code

Best for: CLI-driven repo-wide IaC refactors and CI automation

PricingIn Claude Pro $17/mo (annual); Max from $100/mo

+Runs headless for jobs like refactoring every module to a new provider version

+Runs inside CI to fix a failing terraform validate step

+Ships inside Claude Pro at $17/month

−No awareness of live cloud state without an MCP server or shell access

−Output is a draft and can suggest a destructive replace if applied directly

Visit Claude Code →

4

### Datadog Bits AI

Best for: Correlating traces, logs, and metrics during incidents

Pricing$500 per 500 credits/mo (annual); $1.30/credit on-demand

+Natural-language search across traces, logs, and metrics

+Bits Investigation pulls an incident timeline together faster than clicking five dashboards

+Offers AI-assisted code fixes straight from an error trace

−Only makes sense if you already pay for Datadog's core observability stack

−Credit model makes costs hard to forecast; a bad incident week burns the allotment fast

Visit Datadog Bits AI →

5

### PagerDuty AI

Best for: Auto-triaging and summarizing incidents on-call

PricingFrom $21/user/mo (Professional, annual); Business $41/user/mo

+Auto-generates incident summaries and status updates

+Suggests likely responders based on past incidents

+Drafts runbook-style remediation steps from historical data

−One-time AI Actions don't refresh monthly, so high-volume teams buy more mid-cycle

−Only as good as your service catalog; untagged services give it nothing useful to summarize

Visit PagerDuty AI →

6

### k8sGPT

Best for: Diagnosing cluster issues without a subscription

PricingFree, open source (Apache-2.0)

+Free and open source under Apache-2.0

+Purpose-built for Kubernetes: explains CrashLoopBackOff, ImagePullBackOff, misconfigured HPAs

+Bring your own LLM backend (OpenAI, Claude, or a local model via Ollama)

−Diagnoses but doesn't fix unless you enable auto-remediation, risky in production

−Needs an LLM API key, so cluster error output leaves the cluster unless you run a local model

Visit k8sGPT →

7

### Harness AI

Best for: Teams already standardized on Harness pipelines

PricingFree tier; Essentials and Enterprise quote-only

+Pipeline failure analysis points at the likely cause of a broken build

+AI-assisted test triage separates real failures from flaky tests

+Free plan for individuals and small teams on the open-source starter modules

−No public pricing above the free tier, so budgeting needs a sales call

−Only worth adopting if you run CI/CD on Harness itself, not bolted onto Jenkins

Visit Harness AI →

8

### Firefly

Best for: Multi-cloud shops that need IaC drift detection

Pricing14-day free trial; Essential $2,499/mo (annual, up to 20K assets)

+Discovers cloud resources created outside your IaC

+Flags drift between live state and your Terraform or Pulumi definitions

+Generates remediation code to sync live state back to your IaC

−$2,499/month is priced for mid-to-large teams, not a small startup

−The 20,000-asset cap fills fast since every security group, IAM role, and tag counts separately

Visit Firefly →

9

### BigPanda

Best for: Enterprise NOC teams drowning in alert noise

PricingCustom quote; no public pricing

+Ingests alerts from Datadog, Prometheus, CloudWatch, and Splunk and correlates them

+Cuts forty alerts for one outage down to a single actionable incident

+L1 Agent can auto-triage and even auto-remediate known incident patterns

−No public pricing, so every deal starts with a sales demo

−Overkill if your stack is small enough for one engineer to eyeball the alert channel

Visit BigPanda →

## What it is

AI for DevOps is not one product but three overlapping toolsets. The first drafts and reviews infrastructure-as-code: describe what you want in plain language and the tool writes Terraform, Pulumi, or a Helm chart as a reviewable diff, then flags risky changes like a public S3 bucket before a human looks.

The second reads live Kubernetes state, scanning pods and controllers to explain why something is broken right now, not just what the YAML implies.

The third is AIOps: it ingests alerts, logs, and traces from every monitoring tool you run and correlates them into a much smaller set of actionable incidents, then summarizes timelines and suggests likely root causes.

Some tools live inside a platform you already pay for like Datadog or PagerDuty, turning on as an add-on; others are standalone like k8sGPT and BigPanda. What they share is a goal: cut the toil between a signal and an explanation, without owning the decision about what to do next.

## Why it matters

The wrong pick either does nothing for your problem or locks you into a stack you don't run. An IaC generator is useless at 3 a.m. when you're drowning in pages, and an AIOps platform won't write you a VPC module.

Category fit comes first: match the tool to the job, or you're paying for a feature you'll never open.

Cost and lock-in come next. The spread runs from free to thousands a month: k8sGPT plus Claude Pro covers Kubernetes diagnostics and IaC drafting for under $20, while Firefly Essential is $2,499/month and BigPanda sells only by quote.

Pulumi Neo only helps if you run Pulumi; Harness AI only helps if you run Harness. The fastest, cheapest wins come from turning on AI inside tools you already pay for, then filling gaps with something free.

## Key features to look for

Category fitEssential
The tool has to match the job. An IaC generator won't triage an incident, and an AIOps platform won't write a Terraform module. Decide whether you need code generation, Kubernetes diagnosis, or alert correlation before comparing features.

Reviewable output, not blind applyEssential
Good tools produce a diff or pull request you review before anything changes. Treat generated infrastructure code as a first draft, run it through plan and a scanner, and never let an agent apply to stateful resources unattended.

Live state awareness
Reads your actual cloud or cluster state, not just the config file. k8sGPT introspects a running cluster; general coding assistants only see the YAML unless you wire them to run terraform plan or kubectl against live infrastructure.

Transparent pricing
You can budget without a sales call. Flat per-seat pricing like Cursor's is easy to forecast, while credit models like Datadog's and Pulumi's and quote-only enterprise tiers like Harness and BigPanda make costs harder to predict.

Stack integration
Turning on AI inside a tool you already pay for beats adding a new vendor. Datadog Bits and PagerDuty Advance are add-ons to platforms many teams already run, so the setup cost and procurement friction are close to zero.

Data and security boundary
Know what leaves your environment. k8sGPT sends cluster error output, which can include sensitive config, to an LLM unless you run a local model, and AI-generated IaC can open a security group too broadly if merged without a scan.

Mistakes to avoid

×Buying by the category label. "AI for DevOps" spans three unrelated jobs; a tool that drafts Terraform does nothing for alert fatigue, and vice versa. Pick for the specific job, not the buzzword.

×Applying AI-generated infrastructure code directly. Cursor and Claude Code will happily generate a change that forces a destructive replace, because they don't know your live cloud state. Always run plan and a scanner first.

×Adding a new vendor before checking your stack. On Datadog or PagerDuty, the AI is a credits add-on or partially included, not a separate purchase. Teams often buy a standalone tool when they already own the capability.

Expert tips

→Start with the job, then the budget. k8sGPT plus Claude Pro covers Kubernetes diagnostics and IaC drafting for under $20/month before you consider any enterprise contract.

→Wire coding assistants to your live state. Give Cursor or Claude Code a state-reading MCP server or shell access to run terraform plan, so they stop suggesting changes that already exist or destroy resources.

→Pair AI-generated IaC with automated scanning. Run every change through tfsec, Checkov, or OPA/Sentinel and require human review before apply, the same discipline you'd give a PR from a new hire.

## The bottom line

There's no single best AI for DevOps, because the category is really three jobs. For writing and reviewing infrastructure code, Pulumi AI (Neo) leads if you run Pulumi, and Cursor or Claude Code cover everyone else editing Terraform and Helm in the repo.

For Kubernetes, k8sGPT is the standout: free, open source, and purpose-built for reading live cluster state.

For 3 a.m. pages and alert noise, PagerDuty Advance handles on-call summaries at a per-seat price, while BigPanda is the enterprise answer once alert correlation across many monitoring tools becomes a real problem. Datadog Bits and Firefly pay off only if you already live in Datadog or need serious multi-cloud drift governance.

Start with the free or already-paid options, then add a specialist tool for the one gap that keeps costing you time.

## Frequently asked questions

What is the best AI for DevOps in 2026?

There's no single winner, since the category spans three jobs. For infrastructure code, Pulumi AI and Claude Code lead. For Kubernetes diagnostics, k8sGPT is the most purpose-built and it's free. For incident response, PagerDuty Advance and BigPanda are strongest, at very different price points.

Is there a free or open-source option for DevOps?

Yes. k8sGPT is fully open source under Apache-2.0 with no paid tier; you just supply an LLM API key or run a local model. Pulumi's Free plan covers one user at no cost, and Cursor's Hobby plan covers light IaC editing.

Can I trust AI-generated infrastructure code in production?

Treat it as a first draft, never something to apply directly. Run every change through plan, scan it with tfsec, Checkov, or OPA/Sentinel, and require human review before apply on anything stateful. Cursor and Claude Code don't know your live cloud state unless you wire it in.

How much should a team budget for AI DevOps tooling?

Anywhere from $0 to thousands a month. k8sGPT plus Claude Pro runs under $20; Firefly Essential is $2,499/month, and Harness or BigPanda enterprise contracts cost more. Start with free or already-paid tools in your stack, then move up once a specific, recurring pain justifies it.

Does AI understand Kubernetes, or just generic YAML?

General coding assistants write manifests well but don't introspect a live cluster unless you give them access to run kubectl. k8sGPT is purpose-built for this: it reads actual cluster state and explains what's broken now, not just what the YAML implies.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Pulumi AI pricing](https://pulumi.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Datadog Bits AI pricing](https://datadoghq.com/pricing)checked Sep 2026
- [PagerDuty AI pricing](https://pagerduty.com/pricing)checked Sep 2026
- [k8sGPT pricing](https://k8sgpt.ai)checked Sep 2026
- [Harness AI pricing](https://harness.io/pricing)checked Sep 2026
- [Firefly pricing](https://firefly.ai/pricing)checked Sep 2026
- [BigPanda pricing](https://bigpanda.io/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsAi For Code ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best API Documentation Tools in 2026

URL: https://devshot.co/reviews/best-api-documentation-tools
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Mintlify, ReadMe, Redocly, Stoplight, Bump.sh, Postman, Scalar, GitBook and Docusaurus compared on price, OpenAPI handling and who each one actually fits…

Expert Guide

## The Best API Documentation Tools in 2026

Nine platforms compared on 2026 pricing, OpenAPI support and hosting model, from a free self-hosted generator to a flat-rate AI-native docs platform.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 9 tools compared

TL;DR

Short answer: Mintlify is the best all-round pick for a team that wants a polished, AI-native docs site fast. It is free for a small team on Starter and $450/mo on Pro once you need automations and an AI agent. ReadMe ($250/mo Pro) suits teams that want an interactive reference tied to real API keys. Redocly and Scalar fit teams that generate docs straight from an OpenAPI file and want per-seat pricing instead of a flat plan. Docusaurus is free forever if you can self-host and write in Markdown.

Key facts9 tools compared: Mintlify, ReadMe, Redocly, Stoplight, Bump.sh, Postman, Scalar, GitBook…

- Updated: September 25, 2026
- Top pick: Mintlify (best for: Teams that want a polished, AI-native docs site without per-seat billing)
- Top pick price as of September 25, 2026: Mintlify: Starter free (5 editor seats); Pro $450/mo (unlimited seats); Enterprise custom.

- 9 tools compared: Mintlify, ReadMe, Redocly, Stoplight, Bump.sh, Postman, Scalar, GitBook, Docusaurus
- ReadMe (best for: Teams that want an interactive reference readers can test with their own API key): Starter free (1 admin); Pro $250/mo billed annually (5 admins, $20 per extra); Enterprise custom.
- Redocly (best for: Spec-first teams that want per-seat pricing instead of a flat platform fee): Pro $10/seat/mo (1 project, 100 pages); Enterprise $24/seat/mo; Enterprise+ custom.
- Stoplight (best for: Teams that design the API and publish the docs in the same tool): Free (1 user); Basic from $44/mo; Startup from $113/mo; Pro Team from $362/mo (annual); Enterprise custom.

Every tool on this list turns an OpenAPI spec or a set of Markdown pages into a docs site a developer can read and test against.

The split is in who hosts it, how the price scales with seats or endpoints, and how much of the writing an AI agent now does for you, not in whether the result looks clean.

Devshot data: [Toolradar](https://toolradar.com/best/api-tools), the software directory we run, evaluated 465 tools for its September 2026 ranking of API development, testing and documentation software, fetched today.

This page narrows that field to the nine tools built specifically for publishing API reference docs, the cut a general API-tools list usually buries under testing clients.

Prices below were checked on each vendor's own pricing page today, 25 September 2026.

Swagger UI and Swagger Editor are still free and open source under SmartBear, but they render a spec rather than host or manage a docs site, so we cover the SmartBear-owned tool that does that job, Stoplight, instead of the bare renderer.

## Top Picks

Based on features, real-world fit, and value for money.

Best API Documentation Tools in 2026: 9 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Mintlify | Starter free (5 editor seats); Pro $450/mo (unlimited seats); Enterprise custom. | Teams that want a polished, AI-native docs site without per-seat billing |

ReadMe | Starter free (1 admin); Pro $250/mo billed annually (5 admins, $20 per extra); Enterprise custom. | Teams that want an interactive reference readers can test with their own API key |

Redocly | Pro $10/seat/mo (1 project, 100 pages); Enterprise $24/seat/mo; Enterprise+ custom. | Spec-first teams that want per-seat pricing instead of a flat platform fee |

Stoplight | Free (1 user); Basic from $44/mo; Startup from $113/mo; Pro Team from $362/mo (annual); Enterprise custom. | Teams that design the API and publish the docs in the same tool |

Bump.sh | Basic $50/mo (10 API docs, 5 MCP tools); Pro $120/mo (30 API docs, 50 MCP tools); Custom for SSO and on-premise. | Teams that want automatic changelogs generated from spec diffs |

[Postman](https://toolradar.com/tools/postman) | Free (Postman-branded docs); Solo $9/mo billed annually; Team $19/user/mo billed annually; Enterprise custom. | Teams already testing APIs in Postman who want docs in the same workspace |

Scalar | Free (1 editor seat, up to 3 APIs); Pro from $125/mo annual; Business from $500/mo annual; Enterprise custom. | Teams that want a modern reference plus generated client SDKs from one spec |

[GitBook](https://toolradar.com/tools/gitbook) | Free (per site); Premium $65/site/mo, Ultimate $249/site/mo (annual), plus $12/user/mo; Enterprise custom. | Teams that want product docs and API reference in one shared knowledge base |

Docusaurus | Free and open source; you pay only for hosting (commonly $0 on Vercel or Netlify's free tier). | Teams comfortable self-hosting a static site who do not want a recurring docs bill |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Mintlify

Top Pick

Best for: Teams that want a polished, AI-native docs site without per-seat billing

PricingStarter free (5 editor seats); Pro $450/mo (unlimited seats); Enterprise custom.

+Starter is free for up to 5 editor seats, with a custom domain and an MCP server included, not gated to a paid tier

+Pro removes the seat cap entirely at one flat price instead of charging per person added

+The web editor, API playground and authentication docs ship on every plan, including free

−The jump from free to Pro is steep, with no smaller paid tier in between for a growing team

−Automations, the AI Agent and Admin APIs are Pro-only, so a 6-person team pays the full jump for one extra seat

Visit Mintlify →

2

### ReadMe

Best for: Teams that want an interactive reference readers can test with their own API key

PricingStarter free (1 admin); Pro $250/mo billed annually (5 admins, $20 per extra); Enterprise custom.

+The interactive API reference and bi-directional Git sync are on the free Starter plan, not held back for Pro

+Pro's multiplayer editing and branching let several writers work the same docs without overwriting each other

+Ask AI Lite and an AI Linter are included on Pro, not a separate paid add-on

−Starter allows only 1 admin seat, so any second writer forces an upgrade to Pro

−Extra Pro seats cost $20 each on top of the base plan, which adds up past a five-person team

Visit ReadMe →

3

### Redocly

Best for: Spec-first teams that want per-seat pricing instead of a flat platform fee

PricingPro $10/seat/mo (1 project, 100 pages); Enterprise $24/seat/mo; Enterprise+ custom.

+At $10/seat/mo, a 3-person team pays $30/mo, cheaper than most flat-fee competitors' entry paid plan

+Reunite, the core docs product, comes with custom domain hosting on the entry Pro tier

+MCP servers and AI search are included at the $24/seat Enterprise tier, not a separate SKU

−Pro caps you at 1 project and 100 pages; a second product line needs the pricier Enterprise tier

−The optional Revel and Reef add-ons each add a further per-seat charge, so a fully loaded seat costs well past the entry rate

Visit Redocly →

4

### Stoplight

Best for: Teams that design the API and publish the docs in the same tool

PricingFree (1 user); Basic from $44/mo; Startup from $113/mo; Pro Team from $362/mo (annual); Enterprise custom.

+Free covers a solo user with the visual API editor, docs and instant mock servers, no card required

+Every included-user count (3, 8, 15) already covers a small team without buying extra seats immediately

+Pro Team adds self-hosted Git providers, LDAP and SAML SSO for larger orgs standardizing on one platform

−Monthly billing runs noticeably higher than annual at every tier, for example Basic at $56/mo instead of $44/mo

−Custom domains and private projects require Startup at $113/mo, not the entry Basic plan

Visit Stoplight →

5

### Bump.sh

Best for: Teams that want automatic changelogs generated from spec diffs

PricingBasic $50/mo (10 API docs, 5 MCP tools); Pro $120/mo (30 API docs, 50 MCP tools); Custom for SSO and on-premise.

+Automatic changelog generation from spec diffs is a built-in feature, not a manual writing task

+Both Basic and Pro include a custom domain and logo, no branding-removal upcharge

+Full human support ships on every tier, including the entry Basic plan

−Basic caps you at 10 API docs total, tight for a company running microservices with separate specs per service

−The API Explorer and private MCP servers require the Pro tier, not Basic

Visit Bump.sh →

6

### Postman

Best for: Teams already testing APIs in Postman who want docs in the same workspace

PricingFree (Postman-branded docs); Solo $9/mo billed annually; Team $19/user/mo billed annually; Enterprise custom.

+Solo is the lowest paid entry price of any tool covered here, billed annually

+Documentation lives next to the same collections and environments your team already tests APIs in

+Custom-branded docs with unlimited custom domains are included from Solo, not a higher tier

−The free plan publishes docs with visible Postman branding, unlike free tiers on Mintlify, ReadMe or Stoplight

−Team billing is per user at $19/mo each, so a 10-person team runs $190/mo just for seats

Visit Postman →

7

### Scalar

Best for: Teams that want a modern reference plus generated client SDKs from one spec

PricingFree (1 editor seat, up to 3 APIs); Pro from $125/mo annual; Business from $500/mo annual; Enterprise custom.

+Free already includes 1 generated SDK and MCP servers for up to 3 APIs, not stripped down to docs alone

+Annual billing saves 2 months versus paying monthly at every paid tier

+Business adds SSO and subpath hosting, useful for embedding docs under an existing domain path

−Additional SDKs beyond the one included cost $150/mo or $600/mo depending on endpoint count, on every paid plan

−The jump from Pro to Business is 4x the price for the same feature set plus SSO and 10 more endpoints

Visit Scalar →

8

### GitBook

Best for: Teams that want product docs and API reference in one shared knowledge base

PricingFree (per site); Premium $65/site/mo, Ultimate $249/site/mo (annual), plus $12/user/mo; Enterprise custom.

+Free already includes interactive API playgrounds and GitHub or GitLab sync, not held back for a paid plan

+One license covers general product docs and API reference together, useful if you do not want two separate tools

+Annual billing includes 2 months free on every paid tier

−Pricing is per site plus per user, so a company running docs for three products pays the $65 or $249 site fee three times over

−AI Assistant and adaptive content require Ultimate at $249/site/mo, a large step past Premium

Visit GitBook →

9

### Docusaurus

Best for: Teams comfortable self-hosting a static site who do not want a recurring docs bill

PricingFree and open source; you pay only for hosting (commonly $0 on Vercel or Netlify's free tier).

+No license cost at any team size, unlike every other tool on this list past a free seat cap

+Content is plain Markdown and MDX in your own Git repo, so there is no vendor lock-in on the writing side

+A large plugin ecosystem covers versioning, search and internationalization without a paid add-on

−There is no built-in interactive Try It panel or API key auth; you add a plugin or a separate API explorer yourself

−Hosting, custom domains, SSL and uptime are entirely your team's responsibility, not a vendor SLA

Visit Docusaurus →

## What it is

An API documentation tool takes an OpenAPI or AsyncAPI spec, or a set of hand-written pages, and publishes a site developers can read, search and often call live through a built-in Try It panel.

Some tools, like [Mintlify](https://toolradar.com/tools/mintlify) or [Scalar](https://toolradar.com/tools/scalar), generate that site from your spec on every deploy.

Others give writers a visual editor and sync the result back to your Git repo.

Most now bundle an MCP server, a way for an AI coding agent to query your docs directly instead of scraping the rendered page, and several add an in-house AI assistant that drafts or edits pages from a prompt.

If your team also tests the API itself, not just documents it, our API testing tools guide covers that separate buying decision.

## Why it matters

The bill here scales two different ways. Mintlify, ReadMe, Scalar and GitBook charge per site or per platform tier, so the cost jumps once you outgrow the free seat count.

Redocly and [Postman](https://toolradar.com/tools/postman) charge per seat, which is cheaper for a two-person team and pricier once ten people need write access.

Hosting model matters just as much as price.

Docusaurus costs nothing to license but you pay for and maintain the hosting yourself. Every other tool here hosts the site for you, which is the convenience you are actually buying once the price climbs past a free tier.

Teams that also need to design and mock the API before writing a line of docs should compare our API gateways guide and GraphQL tools guide for the adjacent buying decisions.

## Key features to look for

OpenAPI and AsyncAPI handlingEssential
Redocly, Stoplight, Bump.sh and Scalar are built around a spec file and regenerate the docs whenever it changes. Mintlify, ReadMe and GitBook accept a spec too but lean more on hand-written Markdown pages around it.

Interactive Try It / API ExplorerEssential
ReadMe, Postman, Bump.sh Pro and Scalar let a reader send a real request with their own API key from the docs page. <a href="https://toolradar.com/tools/gitbook">GitBook</a> bundles a lighter version of the same panel. Docusaurus needs a plugin to get the same result.

Hosting modelEssential
Every tool except Docusaurus hosts the published site for you on a custom domain. Docusaurus builds a static site you deploy yourself, on Vercel, Netlify or your own server.

MCP server and AI features
Mintlify, ReadMe, Redocly and Bump.sh now ship an MCP server so an AI coding agent can query the docs directly, plus an in-app AI assistant for drafting pages. Docusaurus and Postman's documentation feature do not.

Seat and editor limits
Free tiers cap editors at one to five people. Mintlify's Pro plan is the outlier with unlimited editor seats at one flat monthly price instead of a per-seat charge.

Diff, changelog and version history
<a href="https://toolradar.com/tools/bump-sh">Bump.sh</a> is built specifically to diff your spec between deploys and post a changelog automatically. Redocly and <a href="https://toolradar.com/tools/stoplight">Stoplight</a> handle branching and multi-version docs; the others publish a single current version.

## Pricing

Three tools give a real free tier that covers actual usage, not a demo: Mintlify Starter (5 seats), [ReadMe](https://toolradar.com/tools/readme) Starter (1 admin) and Stoplight Free (1 user), plus Scalar and GitBook at the low end.

[Docusaurus](https://toolradar.com/tools/docusaurus) costs nothing to license at any size, since you host it yourself.

Postman has the cheapest named paid entry price of any tool here, though its free docs carry Postman branding that the others do not force on you.

Bump.sh is the only tool here with no free tier at all, starting after a 14-day trial.

At the top, Mintlify Pro and Scalar Business are the most expensive named flat-rate plans, while [Redocly](https://toolradar.com/tools/redocly) and Postman scale per seat instead, which is cheaper for a two or three-person team and can cost more once ten or more people need write access.

Plan | Price | Best for |

Mintlify Starter | Free | 5 editor seats, custom domain, MCP server included |

Mintlify Pro | $450/mo | Unlimited seats, AI agent, automations and admin APIs |

ReadMe Starter | Free | 1 admin, interactive API reference, custom domain |

ReadMe Pro | $250/mo, billed annually | 5 admins ($20 per extra), multiplayer editing and Ask AI Lite |

Redocly Pro | $10/seat/mo | 1 project, 100 pages, custom domain hosting |

Redocly Enterprise | $24/seat/mo | 500 pages, SSO, RBAC, AI search and MCP servers |

Stoplight Free | Free | 1 user, API editor, instant mock servers |

Stoplight Basic | From $44/mo annual | 3 users included, custom domains not included |

Stoplight Startup | From $113/mo annual | 8 users, custom domains, multi-branch support |

Stoplight Pro Team | From $362/mo annual | 15 users, LDAP and SAML SSO, branding removal |

Bump.sh Basic | $50/mo | 10 API docs, 5 MCP tools, custom domain and logo |

Bump.sh Pro | $120/mo | 30 API docs, 50 MCP tools, API Explorer and rollback |

Postman Free | Free | 1 user, Postman-branded public and private docs |

Postman Solo | $9/mo billed annually | Custom-branded docs, unlimited custom domains |

Postman Team | $19/user/mo billed annually | SDK generation and distribution added |

Scalar Free | Free | 1 editor seat, up to 3 APIs, 1 SDK included |

Scalar Pro | From $125/mo annual | 5 editor seats, up to 15 APIs and MCP servers |

Scalar Business | From $500/mo annual | 10 editor seats, SSO and subpath hosting |

GitBook Free | Free | Interactive API playgrounds, GitHub or GitLab sync |

GitBook Premium | $65/site/mo plus $12/user/mo, annual | Custom domain, AI search and advanced branding |

GitBook Ultimate | $249/site/mo plus $12/user/mo, annual | AI Assistant, GitBook Agent and adaptive content |

Docusaurus | Free | Self-hosted static site generator, MIT-style open source |

Mistakes to avoid

×Picking Mintlify Pro for a 6-person team because free Starter caps out at 5 seats. Check ReadMe Pro or Redocly's per-seat plan first: both add one more seat far cheaper than the full Mintlify jump.

×Choosing Docusaurus to save money without budgeting the engineering time to build, host and maintain it. The license is free; the hosting, uptime and plugin upkeep are not, and that time has a real cost most teams skip when comparing.

×Standardizing on Postman for docs because the team already pays for API testing there. Postman's free docs carry Postman branding, and spec-first features like automatic changelogs or multi-version docs live in Bump.sh and Redocly instead.

Expert tips

→Start on a free tier before comparing paid plans. Mintlify, ReadMe, Stoplight, Scalar and GitBook all let you publish a real docs site for $0 first, so you can judge the editor and the reader experience before a bill starts.

→If your spec changes weekly, weigh Bump.sh's automatic changelog generation against the writing time you would otherwise spend updating a changelog by hand in Mintlify or ReadMe.

→For a two or three-person team, run the per-seat math on Redocly or Postman against the flat platform fees on Mintlify or Bump.sh before assuming the named entry price is the actual cost.

## The bottom line

Mintlify is the strongest default for most teams that want a modern, AI-native docs site without managing hosting, free for a small team and priced as one flat plan once you need the AI agent and automations. ReadMe is the pick when readers need to test live calls against their own key.

Redocly, Scalar and Bump.sh suit spec-first teams that regenerate docs from an OpenAPI file on every deploy, with Bump.sh the only one built specifically for automatic changelogs.

Stoplight fits teams that want API design and docs in one SmartBear-owned tool, Postman fits teams already testing there, GitBook fits a shared product-and-API knowledge base, and Docusaurus is the right call when a zero license cost matters more than a managed host.

Methodology: we shortlisted these nine from Toolradar's 465-tool API tools ranking by documentation focus, then read every vendor's own pricing page today; no placement on this page is paid. Cite this: Devshot, "Best API Documentation Tools in 2026", September 2026.

## Frequently asked questions

What is the best API documentation tool in 2026?

Mintlify for most teams: it is free for up to 5 editor seats, then one flat Pro price once you need unlimited seats and an AI writing agent. ReadMe is the better pick if readers need to test live API calls with their own key, and Redocly or Scalar fit teams that want docs generated straight from an OpenAPI spec at a lower per-seat price.

Is there a free API documentation tool?

Docusaurus is free at any team size since you self-host it, though you take on hosting and maintenance yourself. Mintlify, ReadMe, Stoplight, Scalar and GitBook all offer a real free tier too, capped at one to five editor seats rather than a time-limited trial. Bump.sh is the one tool covered here with no free tier at all.

How much does Mintlify cost?

Mintlify Starter is free for up to 5 editor seats with a custom domain and MCP server included, checked on Mintlify's pricing page 25 September 2026. Pro is $450/mo with unlimited seats, an AI agent, automations and admin APIs. Enterprise adds SSO, SCIM and a performance SLA at a custom quote. See our AI coding assistants guide for how an MCP-connected docs site plugs into an agent workflow.

ReadMe vs Mintlify: what's the real difference?

ReadMe leans on an interactive API reference where readers send real calls with their own API key, with its Pro plan capped at 5 admins plus a per-seat add-on. Mintlify leans on AI-native page generation and an MCP server for coding agents, with unlimited seats on its Pro plan instead of a per-admin cap.

Which tool is best for OpenAPI spec-driven docs?

Redocly, Stoplight, Bump.sh and Scalar all regenerate docs directly from an OpenAPI or AsyncAPI file. Bump.sh is the only one built specifically to diff the spec on every deploy and post a changelog automatically, starting at $50/mo with no free tier.

Do these tools include an MCP server for AI agents?

Mintlify, ReadMe, Redocly and Bump.sh all ship an MCP server so an AI coding agent can query the docs directly instead of scraping the rendered page. Docusaurus and Postman's documentation feature do not include one out of the box.

Is Postman good enough for API documentation on its own?

It works, and Solo at $9/mo billed annually is the cheapest paid entry price on this list, but docs are one feature inside a broader API testing platform. Free Postman docs carry Postman branding, and teams that want spec-first automation like automatic changelogs get more from Bump.sh or Redocly.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Postman pricing](https://www.postman.com/pricing/)checked Sep 2026
- [GitBook pricing](https://gitbook.com/pricing)checked Sep 2026

Related guides

Api Testing ToolsApi GatewaysCode EditorsGraphql ToolsStatic Site GeneratorsPostman AlternativesAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best API Gateways in 2026

URL: https://devshot.co/reviews/best-api-gateways
Type: review
Published: 2026-09-03
Updated: 2026-09-25
Summary: The API gateways worth running in 2026, compared on per-million request fees, control-plane subscriptions and what the open-source builds still do for free.

Expert Guide

## The Best API Gateways in 2026

AWS bills per million requests. Kong bills a control plane. The open-source ones bill the box you run them on.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

AWS API Gateway HTTP APIs are about $1 per million requests (AWS's 432 million example is $418.80). REST is $3.50 per million on the first 333 million, plus data transfer. Kong Konnect Plus starts at $25 a month for a serverless control plane, or $200 hybrid. Kong Gateway and Apache APISIX are free to self-host. Tyk has an open-source gateway; the cloud is a separate bill.

Key facts5 tools compared: AWS API Gateway, Kong Konnect, Kong Gateway (open source), Apache APISIX, Tyk

- Updated: September 25, 2026
- Top pick: AWS API Gateway (best for: APIs that already live on AWS and want a managed endpoint without running a proxy)
- Top pick price as of September 25, 2026: AWS API Gateway: From about $1 per million requests (HTTP APIs); free tier of 1 million REST and 1 million HTTP requests/mo for 12 months

- 5 tools compared: AWS API Gateway, Kong Konnect, Kong Gateway (open source), Apache APISIX, Tyk
- Kong Konnect (best for: Teams that want Kong's plugin model with a hosted control plane): From $25/month (Plus serverless control plane); open-source gateway is free, Enterprise is quote-only
- Kong Gateway (open source) (best for: Teams that will run the proxy themselves and keep config in Git): Free (Apache license); you pay for the machines and any Konnect or Enterprise add-on
- Apache APISIX (best for: Teams that want a fast, etcd-backed gateway without paying Kong prices): Free (Apache License 2.0); API7 Cloud is $250 per gateway group/month plus $2 per million calls

API gateways all terminate HTTP and apply auth, rate limits and routing. They do not sell the same unit. AWS meters requests. Kong Konnect meters a control plane, then requests after a million. Self-hosted Kong and APISIX meter nothing but CPU.

A comparison that lines up $1, $25 and $0 as if they were the same SKU is how a team picks the wrong bill.

The useful question is who runs the data plane, because that is what you are paying for when the traffic is real.

## Top Picks

Based on features, real-world fit, and value for money.

Best API Gateways in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

AWS API Gateway | From about $1 per million requests (HTTP APIs); free tier of 1 million REST and 1 million HTTP requests/mo for 12 months | APIs that already live on AWS and want a managed endpoint without running a proxy |

Kong Konnect | From $25/month (Plus serverless control plane); open-source gateway is free, Enterprise is quote-only | Teams that want Kong's plugin model with a hosted control plane |

Kong Gateway (open source) | Free (Apache license); you pay for the machines and any Konnect or Enterprise add-on | Teams that will run the proxy themselves and keep config in Git |

Apache APISIX | Free (Apache License 2.0); API7 Cloud is $250 per gateway group/month plus $2 per million calls | Teams that want a fast, etcd-backed gateway without paying Kong prices |

Tyk | Free to self-host; Tyk Cloud and licensed enterprise are custom quotes | Teams that like Tyk's middleware model and will either self-host the open-source gateway or pay for Cloud |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

1

### AWS API Gateway

Top Pick

Best for: APIs that already live on AWS and want a managed endpoint without running a proxy

PricingFrom about $1 per million requests (HTTP APIs); free tier of 1 million REST and 1 million HTTP requests/mo for 12 months

+Published request math, including a worked monthly example

+No proxy fleet to patch

+Free tier is enough to learn the product before traffic is real

−REST is more than 3x HTTP on the first tier, so the API type you pick is the price

−Data transfer is a second meter people forget

Visit AWS API Gateway →

2

### Kong Konnect

Best for: Teams that want Kong's plugin model with a hosted control plane

PricingFrom $25/month (Plus serverless control plane); open-source gateway is free, Enterprise is quote-only

+Same plugin model as OSS Kong, so you are not rewriting config to go hosted

+Plus publishes control-plane and overage numbers, which most API management suites do not

+Hybrid lets you keep the data plane on your nodes

−The $25 plan and the $200 plan are different products. Do not budget the cheap one for hybrid

−Request overage at $200 per million can beat AWS HTTP on a busy public API

Visit Kong Konnect →

3

### Kong Gateway (open source)

Best for: Teams that will run the proxy themselves and keep config in Git

PricingFree (Apache license); you pay for the machines and any Konnect or Enterprise add-on

+No per-million fee

+Huge plugin list and a lot of existing examples

+Declarative config fits CI

−You patch, scale and observe it

−Admin GUI and some enterprise plugins are not in OSS

Visit Kong Gateway (open source) →

4

### Apache APISIX

Best for: Teams that want a fast, etcd-backed gateway without paying Kong prices

PricingFree (Apache License 2.0); API7 Cloud is $250 per gateway group/month plus $2 per million calls

+Genuinely free, with a plugin model that covers auth and limits

+etcd config suits Kubernetes shops

+No vendor request meter on the Apache build

−You operate etcd and the data plane

−Commercial features live in API7, a separate product billed per gateway group and per call

Visit Apache APISIX →

5

### Tyk

Best for: Teams that like Tyk's middleware model and will either self-host the open-source gateway or pay for Cloud

PricingFree to self-host; Tyk Cloud and licensed enterprise are custom quotes

+Open-source gateway you can run without a license key

+Analytics and a dashboard exist in the commercial line if you want them

+Go codebase is easy to read if you will patch it

−Cloud pricing did not get a locked public number in this pass

−OSS versus Cloud feature split is easy to miss in a demo

Visit Tyk →

## What it is

An API gateway sits in front of services and handles TLS, auth, routing, rate limits and sometimes transformation. The category split is simple. Cloud-native gateways are a managed endpoint you point DNS at. Self-hosted gateways are a process you run next to the services. Most vendors now sell both: an OSS binary, and a control plane that bills monthly.

## Why it matters

The hidden meter is not features. It is requests versus control plane. A quiet internal API on AWS still pays per million. A noisy public API on self-hosted Kong pays for the instance. Konnect Plus includes 1 million requests, then $200 per extra million on some SKUs.

Those units do not convert cleanly, which is why a "cheapest gateway" table is usually wrong.

## Key features to look for

Billing unit
Per million requests, a monthly control plane, or the VM you already have. This is the invoice.

Who runs the data plane
AWS runs theirs. Self-hosted Kong and APISIX run on your nodes. Hybrid Konnect splits the two.

Auth plugins
JWT, API keys, OIDC, mTLS. A gateway without the plugin you need is a reverse proxy with extra steps.

Rate limits and quotas
Whether limits are per consumer, per route and consistent across instances. This is the feature teams actually turn on first.

Config as code
Declarative YAML or Terraform versus a UI that drifts. CI only works if the gateway is a file.

## Pricing

AWS API Gateway and Kong Konnect Plus publish prices. API7 Cloud publishes $250 per gateway group a month plus $2 per million calls. Kong Enterprise, API7 Enterprise, Tyk Cloud and Tyk licensed enterprise are quote-only. Kong Gateway, Apache APISIX and Tyk Gateway are free to self-host, so the bill is the machines and the people who run them.

The cheapest published paid entry is AWS HTTP APIs at about $1 per million requests, with a free tier of 1 million REST and 1 million HTTP requests a month for 12 months. Kong Konnect Plus starts at $25 a month for a serverless control plane, or $200 a month for hybrid.

Costs jump on REST at $3.50 per million for the first 333 million plus $0.09/GB in US East, on a dedicated cloud gateway at $500 a month plus $0.15/GB, and on Plus overage at $200 per additional million after 1 million included requests (checked 26 August 2026).

Plan | Price | Best for |

AWS API Gateway HTTP APIs | About $1 per million requests | Per million HTTP API requests |

AWS API Gateway HTTP example | $418.80 for 432 million requests | AWS monthly example for HTTP APIs |

AWS API Gateway REST APIs | $3.50 per million (first 333 million) | First 333 million, then volume tiers |

AWS API Gateway data transfer | $0.09/GB in US East | Data transfer in US East on top of requests |

AWS API Gateway free tier | Free for 12 months | 1 million REST and 1 million HTTP requests a month |

Kong Konnect Plus Serverless | $25/month | Plus serverless control plane |

Kong Konnect Plus Hybrid | $200/month | Plus hybrid control plane |

Kong Konnect Dedicated Cloud Gateway | $500/month plus $0.15/GB | Dedicated Cloud Gateway plus data transfer |

Kong Konnect Plus overage | $200 per additional million | 1 million requests included, then overage on Plus |

Kong Konnect Enterprise | Custom quote | Enterprise features are sales-quoted |

Kong Gateway (open source) | Free | Apache-licensed; pay for machines and Konnect or Enterprise |

Apache APISIX | Free | Apache License 2.0; you operate the gateway |

API7 Cloud | $250/gateway group/month + $2 per million calls | Hosted APISIX control plane, plus $10 per service; 99.95% SLA |

API7 Enterprise | Custom quote | Separate commercial product; confirm the price with API7 |

Tyk Gateway | Free | Open-source gateway you run yourself |

Tyk Cloud | Custom quote | Confirm current Cloud dollars on tyk.io/pricing |

Tyk licensed enterprise | Custom quote | Licensed enterprise is a commercial product |

Mistakes to avoid

×Picking REST API Gateway by default on AWS. HTTP APIs are the $1-per-million product. REST is $3.50 before transfer.

×Budgeting Kong Konnect at $25 when you need hybrid or a dedicated cloud gateway. Those are $200 and $500 starting points.

×Calling a self-hosted gateway free after you ignore the people who keep it up.

Expert tips

→Estimate AWS from request type and transfer, not from the homepage. REST plus $0.09/GB changes the story on large payloads.

→On Konnect, read whether the SKU is serverless control, hybrid or dedicated cloud gateway before you compare it to AWS.

→Start on OSS Kong or APISIX if traffic is internal and you already run Kubernetes. Pay the hosted control plane when the admin API becomes the bottleneck.

## The bottom line

Use AWS API Gateway HTTP APIs when the services are already on AWS and you want a published per-million price. Use Kong Konnect when you want Kong plugins and a hosted control plane, and budget hybrid or dedicated if that is what the demo used.

Run Kong Gateway or Apache APISIX yourself when you will staff the proxy. Treat Tyk Cloud as a separate quote from the free Tyk gateway.

## Frequently asked questions

What is the best free API gateway?

Kong Gateway OSS, Apache APISIX and Tyk Gateway are all free to self-host. You pay for compute and for the person who operates them. None of those is a managed AWS endpoint.

How much does AWS API Gateway cost?

Checked 26 August 2026: HTTP APIs about $1 per million requests (AWS example: 432 million requests is $418.80). REST $3.50 per million on the first 333 million, plus data transfer. Free tier is 1 million REST and 1 million HTTP requests per month for 12 months.

How much does Kong Konnect cost?

Checked 26 August 2026: Plus serverless control plane $25 a month, hybrid $200, dedicated cloud gateway $500 plus $0.15/GB. Published Plus SKUs include 1 million requests, then $200 per extra million. Enterprise is sales-quoted. OSS Kong is free.

API gateway versus API security tool?

A gateway routes and limits. A WAF or scanner inspects for abuse. You often want both. They are not the same invoice. See Cyberpresso's API security list if the job is inspection.

Related guides

Openrouter AlternativesApi Testing ToolsServerless PlatformsCi Cd ToolsFree LLM APIs in 2026: Rate Limits and Training TermsLlm ApisAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best API Testing Tools in 2026

URL: https://devshot.co/reviews/best-api-testing-tools
Type: review
Published: 2026-08-12
Updated: 2026-09-25
Summary: The API testing tools engineers actually use in 2026: Postman, Bruno, Insomnia, Hoppscotch and Apidog compared on price, Git workflow, CI and offline support.

Expert Guide

## The Best API Testing Tools in 2026

Clients and test runners that keep your requests versioned in Git instead of trapped in someone's laptop.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

The best API testing tools in 2026 are Postman if you want the full platform and your team already lives in it, Bruno if you want your collections in Git as plain text files, Insomnia for a clean client with a generous free tier, Hoppscotch for a fast open-source web client you can self-host, and Apidog if you want design, mocking and testing in one tool. The real decision is not features, it is whether your requests live in a vendor cloud or in your repo.

Key facts5 tools compared: Postman, Bruno, Insomnia, Hoppscotch, Apidog

- Updated: September 25, 2026
- Top pick: Postman (best for: Teams that want one platform for the whole API lifecycle)
- Top pick price as of September 25, 2026: Postman: From $9/mo (Solo, billed annually); free plan yes ($0)

- 5 tools compared: Postman, Bruno, Insomnia, Hoppscotch, Apidog
- Bruno (best for: Teams that want collections in the repo, reviewed like code): From $6/user/mo (Pro, billed annually); free open source; 14-day trial, no card
- Insomnia (best for: Small teams that want a clean client without paying immediately): From $12/user/mo (Pro, billed annually); Essentials free
- Hoppscotch (best for: Fast browser-based testing and self-hosted deployments): Free on cloud or self-hosted (open source); Organization $6/user/mo billed annually

Every backend team ends up with the same problem: the requests that prove the API works exist only in one engineer's client, in a workspace nobody else can see. The tools below all send HTTP requests competently.

Where they differ, and where the last two years have reshaped the category, is how collections are stored, whether they run in CI, and how much of your workflow is locked to a vendor account.

## Top Picks

Based on features, real-world fit, and value for money.

Best API Testing Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Postman](https://toolradar.com/tools/postman) | From $9/mo (Solo, billed annually); free plan yes ($0) | Teams that want one platform for the whole API lifecycle |

[Bruno](https://toolradar.com/tools/bruno) | From $6/user/mo (Pro, billed annually); free open source; 14-day trial, no card | Teams that want collections in the repo, reviewed like code |

[Insomnia](https://toolradar.com/tools/insomnia) | From $12/user/mo (Pro, billed annually); Essentials free | Small teams that want a clean client without paying immediately |

[Hoppscotch](https://toolradar.com/tools/hoppscotch) | Free on cloud or self-hosted (open source); Organization $6/user/mo billed annually | Fast browser-based testing and self-hosted deployments |

[Apidog](https://toolradar.com/tools/apidog) | Free tier yes; paid team plans, price not published | Teams that want API design, mocking and testing in one place |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Postman

Top Pick

Best for: Teams that want one platform for the whole API lifecycle

PricingFrom $9/mo (Solo, billed annually); free plan yes ($0)

+Broadest feature set by a distance, including monitors, mock servers and generated docs

+Native Git support now available even on the free tier

+Enormous ecosystem: most vendors publish an official Postman collection

−Heavy application that has grown slower with each platform addition

−Historically pushed cloud sync over local files, which soured many teams

Visit Postman →

2

### Bruno

Best for: Teams that want collections in the repo, reviewed like code

PricingFrom $6/user/mo (Pro, billed annually); free open source; 14-day trial, no card

+Collections are files in your repo, so pull requests review API changes like any other code

+Fully offline, no mandatory account or cloud sync

+Fast and light compared with the incumbents

−Younger ecosystem, fewer prebuilt vendor collections

−Free tier caps you at 2 workspaces and public Git only

Visit Bruno →

3

### Insomnia

Best for: Small teams that want a clean client without paying immediately

PricingFrom $12/user/mo (Pro, billed annually); Essentials free

+The most generous free tier of the commercial options, including Git Sync for up to 3 users

+Clean, fast interface that stayed focused while Postman expanded

+First-class GraphQL and gRPC support

−Ownership by Kong pushed some users away after the 2023 account requirement controversy

−Enterprise pricing jumps sharply from Pro

Visit Insomnia →

4

### Hoppscotch

Best for: Fast browser-based testing and self-hosted deployments

PricingFree on cloud or self-hosted (open source); Organization $6/user/mo billed annually

+Runs in the browser with nothing to install, useful for quick checks and onboarding

+Fully open source and self-hostable, which solves most compliance objections

+Supports REST, GraphQL, WebSocket, SSE and MQTT in one interface

−Fewer collaboration and governance features than the commercial platforms

−Self-hosting means you own the upgrades and backups

Visit Hoppscotch →

5

### Apidog

Best for: Teams that want API design, mocking and testing in one place

PricingFree tier yes; paid team plans, price not published

+Design-first workflow keeps the OpenAPI spec, mocks and tests in sync automatically

+Smart mock generation produces realistic data without hand-written rules

+Bundles what would otherwise be three tools

−Less known outside its core user base, so fewer community resources

−The all-in-one approach means you adopt its whole workflow or little of it

Visit Apidog →

## What it is

An API testing tool combines an API client (compose a request, inspect the response, manage environments and auth) with a test runner (assert on responses, chain requests, run the suite in CI).

Modern entrants add mock servers, OpenAPI import and sync, and collaboration. The category split cleanly in 2024 to 2025 between cloud-first platforms and local-first tools that keep everything in files you commit.

## Why it matters

API tests are the cheapest integration tests you will ever write, and the fastest way to catch a breaking change before a client does. But they only pay off if the whole team can run them, which means collections have to be shared, versioned alongside the code they test, and executable in your pipeline.

A tool that keeps your requests in a private cloud workspace quietly turns your test suite into tribal knowledge.

## Key features to look for

Git-native storage
Collections stored as plain text in your repo, reviewed in pull requests and versioned with the code. The single biggest workflow difference between these tools.

CLI and CI execution
A command-line runner so the same suite runs on every pull request. Without it, API tests are a manual ritual nobody performs under deadline.

Environment and secret handling
Per-environment variables with secrets kept out of the committed files. Check how each tool separates the two before you commit anything.

OpenAPI import and sync
Generate the collection from your spec and keep it in sync as the spec changes, instead of hand-maintaining two sources of truth.

Offline and self-host options
Whether the tool works with no account and no network. For regulated environments this is often the deciding constraint.

## Pricing

Postman, Bruno, Insomnia and Hoppscotch publish list prices for their paid seats, while Apidog's paid team plans and Postman Enterprise are not listed. All five have a free path, and Bruno's paid plans include a 14-day trial with no card required.

The cheapest paid entry is a tie at $6 per user per month billed annually, Bruno Pro and Hoppscotch Organization, then Postman Solo at $9 a month billed annually and Insomnia Pro at $12 per user per month billed annually.

Costs jump on the top seats: Bruno Ultimate is $11 per user per month, Postman moves from Team at $19 to a quote-only Enterprise plan, and Insomnia Enterprise is $45 per user per month.

Plan | Price | Best for |

Postman Free | $0 | 1 user, 50 AI credits/mo, 10,000 API calls/mo |

Postman Solo | $9/mo billed annually | Solo plan, billed annually |

Postman Team | $19/user/mo billed annually | Per-user team plan, 400 AI credits per user |

Postman Enterprise | Custom quote | 800 pooled AI credits per user, API Catalog, advanced RBAC |

Bruno Open source | $0 | 2 workspaces, public Git |

Bruno Pro | $6/user/mo billed annually | 14-day trial on paid plans, no card required |

Bruno Ultimate | $11/user/mo | Ultimate paid tier |

Insomnia Essentials | Free, 1,000 mock requests/mo | Unlimited local and cloud projects, Git Sync for up to 3 users |

Insomnia Pro | $12/user/mo billed annually | Pro plan, billed annually |

Insomnia Enterprise | $45/user/mo | Enterprise per-user plan |

Hoppscotch self-host | Free, open source | Free to self-host |

Hoppscotch Organization | $6/user/mo billed annually | Admin dashboard, dedicated support, custom payment options |

Apidog Free | Free tier | Free tier included |

Apidog team plans | Custom quote | Paid team plans, check current pricing |

Mistakes to avoid

×Letting collections live only in a personal cloud workspace. When that engineer leaves, the institutional knowledge of how the API is exercised leaves with them.

×Committing environment files with real secrets. Every tool here separates variables from secrets; configure that before the first commit, not after the leak.

×Writing tests that only assert on status codes. A 200 with a malformed body is the failure users actually notice.

×Skipping the CLI. If the suite does not run on every pull request, it is documentation, not testing.

Expert tips

→Import your OpenAPI spec rather than hand-building collections, then keep the spec as the source of truth so drift cannot start.

→Store collections next to the service they test, in the same repo, so a breaking change and its test update land in one pull request.

→Run the smoke subset on every pull request and the full suite nightly. Fast feedback beats complete feedback in CI.

→Trial Bruno or Hoppscotch for a week before renewing a large per-seat contract. Most teams use a fraction of the platform they pay for.

## The bottom line

If your team is already deep in Postman and uses monitors and docs, stay. If you are choosing fresh in 2026, Bruno is the one to try first: collections as files in Git is the workflow most teams end up wanting, at a tied-lowest price per seat.

Insomnia is the best compromise for small teams on a free tier, and Hoppscotch wins the moment self-hosting is a requirement.

## Frequently asked questions

What is the best free API testing tool?

Bruno's open-source version is the strongest free option if you want collections in Git, with no account required and no cloud dependency. Insomnia's Essentials tier is the most generous commercial free plan, including Git Sync for up to 3 users. Hoppscotch is fully open source and self-hostable at no cost.

Is Postman still worth it in 2026?

Yes if you use the platform, not just the client. Monitors, mock servers, generated documentation and the ecosystem of official vendor collections are real advantages nothing else matches. If you only send requests and run tests in CI, you are paying for a lot of surface area you do not touch, and Bruno or Insomnia will feel faster.

What is the difference between Bruno and Postman?

Storage and philosophy. Bruno keeps every collection as plain text files in your repository, so API tests are versioned and code-reviewed like source code, and it works fully offline. Postman is a cloud platform first, with far more features (monitoring, docs, mocks) but historically a stronger pull toward its own sync. Bruno also costs less per seat.

Can API tests run in CI/CD?

Yes, and they should. Postman ships Newman and its own CLI, Insomnia has Inso CLI, and Bruno has a native CLI runner. The usual pattern is a fast smoke suite on every pull request and the full regression suite on a nightly schedule, with secrets injected from your CI provider rather than committed.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Postman pricing](https://www.postman.com/pricing/)checked Sep 2026
- [Bruno pricing](https://usebruno.com/pricing)checked Sep 2026
- [Insomnia pricing](https://insomnia.rest/pricing)checked Sep 2026
- [Hoppscotch pricing](https://hoppscotch.io/pricing)checked Sep 2026
- [Apidog pricing](https://apidog.com/pricing)checked Sep 2026

Related guides

Graphql ToolsWeb Scraping ToolsSelf Hosting ToolsPostman AlternativesAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best CI/CD Tools in 2026

URL: https://devshot.co/reviews/best-ci-cd-tools
Type: review
Published: 2026-08-20
Updated: 2026-09-25
Summary: The best CI/CD tools in 2026, compared on billing unit rather than feature grids. GitHub Actions, GitLab CI, CircleCI, Buildkite and Jenkins…

Expert Guide

## The Best CI/CD Tools in 2026

Five pipelines compared on the unit each one actually bills, because $15, $29 and $30 a month buy completely different things here.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

The headline prices in this category are not comparable, because no two vendors sell the same unit. CircleCI bills compute credits, Buildkite bills active users plus concurrent agents, GitLab bills seats, and GitHub Actions bills minutes against an included allowance. A team of eight running heavy pipelines and a team of forty running light ones will get opposite answers from the same table.

Start from your build minutes and your concurrency, not from the plan names.

Key facts5 tools compared: GitHub Actions, GitLab CI, CircleCI, Buildkite, Jenkins

- Updated: September 25, 2026
- Top pick: GitHub Actions (best for: Teams already on GitHub who want zero additional vendors)
- Top pick price as of September 25, 2026: GitHub Actions: Free for public repositories; Team $4/user/mo (first 12 months) with 3,000 minutes, then Linux runners from $0.002/min

- 5 tools compared: GitHub Actions, GitLab CI, CircleCI, Buildkite, Jenkins
- GitLab CI (best for: Teams that want source control, CI and security scanning on one bill): Free at $0/user/mo; Premium $29/user/mo billed annually; Ultimate quote-only
- CircleCI (best for: Teams whose cost driver is build volume rather than headcount): From $15/mo on compute credits; free plan includes 30,000 credits a month, no card required
- Buildkite (best for: Teams running their own compute at scale): Free for solo developers and small teams; Pro $30 per active user per month

Every comparison of CI/CD tools leads with a feature matrix, and every tool on that matrix can run tests, build containers and deploy. The features converged years ago. What did not converge is the billing unit, and that is where the money and the arguments are.

A pipeline that costs one team almost nothing costs another several thousand a month on the same plan, because one team's cost driver is minutes and the other's is seats.

## Top Picks

Based on features, real-world fit, and value for money.

Best CI/CD Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

GitHub Actions | Free for public repositories; Team $4/user/mo (first 12 months) with 3,000 minutes, then Linux runners from $0.002/min | Teams already on GitHub who want zero additional vendors |

GitLab CI | Free at $0/user/mo; Premium $29/user/mo billed annually; Ultimate quote-only | Teams that want source control, CI and security scanning on one bill |

CircleCI | From $15/mo on compute credits; free plan includes 30,000 credits a month, no card required | Teams whose cost driver is build volume rather than headcount |

Buildkite | Free for solo developers and small teams; Pro $30 per active user per month | Teams running their own compute at scale |

Jenkins | Free and open source; upkeep of a mature installation often exceeds a commercial subscription | Teams with hard self-hosting requirements and someone to own it |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 5 do not publish a comparable monthly price and are left out rather than estimated.

1

### GitHub Actions

Top Pick

Best for: Teams already on GitHub who want zero additional vendors

PricingFree for public repositories; Team $4/user/mo (first 12 months) with 3,000 minutes, then Linux runners from $0.002/min

Visit GitHub Actions →

2

### GitLab CI

Best for: Teams that want source control, CI and security scanning on one bill

PricingFree at $0/user/mo; Premium $29/user/mo billed annually; Ultimate quote-only

Visit GitLab CI →

3

### CircleCI

Best for: Teams whose cost driver is build volume rather than headcount

PricingFrom $15/mo on compute credits; free plan includes 30,000 credits a month, no card required

Visit CircleCI →

4

### Buildkite

Best for: Teams running their own compute at scale

PricingFree for solo developers and small teams; Pro $30 per active user per month

Visit Buildkite →

5

### Jenkins

Best for: Teams with hard self-hosting requirements and someone to own it

PricingFree and open source; upkeep of a mature installation often exceeds a commercial subscription

Visit Jenkins →

## What it is

CI/CD is the automation that runs on every commit: build, test, and in most teams deploy. Continuous integration is the part that verifies a change works with everything already merged. Continuous delivery is the part that gets it to an environment. In practice the two live in one configuration file and one bill.

## Why it matters

Two costs, and teams budget for the wrong one. The visible cost is the subscription. The invisible one, usually larger, is engineering time waiting on queued builds. A pipeline that takes twelve minutes instead of four, times a hundred commits a week, times a team of ten, is a headcount-sized number that never appears on an invoice.

Concurrency limits, which are priced very differently across these vendors, are what decide it.

## Key features to look for

Billing unit
Minutes, credits, active users, concurrent agents or seats. This decides your bill far more than the advertised monthly price, and no two vendors here use the same one.

Concurrency
How many jobs run at once. The single biggest lever on how long engineers wait, and the thing most often capped on cheaper tiers.

Self-hosted runners
Running jobs on your own machines. Cuts the compute bill substantially and moves the cost to infrastructure you already maintain.

Included allowance
What the free or entry tier covers before metering starts. Ranges from 400 compute minutes on GitLab Free to unlimited on self-hosted.

Monorepo support
Whether the tool can build only what changed. On a large repository this is the difference between a four-minute pipeline and a forty-minute one.

Where it runs
Managed cloud, self-hosted, or both. Regulated teams often have no choice, and the pricing model changes completely between them.

## Pricing

GitHub Actions, GitLab, CircleCI and Buildkite publish prices. GitLab Ultimate is quote-only, and Jenkins is free and open source, so the bill is servers and maintenance.

The cheapest published paid entry is GitHub Team at $4 per user per month for the first 12 months, with 3,000 included Actions minutes a month and Linux runners from $0.002 per minute after that allowance.

CircleCI paid usage starts at $15 a month after 30,000 free credits (about 3,000 minutes on a Linux Medium machine), GitLab Premium is $29 per user per month billed annually with 10,000 compute minutes, and Buildkite Pro is $30 per active user per month.

Costs jump at CircleCI bulk credits from $600 per million, at Buildkite on extra agents at $3.50 and workflow executions at $15 per million, and on GitLab Ultimate, a custom quote with 50,000 compute minutes.

Plan | Price | Best for |

GitHub Actions public repositories | Free | Public repositories are free |

GitHub Actions Team | $4 per user per month for the first 12 months | Private repositories draw on the Team allowance |

GitHub Actions compute | From $0.002 per minute | Linux 1-core rate; Linux 2-core is $0.006 per minute |

GitHub Actions included minutes | 3,000 minutes/month on Team | 2,000 on Free and 50,000 on Enterprise, for private repositories |

GitLab Free | $0 per user per month | 400 compute minutes a month, 5 users per top-level group |

GitLab Premium | $29/user/mo, billed annually | 10,000 compute minutes, plus $12 of AI credits per user monthly, a limited-time promotion |

GitLab Ultimate | Custom quote | 50,000 compute minutes, plus $24 of AI credits per user, a limited-time promotion |

CircleCI Free | 30,000 credits a month | Free tier, no card required |

CircleCI paid usage | $15 a month | Paid usage billed on compute credits |

CircleCI bulk credits | From $600 per million | Volume discounts above five million |

Buildkite Free | Free | For solo developers and small teams |

Buildkite Pro | $30 per active user per month | Pro plan, 10 agents included |

Buildkite additional agent | $3.50 per agent per month | Each additional agent past the 10 included |

Buildkite workflow executions | $15 per million | Workflow runs past the included allowance |

Jenkins open source | Free | You pay for servers and maintenance |

Mistakes to avoid

×Comparing the monthly prices directly. $15 on CircleCI, $29 on GitLab and $30 on Buildkite buy compute, a seat and an active user respectively. The same team gets a different ranking depending on whether it has more engineers or more builds.

×Ignoring concurrency on the entry tier. The subscription looks cheap and then every push queues behind the last one. Waiting engineers cost more than the plan you saved on.

×Treating Jenkins as free. It has no licence cost and a real maintenance cost, and the second one does not appear until the person who set it up leaves.

Expert tips

→Count your build minutes for one real week before reading any pricing page. It converts an unanswerable comparison into arithmetic, and it usually eliminates two of the five options immediately.

→Cache dependencies before buying faster compute. Most slow pipelines are slow because they reinstall everything on every run, and that fix is free.

→Use self-hosted runners for the heavy jobs and managed compute for the rest. Buildkite is built for exactly this split, and GitHub Actions supports it well.

## The bottom line

For most teams already on GitHub, GitHub Actions is the correct answer and the comparison stops there: the pipeline lives beside the code and the allowance covers normal usage. Move off it for a reason, and the two real reasons are cost shape and self-hosting.

If builds are heavy and the team is small, CircleCI's compute billing fits better than any per-seat plan. If you run your own machines at scale, Buildkite is cheaper than paying anyone for compute. GitLab wins when you want one vendor for source control, CI and security, and are willing to pay per seat for it.

## Frequently asked questions

What is the cheapest CI/CD tool?

It depends entirely on your cost driver. CircleCI's free tier includes 30,000 credits a month (about 3,000 minutes on a Linux Medium machine), GitLab and Buildkite both have free plans, and GitHub Actions is free on public repositories. For a small team running heavy pipelines, compute-billed CircleCI is usually cheapest. For a large team running light ones, per-seat pricing wins.

Is GitHub Actions enough for production deployments?

Yes for most teams. The common reasons to add or replace it are self-hosting requirements, monorepo builds that need finer control over what runs, and compute costs at high volume where self-hosted runners on another orchestrator are cheaper.

Does Jenkins still make sense in 2026?

When self-hosting is a hard requirement it remains the most capable option, and the plugin ecosystem covers cases the managed platforms do not. Budget the maintenance honestly: it is the cost, and it is paid in engineering time rather than on an invoice.

What is the difference between build minutes and credits?

Minutes are wall-clock time on a standard machine. Credits price the machine as well as the time, so a larger runner consumes credits faster than a small one for the same minute. Credit systems are more accurate and harder to forecast.

Related guides

Monitoring ToolsAi For DevopsTerraform Iac ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Code Editors in 2026

URL: https://devshot.co/reviews/best-code-editors
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: VS Code, Cursor, JetBrains IDEs, Sublime Text, Neovim, Vim, Zed and Nova compared on 2026 pricing, extensibility, AI editing and who each one actually fits.

Expert Guide

## The Best Code Editors in 2026

Eight editors from free open source builds to yearly per-seat subscriptions, compared on extensions, AI features and what actually changes your bill.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

VS Code is the default pick for most developers in 2026: free, open source, and extended by more plugins than any other editor here. Cursor earns its $20/mo Individual plan with agent-mode AI editing built into a VS Code fork. JetBrains's All Products Pack covers language-aware refactoring across every major stack for $299 the first year, dropping to $179 by year three. Sublime Text is still a one-time license for people who want native speed without a subscription clock.

Key facts8 tools compared: VS Code, Cursor, JetBrains IDEs, Sublime Text, Neovim, Vim, Zed, Nova

- Updated: September 25, 2026
- Top pick: VS Code (best for: Most developers who want one free editor that covers any language through extensions)
- Top pick price as of September 25, 2026: VS Code: Free and open source; no paid tier for the editor itself.

- 8 tools compared: VS Code, Cursor, JetBrains IDEs, Sublime Text, Neovim, Vim, Zed, Nova
- Cursor (best for: Developers who want VS Code's feel with an AI agent built into the editor): Free (Hobby); Individual $20/mo; Teams $40/user/mo, billed monthly.
- JetBrains IDEs (best for: Teams that want deep, language-native refactoring across a specific stack): All Products Pack $299/yr individual (drops to $179/yr by year three); WebStorm alone from $79/yr.
- Sublime Text (best for: Developers who want a fast native editor and a license, not a subscription): $99 one time (personal license, 3 years of updates); Business from $65/seat/yr.

Every editor here opens a file and shows you a cursor. The differences show up in how each one is extended, licensed and billed, not in the text-editing basics.

A free VS Code install and a paid JetBrains subscription can end up doing the same refactor, one with a marketplace extension and a bit of patience, the other with a feature built into the IDE's index of your project.

Devshot data: [Toolradar](https://toolradar.com/tools/vscode), the software directory we run, evaluated 124 IDE and code editor tools for its [September 2026 code editor ranking](https://toolradar.com/best/ide-code-editors), fetched today.

That list spans full IDEs and lightweight editors together; this page is the developer-editor cut of it, with the terminal-only and Mac-only entries a general "best IDE" ranking usually skips.

Prices below were verified on each vendor's own pricing page today, 25 September 2026.

Where JetBrains and Cursor also sell an AI coding agent, we cover the editor itself here and leave the agent comparison to our AI coding assistants guide.

Windsurf, once a common name on this kind of list, is not included: Cognition acquired it in 2025 and renamed it Devin Desktop in June 2026, and its old pricing page now redirects straight to Devin's.

## Top Picks

Based on features, real-world fit, and value for money.

Best Code Editors in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

VS Code | Free and open source; no paid tier for the editor itself. | Most developers who want one free editor that covers any language through extensions |

[Cursor](https://toolradar.com/tools/cursor) | Free (Hobby); Individual $20/mo; Teams $40/user/mo, billed monthly. | Developers who want VS Code's feel with an AI agent built into the editor |

JetBrains IDEs | All Products Pack $299/yr individual (drops to $179/yr by year three); WebStorm alone from $79/yr. | Teams that want deep, language-native refactoring across a specific stack |

Sublime Text | $99 one time (personal license, 3 years of updates); Business from $65/seat/yr. | Developers who want a fast native editor and a license, not a subscription |

Neovim | Free and open source, Apache 2.0 license. | Terminal-first developers who want built-in LSP support and a modern plugin system |

Vim | Free; distributed as charityware, with donations supporting causes in Uganda. | Developers who want the original modal editor everywhere, including minimal servers |

[Zed](https://toolradar.com/tools/zed) | Personal free; Pro $10/mo; Business $30/seat/mo. | Developers who want the fastest GUI editor here with a built-in AI panel |

Nova | $99 once (includes 1 year of updates); $49/yr to keep updating after that. | Mac developers who want a native editor built by a small, focused team |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 5 of 8 do not publish a comparable monthly price and are left out rather than estimated.

1

### VS Code

Top Pick

Best for: Most developers who want one free editor that covers any language through extensions

PricingFree and open source; no paid tier for the editor itself.

+Free with no seat limit, trial or account required to install and use it

+The largest extension marketplace of any editor in this list, covering nearly every language and framework

+Remote development over SSH, containers and WSL ships built in, not as a paid add-on

−Built on Electron, so it is slower to cold-start and heavier on RAM than Sublime Text, Zed or Vim

−No native AI agent; Copilot is a separate GitHub subscription layered on top

Visit VS Code →

2

### Cursor

Best for: Developers who want VS Code's feel with an AI agent built into the editor

PricingFree (Hobby); Individual $20/mo; Teams $40/user/mo, billed monthly.

+Familiar VS Code shortcuts, themes and most extensions carry over directly

+Agent mode plans a multi-file change instead of one autocomplete suggestion at a time

+Hobby is free with no card required, enough to try the workflow before paying

−You are on a maintained fork, so some VS Code extensions lag or need a workaround

−Heavy agent use in a large monorepo can slow down noticeably

Visit Cursor →

3

### JetBrains IDEs

Best for: Teams that want deep, language-native refactoring across a specific stack

PricingAll Products Pack $299/yr individual (drops to $179/yr by year three); WebStorm alone from $79/yr.

+Refactoring, debugging and code inspection are native to the IDE, not bolted on with an extension

+The continuity discount drops the All Products Pack to $179/yr by the third year on the plan

+A single-product license like WebStorm covers one stack for less than the full pack

−The individual pack's first-year price is the discounted rate; the organization tier with no continuity discount runs $979/yr per seat

−Heavier and slower to index a large project on first open than VS Code or Sublime Text

Visit JetBrains IDEs →

4

### Sublime Text

Best for: Developers who want a fast native editor and a license, not a subscription

Pricing$99 one time (personal license, 3 years of updates); Business from $65/seat/yr.

+Opens instantly and handles very large files that make Electron-based editors stall

+One license covers every OS you run it on, no per-machine repurchase

+Goto Anything and multi-cursor editing remain some of the fastest in any editor here

−After three years an upgrade is required to keep receiving updates, so the license is not truly one-time

−No built-in AI editing; anything beyond autocomplete needs a third-party plugin

Visit Sublime Text →

5

### Neovim

Best for: Terminal-first developers who want built-in LSP support and a modern plugin system

PricingFree and open source, Apache 2.0 license.

+Built-in LSP client gives autocomplete and go-to-definition without Vim's older plugin workarounds

+Lua configuration is faster and more maintainable than Vimscript for a large setup

+Free, and light enough to run comfortably over SSH on a small remote box

−The learning curve for modal editing is real; expect a slower first few weeks than a GUI editor

−No official graphical interface; you are configuring a terminal application, not clicking through settings

Visit Neovim →

6

### Vim

Best for: Developers who want the original modal editor everywhere, including minimal servers

PricingFree; distributed as charityware, with donations supporting causes in Uganda.

+Available by default or a single install command on nearly every Linux and macOS box

+Runs comfortably on minimal hardware, useful for editing a config file on a small remote server

+Decades of documentation, plugins and muscle memory shared with Neovim

−No native LSP support, so modern autocomplete and refactoring need extra plugins Neovim ships with by default

−Vimscript configuration is dated next to Neovim's Lua setup

Visit Vim →

7

### Zed

Best for: Developers who want the fastest GUI editor here with a built-in AI panel

PricingPersonal free; Pro $10/mo; Business $30/seat/mo.

+Rust and GPU rendering make it the fastest GUI editor tested here for large files and fast scrolling

+Personal tier is free forever, not a time-limited trial

+Built-in collaborative editing lets two developers work in the same file in real time

−Extension ecosystem is newer and thinner than VS Code's or JetBrains's

−AI features beyond the free tier require the $10/mo Pro plan and its token allotment

Visit Zed →

8

### Nova

Best for: Mac developers who want a native editor built by a small, focused team

Pricing$99 once (includes 1 year of updates); $49/yr to keep updating after that.

+Fully native to macOS, with the platform polish that implies in scrolling, fonts and window behavior

+Built-in Git client, task runner and live preview for web projects without extra extensions

+The renewal after the first year is cheaper than most subscription editors here

−macOS only; there is no Windows or Linux build to standardize a mixed team on

−Smaller extension library than any cross-platform editor in this list

Visit Nova →

## What it is

A code editor is the application a developer writes and reads source code in: syntax highlighting, a file tree, search across a project, and a way to run or debug what you wrote.

A full IDE like a [JetBrains](https://toolradar.com/tools/jetbrains) product adds a built-in understanding of the language, so it can rename a class across forty files or flag a broken import before you compile.

A lighter editor like VS Code, [Sublime Text](https://toolradar.com/tools/sublime-text) or Zed starts fast and adds that depth through extensions, which you install per project.

Terminal-based editors, [Vim](https://toolradar.com/tools/vim) and Neovim, skip the graphical window entirely and run inside your shell, which matters if you work over SSH on a remote server more than on your own laptop; pair either with a visual Git client if the command-line workflow for commits and diffs feels slower than the editing itself.

## Why it matters

The editor you pick sets your daily startup time, your license bill for the next three years, and how much of your workflow is one paid subscription versus a stack of free tools you assembled yourself.

VS Code and Cursor share a codebase, but Cursor's fork adds a paid agent that plans and edits across files, a real cost difference for the same editing feel.

Licensing shape matters as much as the sticker price.

JetBrains and Sublime Text both sell a subscription that discounts the longer you stay, while Nova sells a one-time purchase with a year of updates included and a cheaper annual renewal after that. Vim and Neovim cost nothing either way, which is why they still anchor so many remote and CI workflows.

## Key features to look for

Extension and plugin ecosystemEssential
VS Code's marketplace and JetBrains's plugin repository dwarf smaller editors. Sublime Text, Zed and Nova have real but thinner ecosystems, and Vim/Neovim depend on the plugin manager you choose yourself.

Built-in language intelligenceEssential
JetBrains IDEs index your project and understand the language natively. VS Code, Sublime Text and Zed rely on Language Server Protocol extensions for the same depth, which usually works but adds a setup step per language.

AI editing and completionEssential
Cursor and JetBrains's AI add-on plan changes across multiple files from a prompt. VS Code ships <a href="/reviews/github-copilot-review">GitHub Copilot</a> as a separate extension and subscription, and Zed's Pro tier bundles a token credit for its own AI panel.

Startup speed and resource use
Native editors, Sublime Text, Zed, Vim and Neovim, open large files and cold-start noticeably faster than Electron-based VS Code or a JetBrains IDE, which matters on an older laptop or a 100,000-line file.

Terminal versus GUI
Vim and Neovim run over SSH with no graphical dependency, useful on a remote server or inside a container. Every other editor here needs a desktop session, even the ones with a terminal panel built in.

Licensing shape
Free and open source (VS Code, Neovim, Vim), one-time purchase with a cheaper renewal (Sublime Text, Nova), a subscription that gets cheaper the longer you stay (JetBrains), or a flat monthly AI plan (Cursor, Zed). This decides your three-year cost more than the feature list does.

## Pricing

Three editors cost nothing outright: VS Code, [Neovim](https://toolradar.com/tools/neovim) and Vim are free and open source, with Vim funded as charityware rather than sold.

Cursor and [Zed](https://toolradar.com/tools/zed) both give away a real free tier and charge for AI features on a monthly per-user plan that scales up again for teams.

JetBrains prices its All Products Pack as a subscription that gets cheaper the longer an individual stays on it, with an organization license carrying no continuity discount and a higher per-seat rate instead.

Sublime Text and Nova both sell a license instead of a subscription, each renewing on cheaper terms after the first purchase. None of these editors bundles an AI coding agent for free beyond a limited trial; the deeper agent comparison, including Claude Code, Copilot and Codex, lives in our AI coding assistants guide.

Plan | Price | Best for |

VS Code | Free | Open source editor; AI (Copilot) is a separate GitHub subscription |

Cursor Hobby | Free | Limited agent requests, no card required |

Cursor Individual | $20/mo | Extended agent limits, billed monthly |

Cursor Teams | $40/user/mo | Centralized billing and admin controls |

JetBrains All Products Pack, individual (year one) | $299/year | 16 tools including every major language IDE and a perpetual fallback license |

JetBrains All Products Pack, individual (year three onward) | $179/year | Same pack, continuity discount after two consecutive years |

JetBrains All Products Pack, organization | $979/year | Per-seat rate with no individual continuity discount |

Sublime Text Personal License | $99 once | Covers every OS on one key; 3 years of updates included |

Sublime Text Business (first 10 seats) | $65/seat/year | Tiered rate drops further at 11+ and 26+ seats |

Neovim | Free | Apache 2.0 license, open source |

Vim | Free | Charityware; donations support causes in Uganda |

Zed Personal | Free | Free forever, not a trial |

Zed Pro | $10/month | Unlimited predictions plus a token credit for AI features |

Zed Business | $30/seat/month | Order forms available at 25 or more seats |

Nova full purchase | $99 once | macOS only; includes one year of updates |

Nova annual renewal | $49/year | Optional after year one; keeps updates coming |

Mistakes to avoid

×Buying JetBrains's All Products Pack for a team that only ever touches one language. A single product license, WebStorm alone from $79/yr, is a smaller bill than the full pack when one IDE covers the whole stack.

×Assuming Cursor's Individual plan has unlimited agent usage. It has extended limits, not infinite ones, and a heavy migration week can push a solo developer toward Teams-level usage.

×Standardizing a mixed Windows and macOS team on Nova because one developer loves it. It is macOS only, and there is no Windows or Linux build to fall back on.

Expert tips

→Start with VS Code or Neovim before paying for anything. Both are free, both cover almost any language through extensions or LSP, and the editing habits carry over if you later move to Cursor or a JetBrains IDE.

→If AI-assisted editing is the reason you are switching, try Cursor's free Hobby tier or Zed's free Personal plan before committing to a paid subscription.

→Renew Nova or Sublime Text on their cheaper terms rather than repurchasing from scratch, and time a fresh purchase to when you actually need the next update cycle.

## The bottom line

VS Code is the right default for most developers: free, extensible, and the base several paid editors here fork from. Pay for Cursor when agent-mode AI editing across files earns back its monthly fee, and pay for a JetBrains IDE when native refactoring across a large codebase is worth the subscription.

Sublime Text and Nova suit developers who want a license instead of a subscription clock, and Vim and Neovim remain the free, install-anywhere choice for terminal-first and remote work.

Zed is the pick when raw startup and scroll speed matter more than ecosystem size.

Methodology: we shortlisted these eight from Toolradar's 124-tool category ranking by developer relevance, then re-read every vendor's own pricing page today; no placement on this page is paid. Cite this: Devshot, "Best Code Editors in 2026", September 2026.

## Frequently asked questions

What is the best code editor for developers in 2026?

VS Code for most developers: it is free, open source, and its extension marketplace covers nearly any language. Cursor is the pick if AI agent editing is worth $20/mo to you, and a JetBrains IDE is the pick for teams that want native, language-aware refactoring and will pay $299 the first year for it.

Is VS Code really free, with no catch?

Yes. VS Code itself has no paid tier, seat limit or trial period. The only cost most people add is GitHub Copilot, a separate subscription layered on as an extension, which is not required to use the editor.

Cursor vs VS Code: what is the actual difference?

Cursor is a maintained fork of VS Code, so shortcuts, themes and most extensions carry over. The difference is Cursor's built-in agent mode, which plans and edits across multiple files from a prompt, available free on a limited Hobby plan or unlocked further on the paid Individual plan.

Is there a free alternative to JetBrains IDEs?

VS Code with a language extension, or Neovim with a Language Server Protocol plugin, cover most of what a JetBrains IDE does for free, though neither indexes a project as deeply out of the box. JetBrains's own community editions of some IDEs, like IntelliJ IDEA Community, are also free but skip the advanced refactoring in the paid editions.

Do I have to pay for Vim or Neovim?

No. Both are free. Vim is distributed as charityware, with the project asking for optional donations toward causes in Uganda rather than selling a license. Neovim is fully open source under the Apache 2.0 license with no paid tier at all.

Which code editor is fastest to start up?

Sublime Text, Zed, Vim and Neovim all open close to instantly, since none of them runs on Electron. VS Code and JetBrains IDEs are noticeably slower to cold-start, especially JetBrains on first index of a large project.

What is the cheapest way to get AI editing features?

Zed's free Personal tier includes a limited allotment of AI features before the $10/mo Pro plan. Cursor's free Hobby tier gives limited agent requests before its paid Individual plan. Neither requires a card to start.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [Zed pricing](https://zed.dev/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsGit ClientsVibe Coding ToolsDatabase ClientsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Database Clients in 2026

URL: https://devshot.co/reviews/best-database-clients
Type: review
Published: 2026-09-01
Updated: 2026-09-25
Summary: The database GUI clients worth using in 2026, compared on one-time licenses versus subscriptions, and on what the free editions still refuse to do.

Expert Guide

## The Best Database Clients in 2026

One of these is $99 once. Another is $259 every year. Two are free. They all open a SQL window.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

TablePlus is $99 once for one device, which is the cheapest way to own a serious client. Beekeeper Studio Community is free; Indie is $9 per month billed yearly. DBeaver Community is free; Lite is $113 a year. DataGrip is $109 the first year for an individual, $259 per user per year for an organisation. pgAdmin is free and only talks to Postgres.

Key facts5 tools compared: TablePlus, Beekeeper Studio, DBeaver, DataGrip, pgAdmin

- Updated: September 25, 2026
- Top pick: TablePlus (best for: People who want to pay once and keep the app)
- Top pick price as of September 25, 2026: TablePlus: From $99 once (Basic, 1 device); limited free trial

- 5 tools compared: TablePlus, Beekeeper Studio, DBeaver, DataGrip, pgAdmin
- Beekeeper Studio (best for: People who want a modern UI and a free tier that is not a demo): From $9/user/mo (Indie, billed yearly); Community free, 14-day trial
- DBeaver (best for: Anyone who needs to talk to many engines without paying, and teams that will pay $113 for the Lite extras): From $113/user/year or $12/mo (Lite); Community free, 14-day PRO trial
- DataGrip (best for: People who already live in JetBrains, or who want the deepest SQL intelligence and will pay every year): From $109 first year (Individual); free for non-commercial use

Database clients are the same product in a screenshot and four different products on an invoice. TablePlus sells a perpetual license. Beekeeper and DBeaver sell a year, or nothing. DataGrip sells a JetBrains subscription that gets cheaper if you pay it yourself and stays expensive if the company does.

You are not choosing a grid and a query editor.

You are choosing whether this tool can disappear next year without taking the license with it.

## Top Picks

Based on features, real-world fit, and value for money.

Best Database Clients in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

TablePlus | From $99 once (Basic, 1 device); limited free trial | People who want to pay once and keep the app |

Beekeeper Studio | From $9/user/mo (Indie, billed yearly); Community free, 14-day trial | People who want a modern UI and a free tier that is not a demo |

DBeaver | From $113/user/year or $12/mo (Lite); Community free, 14-day PRO trial | Anyone who needs to talk to many engines without paying, and teams that will pay $113 for the Lite extras |

DataGrip | From $109 first year (Individual); free for non-commercial use | People who already live in JetBrains, or who want the deepest SQL intelligence and will pay every year |

pgAdmin | Free, open source, Postgres only | Postgres-only work, especially when you do not want another license |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 5 do not publish a comparable monthly price and are left out rather than estimated.

1

### TablePlus

Top Pick

Best for: People who want to pay once and keep the app

PricingFrom $99 once (Basic, 1 device); limited free trial

+Buy once: after the first year you keep the build you have

+Native on macOS, Windows and Linux, and fast enough that people stop using the browser console

+$99 is less than a year of any subscription on this list

−Updates after year one are a paid renewal, so 'perpetual' is the current version, not the future ones

−Team starts at three seats, which is awkward for a two-person shop

Visit TablePlus →

2

### Beekeeper Studio

Best for: People who want a modern UI and a free tier that is not a demo

PricingFrom $9/user/mo (Indie, billed yearly); Community free, 14-day trial

+Community is usable day to day: unlimited connections, no tab cap

+Pay for 12 months and you keep those versions even if you cancel

+Indie at $9 yearly is the cheapest subscription here

−Indie is one device; a home and work laptop means Professional

−The Community cut list is real: some engines and paid features are missing

Visit Beekeeper Studio →

3

### DBeaver

Best for: Anyone who needs to talk to many engines without paying, and teams that will pay $113 for the Lite extras

PricingFrom $113/user/year or $12/mo (Lite); Community free, 14-day PRO trial

+Community talks to an enormous list of engines at $0

+It is what a lot of teams already have installed, which is a kind of support

+Lite at $113 a year is cheaper than DataGrip organisation

−The UI is powerful and dated, and new users get lost in it

−Lite access ends when the year ends, unlike TablePlus

Visit DBeaver →

4

### DataGrip

Best for: People who already live in JetBrains, or who want the deepest SQL intelligence and will pay every year

PricingFrom $109 first year (Individual); free for non-commercial use

+Query insight, refactoring and schema awareness are still ahead of the lightweight clients

+Non-commercial use is free, which covers learning and side projects

+If you already own the All Products Pack, DataGrip is already paid for

−Organization licenses do not get the continuity discount, so year three is still $259

−It is a JetBrains IDE: slower to open, heavier in RAM

Visit DataGrip →

5

### pgAdmin

Best for: Postgres-only work, especially when you do not want another license

PricingFree, open source, Postgres only

+Free, maintained with Postgres, no seat to explain

+Covers the Postgres-specific objects the generic clients treat as an afterthought

+Runs as a desktop app or a web server you host

−Postgres only. A second engine means a second tool

−The interface is a tree of every object, which is complete and slow to scan

Visit pgAdmin →

## What it is

A database client is a desktop (sometimes web) app that connects to Postgres, MySQL, SQL Server, SQLite and the rest, then lets you browse schemas, run SQL, edit rows and export results.

The serious ones add an SSH tunnel, a query history, an ER diagram and enough keyboard shortcuts that the mouse becomes optional.

Pricing has split into buy once, subscribe per year, and free community editions that withhold a few things you will notice in the second week. The engines they talk to are the same. The license is not.

## Why it matters

A $99 one-time license and a $259 yearly seat look close in month one and are not close in year three. Five developers on DataGrip organisation licenses are $1,295 a year, every year. Five TablePlus Team seats at $79 are $395 once.

That is the comparison finance will actually understand.

The other reason is where the connection strings live. A client that stores production credentials in a personal cloud workspace is a different risk from one that keeps them in a local file you already encrypt.

## Key features to look for

License shape
Perpetual, annual subscription, or free community. This decides the three-year cost more than any feature row.

Engines supported
Postgres only, or Postgres plus MySQL, SQL Server, Redis, SQLite and the warehouse of the month. A second engine is how a cheap Postgres-only tool becomes a second purchase.

SSH tunnels and cloud IAM
How you reach a database that is not on localhost. SSO and IAM auth show up on the business tiers, not the indie ones.

What the free edition cuts
Tab limits, missing engines, no export, no ER diagram. Read the cut list. Community editions are real products, not trials, until you hit the cut.

Updates after you pay
TablePlus includes a year of updates on a perpetual license. Subscriptions include them until you stop paying. That difference matters on a tool you open daily.

## Pricing

TablePlus, Beekeeper Studio, DBeaver and DataGrip publish prices, checked 25 September 2026, and none of them is quote-only. Beekeeper Community, DBeaver Community and pgAdmin are free, and DataGrip is free for non-commercial use.

The cheapest paid way to own the app is TablePlus Basic at $99 once for one device. The cheapest subscription is Beekeeper Indie at $9 per user per month billed yearly ($18 monthly).

The bill jumps on DataGrip organisation seats at $259 per user per year, on DBeaver Enterprise and Ultimate at $255 and $510 a year, and on Beekeeper when you pay monthly instead of yearly.

Plan | Price | Best for |

TablePlus Basic | $99 per license | 1 device; one year of updates; USD, VAT may apply |

TablePlus Standard | $129 per license (was $198) | 2 devices on one license |

TablePlus Team | $79/seat, 3-seat minimum | Teams, three seats minimum |

TablePlus update renewal | $59/device | Updates after the included year |

TablePlus extra device | $79 | Add a device to a license |

TablePlus trial | Free trial | Limited to 2 tabs, 2 windows, 2 advanced filters |

Beekeeper Studio Community | Free | Unlimited connections, common databases, no tab limits |

Beekeeper Studio Indie | $9/user/mo yearly ($18 monthly) | 1 device; lifetime access to versions from that year |

Beekeeper Studio Professional | $14/user/mo yearly ($28 monthly) | Multiple devices |

Beekeeper Studio Business | $18/user/mo yearly ($35 monthly) | SSO and air-gapped support |

Beekeeper Studio trial | 14-day trial, no card | Paid features with a 30-day refund |

DBeaver Community | Free | Open source Community Edition |

DBeaver Lite | $113/user/year or $12/month | Access stops if you do not renew |

DBeaver Enterprise | $255/user/year | Annual license, the next edition above Lite |

DBeaver Ultimate | $510/user/year | Top desktop edition, annual license |

DBeaver PRO trial | 14-day trial | Trial of the PRO features |

DataGrip Individual (year 1) | $109 first year | Individual license, first year |

DataGrip Individual (year 2) | $87 second year | Individual price in the second year |

DataGrip Individual (year 3+) | $65 third year onwards | Individual price from the third year |

DataGrip Organization | $259/user/year | Organisation seat, no continuity discount |

DataGrip non-commercial | Free | Non-commercial use only |

DataGrip All Products Pack | Separate bundle | If you already pay for other JetBrains IDEs |

pgAdmin | Free | Open source, Postgres only |

Mistakes to avoid

×Buying DataGrip organisation seats for a team that needed a grid and a tunnel. $259 a year is the right price for people who refactor SQL. It is the wrong price for someone who runs SELECT a few times a week.

×Standardising on pgAdmin and then adding MySQL. You now have two clients and no license savings. Pick a multi-engine tool first if the stack is not Postgres-only.

×Treating TablePlus as free after year one of updates. You keep the app. You do not keep new versions unless you renew.

Expert tips

→If two engines are in play, skip pgAdmin as the standard even if Postgres is the main one. The second client is how these bills double.

→Pay for TablePlus or Beekeeper Indie yourself if the company will not. $99 once or $9 a month is below most expense thresholds. DataGrip organisation is not.

→Keep production credentials out of any vendor cloud workspace. Local connections plus an SSH tunnel you already understand beat a convenient synced vault you have not read the policy for.

## The bottom line

Pay once if you can. TablePlus at $99 for one device is the cleanest deal on this list, and it stays yours. Use Beekeeper Community or DBeaver Community if you do not want to pay yet; both are real clients.

Upgrade Beekeeper to Indie at $9 a month yearly, or DBeaver Lite at $113 a year, only when you hit a specific cut.

DataGrip is the right call when SQL is a large part of the job or when the All Products Pack already covers it.

It is the wrong default for a whole company. pgAdmin is free and correct for Postgres-only. It is a trap the day a second engine appears.

## Frequently asked questions

What is the best free database client?

DBeaver Community if you talk to more than one engine. Beekeeper Community if you want a cleaner UI and can live with the Community cut list. pgAdmin if the only database is Postgres. DataGrip is free for non-commercial use, not for company work.

Is TablePlus a one-time purchase?

Yes. Basic is $99 per license for one device, Standard is $129 for two, Team is $79 per seat with a three-seat minimum, checked 26 August 2026. You get a year of updates. After that the app keeps working. New versions are a $59 renewal per device.

How much does DataGrip cost?

On jetbrains.com/store, checked 26 August 2026: $109 the first year for an individual ($87 then $65), and $259 per user per year for an organisation. Organisation seats do not drop in year two. Non-commercial use is free.

TablePlus vs DBeaver vs DataGrip: which should I get?

TablePlus if you want to pay once and work fast on a few engines. DBeaver Community if the price must be zero and the engine list is long. DataGrip if you want IDE-grade SQL and already tolerate JetBrains, or if the company is buying organisation seats for people who write SQL all day. Do not buy all three.

Related guides

Postgres HostingSelf Hosting ToolsCi Cd ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Diagram as Code Tools in 2026

URL: https://devshot.co/reviews/best-diagram-as-code-tools
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Ten tools for turning text or Python into diagrams, compared on license, what the paid tiers add, and which ones your CI pipeline can render without a browser…

Expert Guide

## The Best Diagram as Code Tools in 2026

Ten diagram-as-code tools compared on what a paid tier actually adds, since most of the category ships free, with prices checked on vendor pricing pages in September 2026.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 10 tools compared

TL;DR

The best diagram-as-code tool for most developers in 2026 is Mermaid. It's free, MIT-licensed, and rendered natively by GitHub, GitLab and Notion, with the hosted Mermaid Chart Plus tier starting at $10/user/mo billed annually. PlantUML, Graphviz and D2 are also fully free and open source, with no paid tier at all.

Pick Diagrams (mingrammer) to generate cloud architecture diagrams straight from Python, and Excalidraw or draw.io when a teammate needs to hand-edit the result. Eraser ($15/member/mo Starter) and Ilograph ($18/mo Pro) are the paid SaaS options, built for AI-assisted or multi-perspective architecture diagrams. Structurizr's hosted cloud service is being shut down with no replacement, so its only paid path now is a self-hosted server priced in GBP.

Key facts10 tools compared: Mermaid, draw.io (diagrams.net), Excalidraw, PlantUML, Graphviz, D2…

- Updated: September 25, 2026
- Top pick: Mermaid (best for: Teams that want a diagram GitHub, GitLab and Notion already render with no extension)
- Top pick price as of September 25, 2026: Mermaid: Free, MIT license (Mermaid.js); Mermaid Chart Plus $10/user/mo, Premium $20/user/mo, both billed annually.

- 10 tools compared: Mermaid, draw.io (diagrams.net), Excalidraw, PlantUML, Graphviz, D2, Diagrams (mingrammer), Eraser, Structurizr, Ilograph
- draw.io (diagrams.net) (best for: Teams that want a real diagram editor with zero license cost and zero vendor lock-in): Free forever, Apache 2.0 license; no paid tier, no Enterprise SSO upsell.
- Excalidraw (best for: Teams that want a hand-drawn look for architecture sketches, not a polished formal diagram): Free, MIT license; Excalidraw+ Plus $6/mo per user, billed monthly (annual billing saves 14%).
- PlantUML (best for: Teams that need dense UML: class, sequence, state and component diagrams from a compact text syntax): Free, LGPL-3.0 license; no paid tier or hosted product.

A diagram as code tool turns a text file, or in one case a Python script, into a rendered diagram. The diagram then lives in version control next to the system it describes, updating in the same pull request.

That is the whole pitch against a drag-and-drop canvas: no drift between the doc and the architecture, and a diff you can actually review.

Toolradar data: the [directory homepage](https://toolradar.com/) lists 10,797+ tools across 401+ categories, checked 25 September 2026.

Its closest tracked category, [best diagramming tools](https://toolradar.com/best/diagramming), evaluated 53 diagramming tools to build that ranking, checked September 2026.

That list leans toward whiteboard tools like Miro and Lucidchart rather than code-first ones, so open [Mermaid](https://toolradar.com/tools/mermaid), [draw.io](https://toolradar.com/tools/drawio), [Excalidraw](https://toolradar.com/tools/excalidraw) and [Eraser](https://toolradar.com/tools/eraser-io) there for the wider catalog listing.

How we ranked: ten tools, each read on its own site or GitHub repository on 25 September 2026.

The order follows license and cost, whether a CI pipeline can render the output without a browser, and what a paid tier buys once the free path runs out.

A whiteboard-first tool like Miro or Lucidchart is a different buy than a text file your CI renders on every push.

That comparison lives in the [diagram tools guide](https://toolradar.com/guides/best-diagram-tools) on Toolradar.

## Top Picks

Based on features, real-world fit, and value for money.

Best Diagram as Code Tools in 2026: 10 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Mermaid | Free, MIT license (Mermaid.js); Mermaid Chart Plus $10/user/mo, Premium $20/user/mo, both billed annually. | Teams that want a diagram GitHub, GitLab and Notion already render with no extension |

draw.io (diagrams.net) | Free forever, Apache 2.0 license; no paid tier, no Enterprise SSO upsell. | Teams that want a real diagram editor with zero license cost and zero vendor lock-in |

Excalidraw | Free, MIT license; Excalidraw+ Plus $6/mo per user, billed monthly (annual billing saves 14%). | Teams that want a hand-drawn look for architecture sketches, not a polished formal diagram |

PlantUML | Free, LGPL-3.0 license; no paid tier or hosted product. | Teams that need dense UML: class, sequence, state and component diagrams from a compact text syntax |

Graphviz | Free, Eclipse Public License 2.0; no paid tier or hosted product. | Teams that need the underlying graph-layout engine other diagram tools build on top of |

D2 | Free, MPL-2.0 license; Terrastruct lists no paid tier for D2 today. | Teams that want a modern syntax with built-in themes, sketch mode and animated diagrams |

Diagrams (mingrammer) | Free, MIT license; a Python library, no paid tier or hosted product. | Teams that want cloud architecture diagrams generated from a Python script, not drawn by hand |

Eraser | Free (3 files, 3 AI diagrams); Starter $15/member/mo annual ($20 mo); Business $45/member/mo annual ($60 mo). | Teams that want AI-generated diagrams alongside Markdown docs in one collaborative workspace |

Structurizr | Cloud service end of life, no replacement; self-hosted server from £300/mo, 1-20 users, billed annually; no USD price. | Teams already committed to the C4 model who can self-host, not new teams starting today |

Ilograph | Free ($0/mo, 14-day diagram history); Pro from $18/mo (annual) or $22/mo (monthly); Team from $25/editor/mo (annual). | Teams that need one architecture model to generate several interactive, zoomable perspectives |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 10 do not publish a comparable monthly price and are left out rather than estimated.

1

### Mermaid

Top Pick

Best for: Teams that want a diagram GitHub, GitLab and Notion already render with no extension

PricingFree, MIT license (Mermaid.js); Mermaid Chart Plus $10/user/mo, Premium $20/user/mo, both billed annually.

+MIT-licensed core library with 90,000+ GitHub stars, free forever with no rendering limit

+Native rendering in GitHub markdown, GitLab and Notion, so most diagrams need no separate viewer

+Mermaid Chart's Plus tier removes the free plan's 3-diagram cap and adds AI generation credits

−The free Mermaid Chart plan caps at 3 diagrams and limited diagram size, a real wall for anyone using the hosted editor daily

−Complex sequence or class diagrams get unreadable fast; PlantUML handles denser UML better

Visit Mermaid →

2

### draw.io (diagrams.net)

Best for: Teams that want a real diagram editor with zero license cost and zero vendor lock-in

PricingFree forever, Apache 2.0 license; no paid tier, no Enterprise SSO upsell.

+Apache 2.0 licensed, with the vendor stating outright there is no Enterprise tier gating SSO

+Diagrams saved in 2005 still load today, so files never go stale against an old app version

+Confluence, Jira, Google Workspace, Microsoft 365 and VS Code integrations ship free, not gated

−Not diagram-as-code in the strict sense: the source is an XML file edited on a canvas, not a text language you'd write by hand

−No built-in CI rendering step; scripting a headless export takes extra tooling

Visit draw.io (diagrams.net) →

3

### Excalidraw

Best for: Teams that want a hand-drawn look for architecture sketches, not a polished formal diagram

PricingFree, MIT license; Excalidraw+ Plus $6/mo per user, billed monthly (annual billing saves 14%).

+MIT-licensed with 132,000+ GitHub stars, and the free tier already includes unlimited collaborators

+Mermaid-to-Excalidraw import turns a text diagram into an editable sketch, bridging the two workflows

+Plus is the cheapest paid tier of any tool on this list, well under Mermaid Chart's or Eraser's entry price

−The free tier caps at 1 infinite scene; Plus is required for unlimited scenes and folders

−Hand-drawn style reads as informal for a customer-facing architecture doc

Visit Excalidraw →

4

### PlantUML

Best for: Teams that need dense UML: class, sequence, state and component diagrams from a compact text syntax

PricingFree, LGPL-3.0 license; no paid tier or hosted product.

+Free and open source (LGPL-3.0), with 13,000+ GitHub stars and two decades of UML syntax coverage

+Renders sequence, class, state, component, and deployment diagrams, more UML diagram types than Mermaid covers

+Runs locally via a JAR file or Docker image, so CI rendering needs no network call to a hosted server

−Syntax is denser and less forgiving than Mermaid's for a newcomer's first flowchart

−No native rendering in GitHub or GitLab; needs a plugin, proxy server, or pre-rendered image

Visit PlantUML →

5

### Graphviz

Best for: Teams that need the underlying graph-layout engine other diagram tools build on top of

PricingFree, Eclipse Public License 2.0; no paid tier or hosted product.

+Free under the Eclipse Public License 2.0, with no commercial edition to weigh against it

+The DOT language it introduced is the layout engine several tools on this list (PlantUML, Diagrams) render through

+Handles large, densely connected graphs that hierarchical tools like Mermaid struggle to lay out cleanly

−DOT syntax is lower-level than Mermaid or D2; positioning and styling take more trial and error

−No native theming system; visual polish requires manual attribute tuning

Visit Graphviz →

6

### D2

Best for: Teams that want a modern syntax with built-in themes, sketch mode and animated diagrams

PricingFree, MPL-2.0 license; Terrastruct lists no paid tier for D2 today.

+Free under the Mozilla Public License 2.0, with 25,000+ GitHub stars on the d2lang/d2 repository

+Ships professionally designed themes and a sketch mode out of the box, ahead of Mermaid or PlantUML's default styling

+Exports to SVG, PNG, PDF, PPTX, GIF and even ASCII, and offers three interchangeable layout engines

−Terrastruct's own pricing page now redirects straight to the D2 project site, so there is no separate paid product to compare

−Smaller ecosystem and fewer native integrations than Mermaid, which GitHub and GitLab render without a plugin

Visit D2 →

7

### Diagrams (mingrammer)

Best for: Teams that want cloud architecture diagrams generated from a Python script, not drawn by hand

PricingFree, MIT license; a Python library, no paid tier or hosted product.

+MIT licensed with built-in icon sets for AWS, Azure, GCP, Kubernetes and on-prem systems, no manual icon hunting

+Diagrams live as Python functions, so the same code review and version control process covers infrastructure and its diagram

+42,000+ GitHub stars and active maintenance make it the most established Python option for this specific job

−Only outputs static PNG images; no interactive viewer or hosted sharing link

−Requires Graphviz installed as a system dependency, an extra setup step compared to a single binary

Visit Diagrams (mingrammer) →

8

### Eraser

Best for: Teams that want AI-generated diagrams alongside Markdown docs in one collaborative workspace

PricingFree (3 files, 3 AI diagrams); Starter $15/member/mo annual ($20 mo); Business $45/member/mo annual ($60 mo).

+Diagram-as-code syntax plus AI generation in one workspace, with GitHub, Notion, Confluence and VS Code integrations even on Free

+Starter opens up unlimited files and 40 AI diagrams a month once the free cap gets tight

+Business adds SAML SSO and unlimited version history at $45/member/mo annually, cheaper than most competing team wikis

−The free plan caps at 3 files total, not 3 diagrams a month, so it stops being usable fast for a real project

−AI diagram count is a hard quota (40 on Starter, 250 on Business) rather than unlimited until Enterprise

Visit Eraser →

9

### Structurizr

Best for: Teams already committed to the C4 model who can self-host, not new teams starting today

PricingCloud service end of life, no replacement; self-hosted server from £300/mo, 1-20 users, billed annually; no USD price.

+Defined the C4 model (context, container, component, code) that several other tools on this list, including Ilograph, support exporting to

+One DSL source generates multiple views (context, container, deployment) instead of separate files per diagram

+Structurizr Lite and the new 'local' command line workflow remain free for a single developer working offline

−The Structurizr cloud service is end of life with no replacement, confirmed on the vendor's own documentation in September 2026

−The only current paid product, Structurizr server, is priced in GBP (from £300/mo for 1-20 users, billed annually) with no USD list price published

Visit Structurizr →

10

### Ilograph

Best for: Teams that need one architecture model to generate several interactive, zoomable perspectives

PricingFree ($0/mo, 14-day diagram history); Pro from $18/mo (annual) or $22/mo (monthly); Team from $25/editor/mo (annual).

+One model generates multiple interactive, clickable perspectives, closer to a live system map than a static image

+The free plan has no diagram-count cap, only a 14-day history window, generous for solo use

+Ilograph Desktop ($11.99 to $14.99/mo depending on billing) offers an offline, on-premises option for teams behind a firewall

−Team pricing runs per editor per month ($25 to $30), which adds up faster than Eraser's or Mermaid Chart's per-user rate at small team sizes

−SSO is locked behind the custom-quote Team+ tier, not available at any published price

Visit Ilograph →

## What it is

Diagram as code means the diagram's source is plain text (or, for Diagrams, Python) that a renderer turns into an SVG, PNG or PDF. The source is diffable, reviewable in a pull request, and reproducible on any machine that has the renderer installed, unlike a file saved inside a proprietary canvas app.

The category splits into two camps.

General-purpose text-to-diagram languages (Mermaid, PlantUML, D2, Graphviz) draw flowcharts, sequence diagrams and graphs from a compact syntax.

Architecture-specific tools (Structurizr, Ilograph, Diagrams) model a system's structure, usually against the C4 model or a cloud provider's icon set, and generate several views from one source.

Excalidraw and draw.io sit at the edge of the category: both are canvas editors first, with a code-like layer (Mermaid-to-Excalidraw import, an XML/JSON file format) bolted on rather than built in.

## Why it matters

Every tool on this list except three is free with no paid tier at all. PlantUML (LGPL-3.0), Graphviz (Eclipse Public License 2.0), D2 (MPL-2.0) and Diagrams (MIT) publish no commercial product at all.

The real decision is which syntax your team will actually keep writing, not which invoice is smaller.

Mermaid's core library is MIT-licensed and free; its maker sells the hosted Mermaid Chart on top once the free plan's 3-diagram cap gets in the way.

Devshot data: 28,000 developer readers get this brief, and the audience file updated 20 September 2026 puts the open rate at 30%.

For that audience, the tool that renders in CI without a browser usually beats one that only works inside a hosted editor. Mermaid, PlantUML, Graphviz, D2 and Diagrams all do this, via a CLI. Structurizr is the cautionary case here.

Its cloud service is end of life with no replacement, confirmed on the vendor's own docs in September 2026, so a team that built workflows around it now needs to migrate to the self-hosted server product or a local export.

Latency and lock-in are the other axis.

A CLI tool you run in a GitHub Action never depends on a vendor's uptime. A hosted SaaS like Eraser or Ilograph trades that independence for AI generation, team libraries and a shareable link.

The CI/CD tools guide and the Git clients guide cover the pipelines these diagrams usually render inside.

## Key features to look for

License and costEssential
Seven of the ten tools here are fully free and open source with no paid tier. The three that sell a hosted product (Mermaid Chart, Eraser, Ilograph) charge for AI generation, team collaboration or version history, not the base rendering.

Renders in CI without a browserEssential
Mermaid, PlantUML, Graphviz, D2 and Diagrams all ship a CLI or library that renders a diagram in a GitHub Action or a build step. Excalidraw, draw.io, Eraser and Ilograph are built around an interactive editor first.

What the diagram modelsEssential
General-purpose languages (Mermaid, PlantUML, D2, Graphviz) draw flowcharts and sequence diagrams. Structurizr and Ilograph model whole software architectures against the C4 model. Diagrams generates cloud infrastructure diagrams from Python, with built-in AWS, Azure and GCP icon sets.

Native rendering support
GitHub, GitLab and Notion all render Mermaid blocks natively, with no extension required. PlantUML, D2 and Graphviz need a renderer, an extension, or a build step to appear inline.

AI diagram generation
Eraser's paid tiers add AI-generated diagrams (3 free, 40 on Starter, 250 on Business) from a text prompt. Mermaid Chart's Plus and Premium tiers scale AI credits the same way. The free, code-only tools have no built-in AI generation step.

## Pricing

Seven of these ten tools are fully free with no paid tier at all. PlantUML (LGPL-3.0), Graphviz (Eclipse Public License 2.0), D2 (MPL-2.0), Diagrams (MIT), draw.io (Apache 2.0), and the core Mermaid and Excalidraw libraries (both MIT) all ship under a permissive license with no commercial edition to buy.

The decision for those seven is syntax and rendering fit, not price.

Mermaid Chart, the hosted product built on top of free Mermaid, is the cheaper of the two per-seat SaaS plans built on an open-source base.

A higher Premium tier exists for teams that need more AI credits. Excalidraw+ undercuts it further, the cheapest paid tier on the page, with a 14% discount for annual billing.

Eraser and Ilograph are the two dedicated paid SaaS products.

Both scale from a mid-teens monthly seat price on their entry paid tier toward a team plan in the twenties to forties per seat; the pricing table below has the exact figures and billing terms. Structurizr is the outlier.

Its cloud service is end of life with no replacement, and the surviving Structurizr server product publishes prices only in GBP, from £300/mo for 1-20 users billed annually, with no USD figure on its pricing page as of September 2026.

Plan | Price | Best for |

Mermaid (core library) | Free | MIT license, no diagram or usage limit |

Mermaid Chart Basic | Free | 3 diagrams, limited size, limited AI credits |

Mermaid Chart Plus | $10/user/mo billed annually | Unlimited diagrams, 300 AI credits a year |

Mermaid Chart Premium | $20/user/mo billed annually | 2,000 AI credits a year, unlimited viewer seats |

Mermaid Chart Enterprise | Custom quote | SSO, custom contract, dedicated CSM |

draw.io (diagrams.net) | Free | Apache 2.0, no paid tier of any kind |

Excalidraw (core app) | Free | MIT license, 1 infinite scene, unlimited collaborators |

Excalidraw+ Plus | $6/mo per user | Billed monthly; annual billing saves 14% |

PlantUML | Free | LGPL-3.0, no paid tier or hosted product |

Graphviz | Free | Eclipse Public License 2.0, no paid tier |

D2 | Free | MPL-2.0; Terrastruct's pricing page redirects to the D2 project site |

Diagrams (mingrammer) | Free | MIT license, Python library, no paid tier |

Eraser Free | $0/member/mo | 3 files, 3 AI diagrams, 7-day version history |

Eraser Starter | $15/member/mo annual ($20 monthly) | Unlimited files, 40 AI diagrams, 90-day history |

Eraser Business | $45/member/mo annual ($60 monthly) | 250 AI diagrams, SAML SSO, unlimited history |

Eraser Enterprise | Custom quote | Unlimited AI diagrams, flexible deployment |

Structurizr server, 1-20 users | £300/mo billed annually | No USD list price published; self-hosted, cloud service is end of life |

Structurizr server, 1,001+ users | £2,400/mo billed annually | Top published server tier; contact for 251+ team pricing above 1,000 |

Ilograph Free | $0/mo | Unlimited diagrams, 14-day history |

Ilograph Pro | $18/mo annual ($22/mo monthly) | Custom icons, unlimited history, 20 private viewers |

Ilograph Team | $25/editor/mo annual ($30/editor/mo monthly) | Team workspace, unlimited viewers |

Ilograph Team+ | Custom quote | SSO via SAML 2.0, API access |

Ilograph Desktop | $11.99/mo annual ($14.99/mo monthly) | Offline desktop app, works with local files |

Mistakes to avoid

×Adopting Structurizr's cloud service for a new project in 2026. It is end of life with no replacement, per the vendor's own documentation, so a new team should start on Structurizr's local workflow or the self-hosted server instead.

×Assuming a diagram-as-code tool renders in CI just because it renders in a browser. Eraser and Ilograph are built around a hosted editor; Mermaid, PlantUML, Graphviz, D2 and Diagrams all ship a CLI or library that renders headless.

×Comparing Mermaid Chart's or Eraser's monthly SaaS price against PlantUML's or Graphviz's $0 as if they compete on the same axis. Seven tools on this list are free forever; the paid three sell AI generation and team collaboration on top of a category that is mostly free by default.

Expert tips

→Start with Mermaid for anything that needs to render inline in a GitHub README or a Notion doc, since neither PlantUML nor D2 renders natively there without an extension.

→If the team already writes infrastructure in Python, generate the architecture diagram with Diagrams (mingrammer) directly from that code instead of hand-drawing it in a separate tool that will drift out of sync.

→Before paying for Eraser's or Ilograph's AI generation, check whether ChatGPT or Claude can already output valid Mermaid or D2 syntax for the same prompt. Both text-to-diagram languages are common enough in LLM training data that a free chat tool often gets close on the first try.

## The bottom line

Default to Mermaid for most day-to-day diagrams: it is free, MIT-licensed, and the only tool here GitHub, GitLab and Notion render natively.

Reach for PlantUML when a diagram needs deep UML support, and Graphviz or D2 when the layout itself, not the notation, is the hard part.

Use Diagrams when the source of truth is already Python, and Excalidraw or draw.io when a non-technical teammate needs to hand-edit the result.

Pay for Eraser or Ilograph only once a team genuinely needs AI generation or multi-perspective modeling; check Structurizr's end-of-life notice before building anything new on its cloud service. Cite this: Devshot, "Best Diagram as Code Tools in 2026", September 2026.

## Frequently asked questions

What is the best diagram-as-code tool in 2026?

Mermaid is the default for most developers: free, MIT-licensed, and rendered natively by GitHub, GitLab and Notion with no plugin, checked on mermaid.js.org and mermaid.ai in September 2026. PlantUML is the strongest pick for dense UML diagrams, and Diagrams (mingrammer) is the best fit when the diagram should be generated from existing Python infrastructure code.

Is Mermaid free?

Yes. The core Mermaid.js library is MIT-licensed and free with no usage limit, checked on GitHub in September 2026. The hosted Mermaid Chart product has a free Basic plan capped at 3 diagrams, with Plus starting at $10/user/mo billed annually and a pricier Premium tier above it for larger teams and more AI credits.

Is there a free diagram-as-code tool?

Most of the category is free: Mermaid, PlantUML, Graphviz, D2, Diagrams, Excalidraw and draw.io all publish a permissive open-source license (MIT, LGPL-3.0, Eclipse Public License 2.0, or MPL-2.0) with no usage cap on the base tool. Only Mermaid Chart, Eraser and Ilograph sell a separate paid tier on top of a free plan.

Mermaid vs PlantUML: which should I use?

Mermaid for anything that needs to render inline in GitHub, GitLab or Notion without a plugin, and for simpler flowcharts, sequence and Gantt diagrams. PlantUML for denser UML work, class and component diagrams especially, where its two-decade head start on syntax coverage shows. Both are free and open source, so the choice is about syntax and where the diagram needs to display, not cost.

What happened to Structurizr's cloud service?

Structurizr's hosted cloud service, along with Structurizr Lite, the Structurizr CLI, and the old on-premises installer, are end of life with no replacement for the cloud service specifically, confirmed on the vendor's own documentation in September 2026. The surviving paid product is Structurizr server, a self-hosted option priced in GBP, from £300/mo for 1-20 users billed annually, with no USD price published.

Can I generate a diagram from Python or another programming language?

Diagrams (mingrammer) generates AWS, Azure, GCP and on-prem architecture diagrams directly from a Python script, and is free under the MIT license. D2 and Graphviz both offer language APIs and CLI tools that other programs can call. Mermaid and PlantUML are primarily text languages rather than libraries, though wrapper packages exist in most languages to generate their syntax programmatically.

What is the cheapest paid diagram tool here?

Excalidraw+ Plus, at $6/mo per user (billed monthly, with a 14% discount for annual billing), is the cheapest paid tier on this list, checked on plus.excalidraw.com in September 2026. Ilograph's Desktop plan is close behind at $11.99 to $14.99/mo depending on billing term. Both sit well under Eraser's or Mermaid Chart's entry-level seat price.

Related guides

Git ClientsCi Cd ToolsTerraform Iac ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Error Tracking Tools in 2026

URL: https://devshot.co/reviews/best-error-tracking-tools
Type: review
Published: 2026-08-14
Updated: 2026-09-25
Summary: The error tracking platforms engineers actually deploy in 2026: Sentry, Rollbar, Bugsnag, Datadog and Honeybadger compared on pricing…

Expert Guide

## The Best Error Tracking Tools in 2026

Knowing production broke before the customer emails, with enough context to fix it without a reproduction.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

The best error tracking tools in 2026 are Sentry as the default for almost every team, Rollbar for aggressive automatic grouping and deploy tracking, Bugsnag for mobile-heavy products with its stability score model, Datadog Error Tracking if you already run Datadog for everything else, and Honeybadger if you want errors, uptime checks and a status page on one flat bill. Sentry's free tier covers 5,000 errors a month, which is enough to start properly.

Key facts5 tools compared: Sentry, Rollbar, Bugsnag, Datadog Error Tracking, Honeybadger

- Updated: September 25, 2026
- Top pick: Sentry (best for: Almost every team, as the default choice)
- Top pick price as of September 25, 2026: Sentry: From $26/mo (Team, billed annually); free Developer plan ($0)

- 5 tools compared: Sentry, Rollbar, Bugsnag, Datadog Error Tracking, Honeybadger
- Rollbar (best for: Teams that want automatic grouping and deploy correlation): Free plan (5,000 occurrences a month); paid plans priced by event volume
- Bugsnag (best for: Mobile-first products): Free plan (7,500 events a month); paid plans, part of SmartBear
- Datadog Error Tracking (best for: Teams already standardised on Datadog): $25/mo flat under 50,000 errors (annual); included with Datadog APM and RUM

The gap between an exception happening in production and an engineer knowing about it used to be a customer support ticket.

Error tracking closed that gap, and the modern versions do considerably more: they group thousands of occurrences into one issue, attach the stack trace with your source mapped back to real filenames, record what the user was doing, and tell you which deploy introduced it.

The differences between the tools are less about capability now and more about noise, pricing model and how much context arrives with each alert.

## Top Picks

Based on features, real-world fit, and value for money.

Best Error Tracking Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Sentry](https://toolradar.com/tools/sentry) | From $26/mo (Team, billed annually); free Developer plan ($0) | Almost every team, as the default choice |

[Rollbar](https://toolradar.com/tools/rollbar) | Free plan (5,000 occurrences a month); paid plans priced by event volume | Teams that want automatic grouping and deploy correlation |

[Bugsnag](https://toolradar.com/tools/bugsnag) | Free plan (7,500 events a month); paid plans, part of SmartBear | Mobile-first products |

[Datadog Error Tracking](https://toolradar.com/tools/datadog) | $25/mo flat under 50,000 errors (annual); included with Datadog APM and RUM | Teams already standardised on Datadog |

Honeybadger | From $26/mo (Team, 50K errors); free Developer plan with 5,000 errors a month | Small teams that want errors, uptime checks and status pages in one bill |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Sentry

Top Pick

Best for: Almost every team, as the default choice

PricingFrom $26/mo (Team, billed annually); free Developer plan ($0)

+Grouping and stack trace quality are the best here, which is what you are actually buying

+Free tier covers 5,000 errors a month with real functionality, not a crippled demo

+Performance monitoring, session replay and profiling in the same product

−Quota-based pricing means an incident can consume your monthly error budget

−The product has broadened a lot, so the interface is busier than it was

Visit Sentry →

2

### Rollbar

Best for: Teams that want automatic grouping and deploy correlation

PricingFree plan (5,000 occurrences a month); paid plans priced by event volume

+Grouping algorithm is aggressive and accurate, producing fewer duplicate issues

+Deploy tracking clearly attributes new errors to the release that caused them

+Automation rules can assign, mute and escalate without human triage

−Narrower scope than Sentry, no performance monitoring or replay

−Smaller ecosystem and community

Visit Rollbar →

3

### Bugsnag

Best for: Mobile-first products

PricingFree plan (7,500 events a month); paid plans, part of SmartBear

+Stability score reframes errors as a release-quality metric leadership can act on

+Best mobile SDK support here, particularly for iOS and Android crash reporting

+Strong release health tracking for staged rollouts

−Web experience is good but less compelling than Sentry's

−Pricing is less transparent since the SmartBear acquisition

Visit Bugsnag →

4

### Datadog Error Tracking

Best for: Teams already standardised on Datadog

Pricing$25/mo flat under 50,000 errors (annual); included with Datadog APM and RUM

+Correlation across errors, traces, logs and infrastructure is genuinely powerful for root cause

+One vendor and one bill if you already run Datadog

+Enterprise-grade access controls and retention

−Standalone use still means adopting the Datadog agent and console, which only pays off inside Datadog

−Error-specific ergonomics are behind the specialists

Visit Datadog Error Tracking →

5

### Honeybadger

Best for: Small teams that want errors, uptime checks and status pages in one bill

PricingFrom $26/mo (Team, 50K errors); free Developer plan with 5,000 errors a month

+Team plan includes unlimited users, 5 uptime monitors and a status page, so there is no second tool to buy

+Free Developer plan matches Sentry's 5,000 errors a month for a solo developer

+Short, readable pricing with the base price and included volume on one line

−No session replay, so reproducing front-end bugs still takes manual work

−Free plan keeps error data for 15 days, against 90 days on Team

Visit Honeybadger →

## What it is

An error tracking tool captures unhandled exceptions and logged errors from your application, deduplicates them into issues, and enriches each one with context: stack trace, source maps, release version, environment, user and request data, breadcrumbs of what happened before, and often a session replay.

It then routes alerts to the humans who can fix them, ideally only for things that matter.

## Why it matters

Two reasons, and the second is the one teams underrate. The first is detection: you find out about failures in minutes rather than when someone complains.

The second is triage: a good tool tells you that this exception affects 340 users on one browser version since Tuesday's deploy, which turns an unbounded investigation into a bounded one. Without that, engineers spend the expensive part of the day trying to reproduce something instead of fixing it.

## Key features to look for

Grouping quality
How well the tool collapses thousands of occurrences into one issue without merging genuinely different bugs. This is the single biggest quality difference between platforms and it is hard to evaluate from a marketing page.

Source maps and release tracking
Minified stack traces mapped back to your real source, tied to the release that introduced them. Without release association, the question of which deploy broke it stays unanswered.

Context and breadcrumbs
What the user did in the minutes before the error, which request, which feature flags. This is what removes the need for a reproduction.

Noise control
Inbound filters, rate limits, spike protection and the ability to ignore known-benign errors. An error tracker nobody looks at is worse than none, and noise is why that happens.

Pricing model
Per-error volume, per-user, or bundled into an observability platform. Volume-based plans need spike protection or an incident can generate the bill as well as the outage.

## Pricing

Sentry, Honeybadger, Rollbar and Datadog publish dollar prices, Bugsnag sends you to a calculator on its site, and Sentry Enterprise is a custom quote.

The cheapest credible entry is Sentry's free Developer plan, with Team billed annually as the first paid plan; Honeybadger Team matches that entry price with 50,000 errors, unlimited users and uptime monitors included. Costs jump at Sentry Business, $80 a month billed annually, which adds unlimited dashboards and anomaly detection.

Rollbar's free plan covers 5,000 occurrences a month and paid plans scale by event volume (Rollbar's own example is $29 a month for 50,000 events on Essentials), while Bugsnag's free plan covers 7,500 events a month. Datadog Error Tracking is included with APM and RUM, and standalone it is a flat $25 a month under 50,000 errors on annual billing.

Plan | Price | Best for |

Sentry Developer | $0 | 1 user, 5k errors/mo, 5M spans, 50 replays |

Sentry Team | $26/mo billed annually | Unlimited users, 50k errors |

Sentry Business | $80/mo billed annually | Unlimited dashboards and anomaly detection |

Sentry Enterprise | Custom quote | Enterprise plan at a custom price |

Rollbar Essentials | $29/mo at 50K events | Priced by event volume; free plan covers 5,000 occurrences |

Bugsnag paid plans | Prices on their site | Free plan covers 7,500 events a month, part of SmartBear |

Datadog Error Tracking | $25/mo under 50K errors | Annual rate, $36 on demand; included free with APM and RUM |

Honeybadger Developer | $0 | 1 user, 5,000 errors a month, 15-day retention |

Honeybadger Team | From $26/mo | Unlimited users, 50K errors, uptime monitors, status page |

Mistakes to avoid

×Alerting on every new error. Within two weeks nobody reads the channel, and the one alert that mattered scrolls past. Alert on rate, on new-since-deploy, and on user impact instead.

×Skipping source maps. A minified stack trace is nearly useless, and uploading source maps in CI takes an afternoon once.

×Not tagging releases. Without release association you cannot answer the most valuable question the tool can answer: which deploy caused this?

×Ignoring quota mechanics. On volume-priced plans, an incident that generates a million errors also generates a bill. Configure spike protection and rate limits before you need them.

Expert tips

→Set up source maps and release tagging in CI on day one, before you tune anything else. Everything else in the tool depends on those two.

→Triage the top five issues by user impact each week rather than by count. A crash affecting 200 users beats a noisy log line firing 50,000 times.

→Attach user context (an id, not personal data) so you can answer whether one customer is disproportionately affected.

→Aggressively ignore known-benign errors, such as browser extension noise and cancelled requests. Signal-to-noise is what determines whether anyone opens the tool in month three.

## The bottom line

For nearly every team, start with Sentry: the free tier is genuinely usable at 5,000 errors a month, grouping quality is the best in the category, and the paid Team plan covers a real production application.

Pick Bugsnag if you ship mobile and want stability scores per release, Rollbar if automatic grouping and triage automation are your priority, and Datadog only if you already live there.

Whatever you pick, source maps and release tagging matter more than the choice itself.

## Frequently asked questions

What is the best free error tracking tool?

Sentry's Developer plan is the strongest free option, covering 5,000 errors a month with full grouping, stack traces and 50 session replays for one user. Honeybadger's Developer plan matches that 5,000-error allowance for one user, and self-hosted Sentry is the route if you want unlimited volume and are willing to run it yourself.

How much does Sentry cost?

The Developer plan is free with 5,000 errors a month for a single user. Team is $26 a month billed annually with unlimited users and 50,000 errors, Business is $80 a month billed annually adding unlimited dashboards and anomaly detection, and Enterprise is custom. Additional volume is available as a paid add-on.

What is the difference between error tracking and APM?

Error tracking captures exceptions and groups them into fixable issues with stack traces and user context. APM measures performance: latency, throughput and where time is spent across services. They answer different questions, which is why platforms increasingly bundle both, but a slow endpoint and a crashing endpoint are separate problems.

Is Sentry or Rollbar better?

Sentry is the broader and more widely adopted platform, with better SDK coverage, an open-source core and performance monitoring plus session replay in the same product. Rollbar is more narrowly focused on error tracking with excellent automatic grouping and triage automation. If you want one tool for errors, performance and replay, choose Sentry.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Sentry pricing](https://sentry.io/pricing)checked Sep 2026
- [Rollbar pricing](https://rollbar.com/pricing)checked Sep 2026
- [Bugsnag pricing](https://bugsnag.com)checked Sep 2026
- [Datadog Error Tracking pricing](https://datadoghq.com/pricing)checked Sep 2026

Related guides

Api Testing ToolsSelf Hosting ToolsDatadog ReviewSentry ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Feature Flag Platforms in 2026

URL: https://devshot.co/reviews/best-feature-flag-platforms
Type: review
Published: 2026-08-28
Updated: 2026-09-25
Summary: Feature flag platforms compared on billing unit. LaunchDarkly bills service connections and MAU, Unleash bills $75 per seat, Flagsmith bills API calls…

Expert Guide

## The Best Feature Flag Platforms in 2026

Four vendors, four incompatible billing units. One advertises no per-seat pricing while another charges $75 a seat, with a five-seat minimum if you self-host.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 tools compared

TL;DR

No two vendors here sell the same unit, which makes the headline prices meaningless on their own. LaunchDarkly is free forever with unlimited seats and bills $10 per service connection plus $8.33 per 1,000 client-side monthly active users. Unleash bills $75 per seat a month, with a five-seat minimum ($375) only when you self-host the Enterprise edition, and its open-source edition is free. Flagsmith starts at $40 and meters API calls. PostHog bills flag requests, from $0.0001 each after the first million free every month.

Work out whether your cost driver is people, traffic or services before reading any of those numbers.

Key facts4 tools compared: LaunchDarkly, Unleash, Flagsmith, PostHog

- Updated: September 25, 2026
- Top pick: LaunchDarkly (best for: Larger engineering organisations where seat count would otherwise dominate the bill)
- Top pick price as of September 25, 2026: LaunchDarkly: From $10/mo per service connection (billed yearly); free forever, unlimited seats

- 4 tools compared: LaunchDarkly, Unleash, Flagsmith, PostHog
- Unleash (best for: Small teams wanting an open source core with a supported cloud option): From $75/seat/month (Pay-As-You-Go); five-seat minimum if self-hosted; open source free
- Flagsmith (best for: Teams that want a low entry price and predictable request-based scaling): From $40/mo (Start-Up, billed yearly; $45 monthly); free plan with 50,000 requests
- PostHog (best for: Teams that want flags, analytics and experiments on one bill): Flags from $0.0001/request after 1 million free each month; analytics from $0.00005/event

Feature flags are a small enough idea that the products converged quickly: a flag, a targeting rule, an SDK, an audit log. What did not converge is what any of these vendors thinks it is selling.

One charges for the engineers who log in, one explicitly refuses to, one charges for the connections your services open, and one folds flags into a usage-based analytics bill. The same team can be cheapest on any of them depending on shape alone.

## Top Picks

Based on features, real-world fit, and value for money.

Best Feature Flag Platforms in 2026: 4 tools compared, updated Sep 2026

Tool | Pricing | Best for |

LaunchDarkly | From $10/mo per service connection (billed yearly); free forever, unlimited seats | Larger engineering organisations where seat count would otherwise dominate the bill |

Unleash | From $75/seat/month (Pay-As-You-Go); five-seat minimum if self-hosted; open source free | Small teams wanting an open source core with a supported cloud option |

Flagsmith | From $40/mo (Start-Up, billed yearly; $45 monthly); free plan with 50,000 requests | Teams that want a low entry price and predictable request-based scaling |

[PostHog](https://toolradar.com/tools/posthog) | Flags from $0.0001/request after 1 million free each month; analytics from $0.00005/event | Teams that want flags, analytics and experiments on one bill |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 4 does not publish a comparable monthly price and is left out rather than estimated.

1

### LaunchDarkly

Top Pick

Best for: Larger engineering organisations where seat count would otherwise dominate the bill

PricingFrom $10/mo per service connection (billed yearly); free forever, unlimited seats

Visit LaunchDarkly →

2

### Unleash

Best for: Small teams wanting an open source core with a supported cloud option

PricingFrom $75/seat/month (Pay-As-You-Go); five-seat minimum if self-hosted; open source free

Visit Unleash →

3

### Flagsmith

Best for: Teams that want a low entry price and predictable request-based scaling

PricingFrom $40/mo (Start-Up, billed yearly; $45 monthly); free plan with 50,000 requests

Visit Flagsmith →

4

### PostHog

Best for: Teams that want flags, analytics and experiments on one bill

PricingFlags from $0.0001/request after 1 million free each month; analytics from $0.00005/event

Visit PostHog →

## What it is

A feature flag platform lets you ship code with a feature turned off and enable it later for specific users, without a deploy. It underpins trunk-based development, gradual rollouts and instant rollback, and it is the difference between shipping on Friday and not.

## Why it matters

The value is not the flag, which anyone can implement with a config file in an afternoon. It is everything around it: targeting rules non-engineers can change safely, an audit trail of who turned what on, and stale flag cleanup.

Teams that build their own get the flag and none of the rest, and the rest is what stops a flag system becoming a second, undocumented configuration language.

## Key features to look for

Billing unit
Seats, service connections, client-side monthly active users, API calls or events. All five appear in this category, and this decides your bill entirely.

Client-side vs server-side
LaunchDarkly prices these differently, at $8.33 per 1,000 client-side MAU against $10 per server-side service connection. Where your flags are evaluated is therefore a pricing decision as much as an architectural one.

Seat minimums
Self-hosted Unleash Enterprise requires five seats at $75 each, a $375 monthly floor regardless of usage, while the hosted plan bills only the seats you use. LaunchDarkly advertises unlimited seats at no cost. Opposite ends of the same category.

Self-hosting
Flagsmith and Unleash both offer it, which converts a subscription into operational work. Worth it when data residency is a requirement, expensive when it is a preference.

Flag lifecycle tooling
Detecting and removing stale flags. The feature that separates a platform from a config file after eighteen months, and the one nobody evaluates on.

Experimentation
Whether the platform also runs A/B tests on the same flags. PostHog's advantage, since the events it bills for are the ones that measure the experiment.

## Pricing

All four vendors publish prices, and none is quote-only. They bill different units (seats, service connections, client-side monthly active users, API calls or events), so the headline numbers are not the same product.

The cheapest paid entry is Flagsmith Start-Up at $40 a month, while LaunchDarkly is free forever with unlimited seats and bills $10 per service connection per month, billed yearly. Costs jump at Unleash's $75 seats (a $375 floor if you self-host), at $8.33 per 1,000 client-side MAU, and on additional Flagsmith API calls from $50 per million.

Unleash includes 53 million API requests before $5 per additional million, and PostHog bills flags at $0.0001 per request after the first million free.

Plan | Price | Best for |

LaunchDarkly free | Free forever | Unlimited seats, no per-seat charge |

LaunchDarkly service connection | $10/service connection/mo, yearly | Pay as you go per service connection |

LaunchDarkly client-side MAU | $8.33 per 1,000 client-side MAU | Per 1,000 client-side monthly active users |

LaunchDarkly server-side | Fractional by connection time | Server-side processes billed by connection time |

Unleash per seat | $75 per seat per month | Pay-As-You-Go, 14-day trial, no card required |

Unleash self-hosted floor | $375 a month | Five-seat minimum when you self-host Enterprise |

Unleash included API | 53 million requests per month | API requests included each month |

Unleash API overage | $5 per additional million | After the included monthly API requests |

Flagsmith free | Free | Free plan with a request allowance |

Flagsmith Start-Up | $40/mo billed yearly, $45 monthly | 1 million requests, 3 team members, unlimited projects |

Flagsmith Scale-Up | $250/mo billed yearly, $300 monthly | 5 million+ requests, 5 to 20 team members, SAML SSO |

Flagsmith extra API calls | From $50 per million per month | Additional API calls; rate falls with volume |

PostHog product analytics | $0.00005 per event | Usage-based; first million events free monthly |

PostHog feature flags | $0.0001 per request | First million flag requests free monthly, rate falls with volume |

Mistakes to avoid

×Comparing $40, $75 and free directly. Flagsmith's $40 buys request capacity, Unleash's $75 buys one seat (five minimum when self-hosted), and LaunchDarkly's free tier gives unlimited seats and charges for connections. These are four different products commercially.

×Missing the seat minimum. Self-hosting Unleash Enterprise carries a five-seat floor, $375 a month even for a team of two, while the hosted plan bills only the seats you use.

×Ignoring client-side traffic. LaunchDarkly's $8.33 per 1,000 client-side MAU is trivial for a B2B tool with 5,000 users and substantial for a consumer app with two million.

Expert tips

→Classify your cost driver before shortlisting. Many engineers and few users favours LaunchDarkly. Few engineers and heavy traffic favours Unleash. Predictable moderate traffic favours Flagsmith.

→Count client-side monthly active users separately from server-side services. Two vendors price them differently and the split changes the answer.

→Ask about stale flag tooling in the trial. Every platform makes creating flags easy; the cost eighteen months later is the ones nobody removed, and only some of these help you find them.

## The bottom line

For a large engineering team, LaunchDarkly's refusal to charge per seat is decisive and the free tier is genuinely usable, provided you model client-side MAU honestly. For a small team serving heavy traffic, Unleash's 53 million included requests are excellent value at $75 a seat on the hosted plan.

Flagsmith at $40 is the most sensible paid entry point for a team that wants neither extreme. And if you already run PostHog, flags sit on the same account with a million free requests a month, which is a better argument than any feature comparison in this category.

## Frequently asked questions

How much do feature flag platforms cost?

LaunchDarkly is free with unlimited seats, then $10 per service connection and $8.33 per 1,000 client-side monthly active users. Unleash is $75 per seat a month, with a five-seat minimum only when self-hosted. Flagsmith starts at $40 a month. PostHog bills flag requests, from $0.0001 each after the first million free.

Is LaunchDarkly really free?

The free tier is free forever with unlimited seats, which is unusual in this category. Costs begin with service connections and client-side monthly active users, so a team with many engineers and modest user traffic can stay very cheap on it.

Should I build feature flags myself?

The flag is easy and the surrounding tooling is not: safe targeting rules for non-engineers, audit trails, and stale flag detection. Teams that build their own generally get the flag and rediscover why the rest exists about a year later.

Why is Unleash more expensive per seat?

It prices the people rather than the traffic, and includes 53 million API requests. That is genuinely cheaper for a small team serving heavy load, and much more expensive for a large team serving little.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [PostHog pricing](https://posthog.com/pricing)checked Sep 2026

Related guides

Ci Cd ToolsMonitoring ToolsError Tracking ToolsLaunchDarkly Alternatives 2026: Feature Flag Prices ComparedAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Git Clients in 2026

URL: https://devshot.co/reviews/best-git-clients
Type: review
Published: 2026-07-09
Updated: 2026-09-25
Summary: The Git clients developers actually use in 2026, from polished GUIs to fast terminal UIs, ranked on speed, features, and price.

Expert Guide

## The Best Git Clients in 2026

Visual commit graphs, painless merge conflicts, and interactive rebase without memorizing flags. The GUI and terminal clients worth installing, ranked.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 6 tools compared

TL;DR

The best Git clients in 2026 are GitKraken for the most polished cross-platform GUI, Fork for a fast native client with a one-time price, Lazygit if you live in the terminal, Tower for power users, and GitHub Desktop if you want free and simple. Your OS and whether you want to pay once, subscribe, or stay free narrow the list fast.

Key facts6 tools compared: GitKraken, Fork, Lazygit, Tower, GitHub Desktop, Sourcetree

- Updated: September 25, 2026
- Top pick: GitKraken (best for: A polished, cross-platform GUI with every feature)
- Top pick price as of September 25, 2026: GitKraken: Free (public repos); Pro $10/user/mo, billed annually

- 6 tools compared: GitKraken, Fork, Lazygit, Tower, GitHub Desktop, Sourcetree
- Fork (best for: A fast native client you pay for once): $59.99 one-time; free evaluation
- Lazygit (best for: Developers who live in the terminal): Free (open source)
- Tower (best for: Power users on Mac and Windows): From $69/year

A Git client does not make you better at Git, but it does make the annoying parts far less painful: staging exactly the right lines, untangling a merge conflict, rewriting a messy history without remembering flags.

The field splits three ways, paid GUIs with every feature, free GUIs that cover the basics, and terminal UIs that keep your hands on the keyboard. We ranked on speed, how cleanly each one handles conflicts and rebases, platform coverage, and cost. Here are six worth installing.

## Top Picks

Based on features, real-world fit, and value for money.

Best Git Clients in 2026: 6 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[GitKraken](https://toolradar.com/tools/gitkraken) | Free (public repos); Pro $10/user/mo, billed annually | A polished, cross-platform GUI with every feature |

[Fork](https://toolradar.com/tools/fork-git) | $59.99 one-time; free evaluation | A fast native client you pay for once |

[Lazygit](https://toolradar.com/tools/lazygit) | Free (open source) | Developers who live in the terminal |

[Tower](https://toolradar.com/tools/tower) | From $69/year | Power users on Mac and Windows |

[GitHub Desktop](https://toolradar.com/tools/github-desktop) | Free (open source) | Free, simple, GitHub-centric work |

[Sourcetree](https://toolradar.com/tools/sourcetree) | Free | A feature-rich free GUI for Atlassian users |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 6 do not publish a comparable monthly price and are left out rather than estimated.

1

### GitKraken

Top Pick

Best for: A polished, cross-platform GUI with every feature

PricingFree (public repos); Pro $10/user/mo, billed annually

+Best-in-class visual commit graph

+Runs on macOS, Windows, and Linux

+Deep GitHub, GitLab, and Bitbucket integration

−Private repos require a paid plan

−Heavier and slower than native clients

Visit GitKraken →

2

### Fork

Best for: A fast native client you pay for once

Pricing$59.99 one-time; free evaluation

+Snappy native app on Mac and Windows

+One-time purchase, no subscription

+Excellent interactive rebase and diff viewer

−No Linux version

−Smaller team than the big commercial clients

Visit Fork →

3

### Lazygit

Best for: Developers who live in the terminal

PricingFree (open source)

+Free and open source

+Fast, keyboard-only workflow

+Runs anywhere your terminal does, including over SSH

−Steeper learning curve than a mouse GUI

−No point-and-click for newcomers

Visit Lazygit →

4

### Tower

Best for: Power users on Mac and Windows

PricingFrom $69/year

+Polished, mature interface

+Undo for almost any action

+Strong at complex rebases and cherry-picks

−Subscription only, no one-time license

−No Linux support

Visit Tower →

5

### GitHub Desktop

Best for: Free, simple, GitHub-centric work

PricingFree (open source)

+Completely free and open source

+Beginner-friendly with minimal clutter

+Tight GitHub integration

−Missing advanced features like interactive rebase

−Weaker for non-GitHub hosts

Visit GitHub Desktop →

6

### Sourcetree

Best for: A feature-rich free GUI for Atlassian users

PricingFree

+Free with a deep feature set

+Good Bitbucket and Jira integration

+Handles Git Flow out of the box

−Sluggish on large repositories

−macOS and Windows only, no Linux

Visit Sourcetree →

## What it is

A Git client wraps Git's command-line interface in a visual or keyboard-driven layer. Instead of typing an interactive rebase command and editing a text file, you drag commits, click to stage individual lines, and see your branch structure as a graph.

Some are full desktop apps like GitKraken, Tower, and Fork, some are lightweight terminal UIs like Lazygit, and all of them sit on top of the same Git you already have installed.

## Why it matters

Git's real power lives in operations most people are scared to run: interactive rebase, cherry-pick, reflog recovery, partial staging. A good client makes those safe by showing you exactly what will happen before you commit to it.

It also speeds up the daily loop of review, stage, commit, push, which adds up over thousands of repetitions. The catch is that a GUI can become a crutch. The developers who get the most from these tools understand the Git underneath and use the client to move faster, not to avoid learning it.

## Key features to look for

Visual commit graphEssential
A clear, color-coded view of branches, merges, and history so you can read the shape of a repository at a glance instead of parsing log output.

Line and hunk stagingEssential
Stage individual lines or chunks of a file, not just whole files, so each commit stays focused and easy to review.

Merge conflict resolutionEssential
A visual three-way merge view that shows both sides and the common base, turning the worst part of Git into a few clicks.

Interactive rebase and history editing
Reorder, squash, edit, and drop commits by dragging, without hand-editing a rebase todo file.

Hosting integrations
Native connections to GitHub, GitLab, and Bitbucket so you can open pull requests and see CI status without leaving the app.

Cross-platform support
Coverage across macOS, Windows, and Linux matters if your team is mixed or you switch machines often.

Mistakes to avoid

×Leaning on the GUI so hard you never learn the Git underneath. When a client cannot undo a mess, you need to understand the plumbing to recover.

×Picking a client that does not run on your OS or your team's. A Mac-only tool splits a mixed team and forces workarounds.

×Ignoring per-seat cost at team scale. A $10-per-user subscription is nothing solo and a real line item across thirty engineers, and free options cover most workflows.

Expert tips

→Learn your client's keyboard shortcuts. The whole point is speed, and reaching for the mouse to stage every file throws that away.

→Use line-level staging to keep commits focused. Splitting unrelated changes into separate commits makes reviews and reverts painless.

→Keep a terminal client like Lazygit handy even if you prefer a GUI. Some operations, especially over SSH on a server, are faster without a full desktop app.

## The bottom line

For most developers who want one polished GUI across every OS, GitKraken is the pick, as long as the subscription for private repos fits your budget. Prefer to pay once and stay native, Fork is the better deal.

If you live in the terminal, Lazygit is free and hard to beat, and GitHub Desktop covers simple, GitHub-centric work for nothing. Match the client to your OS and billing preference first, because that decides more than the feature list does.

## Frequently asked questions

Are Git GUI clients worth it over the command line?

For most people, yes, at least for the hard parts. Visual staging, merge conflict resolution, and interactive rebase are far safer with a good GUI. Plenty of developers keep both and drop to the CLI when scripting or working on a remote server.

Which Git client is best for beginners?

GitHub Desktop. It is free, uncluttered, and hides the intimidating parts of Git while still teaching the core commit-and-push loop. You can graduate to GitKraken or Fork once you want interactive rebase and more control.

Is there a good free Git client?

Several. GitHub Desktop and Sourcetree are free GUIs, and Lazygit is a free open-source terminal UI. You only pay if you want GitKraken's private-repo support, Fork's one-time license, or Tower's power-user features.

Do these clients work on Linux?

Some. GitKraken and Lazygit run on Linux, but Fork, Tower, GitHub Desktop, and Sourcetree do not. Linux users usually land on GitKraken for a GUI or Lazygit in the terminal.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [GitKraken pricing](https://gitkraken.com/pricing)checked Sep 2026
- [Fork pricing](https://git-fork.com)checked Sep 2026
- [Lazygit pricing](https://github.com/jesseduffield/lazygit)checked Sep 2026
- [Tower pricing](https://git-tower.com/pricing)checked Sep 2026
- [GitHub Desktop pricing](https://desktop.github.com)checked Sep 2026
- [Sourcetree pricing](https://sourcetreeapp.com)checked Sep 2026

Related guides

Docker Desktop AlternativesSelf Hosting ToolsStatic Site GeneratorsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best GraphQL Tools in 2026

URL: https://devshot.co/reviews/best-graphql-tools
Type: review
Published: 2026-07-09
Updated: 2026-09-25
Summary: The GraphQL tools developers actually build with in 2026, spanning servers, clients, instant APIs, and testing, ranked on speed and price.

Expert Guide

## The Best GraphQL Tools in 2026

Build a typed API, query it without guesswork, and generate client code straight from your schema. The servers, clients, and tooling worth adopting, ranked.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 6 tools compared

TL;DR

The best GraphQL tools in 2026 are Apollo for the fullest server-plus-client-plus-platform ecosystem, Hasura for an instant GraphQL API on an existing database, GraphQL Yoga for a lightweight modern server, Postman or Insomnia for testing and exploring APIs, and GraphQL Code Generator for end-to-end type safety. Pick based on whether you are building the API, consuming it, or just poking at it.

Key facts6 tools compared: Apollo, Hasura, GraphQL Yoga, Postman, Insomnia, GraphQL Code Generator

- Updated: September 25, 2026
- Top pick: Apollo (best for: A full GraphQL stack from server to client to platform)
- Top pick price as of September 25, 2026: Apollo: Free tier; Developer from $5 per million requests; Standard and Enterprise custom

- 6 tools compared: Apollo, Hasura, GraphQL Yoga, Postman, Insomnia, GraphQL Code Generator
- Hasura (best for: An instant GraphQL API over an existing database): DDN Free; DDN Base from $5 per active model/month
- GraphQL Yoga (best for: A lightweight, modern GraphQL server): Free (open source)
- Postman (best for: Testing and documenting GraphQL APIs): Free tier; Solo $9/mo, Team $19/user/mo (billed annually)

GraphQL tooling is not one product but a stack: servers that expose a schema, clients that consume and cache it, platforms that stitch many services into one graph, and dev tools that test queries and generate typed code. Some pieces are free and open source, some are commercial platforms with real bills at scale.

We ranked on how quickly you get productive, type-safety support, and cost. Note one big 2026 change up front: Apollo retired its managed cloud router plans and now points teams at self-hosted routers. Here are six tools worth adopting.

## Top Picks

Based on features, real-world fit, and value for money.

Best GraphQL Tools in 2026: 6 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Apollo](https://toolradar.com/tools/apollo) | Free tier; Developer from $5 per million requests; Standard and Enterprise custom | A full GraphQL stack from server to client to platform |

[Hasura](https://toolradar.com/tools/hasura) | DDN Free; DDN Base from $5 per active model/month | An instant GraphQL API over an existing database |

GraphQL Yoga | Free (open source) | A lightweight, modern GraphQL server |

[Postman](https://toolradar.com/tools/postman) | Free tier; Solo $9/mo, Team $19/user/mo (billed annually) | Testing and documenting GraphQL APIs |

[Insomnia](https://toolradar.com/tools/insomnia) | Free tier; paid plans from $12/user/mo | A focused API client for GraphQL and REST |

[GraphQL Code Generator](https://toolradar.com/tools/graphql-code-generator) | Free (open source) | End-to-end type safety from your schema |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Apollo

Top Pick

Best for: A full GraphQL stack from server to client to platform

PricingFree tier; Developer from $5 per million requests; Standard and Enterprise custom

+Server, client, and federation all in one ecosystem

+Apollo Client is the standard for React caching

+Apollo Server and Client are free and open source

−Managed cloud router plans were retired in 2026 in favor of self-hosted routers

−GraphOS platform pricing climbs to enterprise fast

Visit Apollo →

2

### Hasura

Best for: An instant GraphQL API over an existing database

PricingDDN Free; DDN Base from $5 per active model/month

+Fastest path from database to API

+Fine-grained row and column permissions

+Real-time subscriptions out of the box

−Easy to over-expose your schema if permissions are sloppy

−Model-based pricing needs watching as you scale

Visit Hasura →

3

### GraphQL Yoga

Best for: A lightweight, modern GraphQL server

PricingFree (open source)

+Free and open source

+Runs on Node, Bun, Deno, and edge runtimes

+Sensible defaults with plugins when you need them

−No managed platform layer of its own

−Smaller name recognition than Apollo

Visit GraphQL Yoga →

4

### Postman

Best for: Testing and documenting GraphQL APIs

PricingFree tier; Solo $9/mo, Team $19/user/mo (billed annually)

+Team already knows it from REST work

+Schema-aware autocomplete and saved collections

+Strong collaboration and documentation features

−Heavier and more cloud-tied than a focused client

−GraphQL is one feature among many, not the focus

Visit Postman →

5

### Insomnia

Best for: A focused API client for GraphQL and REST

PricingFree tier; paid plans from $12/user/mo

+Fast, uncluttered interface

+Good GraphQL query building and introspection

+Free tier covers individual use

−Some features gated behind a login and paid plans

−Smaller ecosystem than Postman

Visit Insomnia →

6

### GraphQL Code Generator

Best for: End-to-end type safety from your schema

PricingFree (open source)

+Free and open source

+Plugins for React, Vue, Svelte, and more

+Turns your schema into a compile-time safety net

−A build step to configure and maintain

−Only useful in typed codebases like TypeScript

Visit GraphQL Code Generator →

## What it is

GraphQL is a query language for APIs where the client asks for exactly the fields it needs and gets back a predictable shape.

The tools around it fall into categories: servers like Apollo Server and GraphQL Yoga that turn resolvers into an API, clients like Apollo Client that fetch and cache data in the frontend, instant-API engines like Hasura that generate a graph from your database, and testing plus codegen tools that make the whole thing safe to work with. Most projects use two or three of these together.

## Why it matters

REST forces the client to over-fetch, under-fetch, or juggle many endpoints. GraphQL lets the frontend request the precise data it needs in one round trip, and the schema doubles as a contract and living documentation.

Good tooling is what makes that pay off: type generation catches breaking changes at compile time, a client cache saves you from re-fetching, and an explorer lets you build queries without reading source.

The failure mode is exposing too much, since an auto-generated API or an open introspection endpoint can leak your entire data model if you skip auth and query limits.

## Key features to look for

Schema definition and type safetyEssential
A strongly typed schema is the contract between client and server, and the source for generating typed code on both sides.

Query execution and resolversEssential
The server layer that maps each field to a data source, with batching to avoid the classic N+1 query problem.

Caching and client stateEssential
A client cache that normalizes results so repeated queries hit memory instead of the network, which is most of the frontend win.

Introspection and an explorer
An interactive UI to browse the schema and build queries with autocomplete, so you test without digging through code.

Federation and schema composition
Combine multiple service graphs into one unified API, which matters once several teams own different parts of the schema.

Code generation
Generate typed operations, hooks, and models straight from the schema so a breaking change fails the build instead of production.

Mistakes to avoid

×Auto-generating an API over your database and shipping it without tight permissions. An instant API is only safe once row, column, and role rules lock down who can read what.

×Leaving introspection open in production. A public introspection endpoint hands attackers your entire schema, so gate or disable it outside of development.

×Ignoring the N+1 problem in resolvers. Without batching through a tool like DataLoader, one nested query fires hundreds of database calls and quietly tanks performance.

Expert tips

→Generate types from your schema and wire the codegen into CI so a breaking change fails the build instead of a user's request.

→Add persisted queries and a query cost or depth limit to stop a single malicious request from asking for the whole graph at once.

→Start with a lightweight server like GraphQL Yoga and only reach for federation or a full platform once multiple teams share the schema.

## The bottom line

For a team building a serious GraphQL API, Apollo is the safe default, since one ecosystem covers server, client, and federation, just know the managed router is now self-hosted.

Need an API over an existing database in an afternoon, Hasura is unmatched, and GraphQL Yoga is the cleaner pick when you want a lightweight server you fully control.

Pair whichever you choose with GraphQL Code Generator for type safety, and keep Postman or Insomnia around for testing.

## Frequently asked questions

Do I need Apollo to use GraphQL?

No. Apollo is the most complete ecosystem, but GraphQL Yoga and other servers work just as well, and you can mix a non-Apollo server with any client. Apollo Client is popular on the frontend, yet plenty of teams use lighter caching libraries instead.

What is the fastest way to get a GraphQL API?

Hasura. Connect it to an existing database and it generates a working GraphQL API with queries, mutations, and subscriptions in minutes. Just configure permissions before exposing it, or you risk leaking your whole data model.

Postman or Insomnia for GraphQL?

Postman if your team already lives in it and wants collaboration and documentation. Insomnia if you want a lighter, more focused client. Both handle GraphQL introspection and query building well, so it often comes down to which feels less cluttered to you.

Is GraphQL federation worth it?

Only once several teams own different parts of the schema. Federation lets each team ship its own subgraph while clients see one unified graph. For a single service owned by one team, it adds complexity you do not need yet.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Apollo pricing](https://apollo.io/pricing)checked Sep 2026
- [Hasura pricing](https://hasura.io/pricing)checked Sep 2026
- [Postman pricing](https://www.postman.com/pricing/)checked Sep 2026
- [Insomnia pricing](https://insomnia.rest/pricing)checked Sep 2026
- [GraphQL Code Generator pricing](https://graphql-code-generator.com/#pricing)

Related guides

Git ClientsSelf Hosting ToolsPostman AlternativesAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Internal Developer Portals in 2026

URL: https://devshot.co/reviews/best-internal-developer-portals
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: The best internal developer portals in 2026, compared on what a platform team actually pays per seat, per user, or per quote…

Expert Guide

## The Best Internal Developer Portals in 2026

Backstage costs nothing to install and Humanitec is the priciest platform here, and the tools between them price by seat, by user, or by quote only.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 9 tools compared

TL;DR

Backstage is free and open source, the CNCF project most of this category is either built on or measured against. Port is the strongest managed alternative: free up to 15 seats, then $30/seat/mo on Basic and $40/seat/mo on Standard, both billed annually. Humanitec is the most expensive platform here, from $2,199/mo for five users.

Between those two ends, Atlassian Compass starts at $7.67/user/mo, Configure8 at $22.89/user/mo, and Cortex, OpsLevel and Harness IDP publish no list price at all: budget for a sales call instead.

Key facts9 tools compared: Backstage, Port, Cortex, OpsLevel, Roadie, Configure8, Atlassian Compass…

- Updated: September 25, 2026
- Top pick: Backstage (best for: Platform teams with 2 or more engineers who want a catalog with no vendor lock-in or license fee)
- Top pick price as of September 25, 2026: Backstage: Free and open source; self-hosted, so the real cost is the engineer who runs and upgrades it.

- 9 tools compared: Backstage, Port, Cortex, OpsLevel, Roadie, Configure8, Atlassian Compass, Harness IDP, Humanitec
- Port (best for: Teams that want Backstage's catalog idea without operating Backstage themselves): Free up to 15 seats; Basic $30/seat/mo, Standard $40/seat/mo, both billed annually; Enterprise: custom.
- Cortex (best for: Larger engineering orgs that want scorecards and maturity tracking and already budget for a sales cycle): Quote-only per-seat pricing that scales with org size; publishes no list price.
- OpsLevel (best for: Teams that want AI-assisted catalog enrichment and are comfortable requesting a quote): Standard and Enterprise are both quote-only ('Get pricing'); publishes no list price.

An internal developer portal is the page a new engineer opens before asking in Slack who owns a service, where its logs live, or which template to clone.

[Backstage](https://backstage.io/) is the reason this exists as software rather than a wiki: Spotify built it, handed it to the CNCF, and most vendors below either sell a hosted version of it or compete directly against it.

Toolradar data: the [Toolradar guide to this category](https://toolradar.com/guides/best-internal-developer-portals) tracks 1,627 developer tools as of September 2026, and puts internal developer portals in the minority of categories where a genuinely free, unlimited-feature option (Backstage) competes directly against platforms billing thousands a month, like [Humanitec](https://toolradar.com/tools/humanitec).

A broader, buyer-facing breakdown of the same category runs on [Dupple's internal developer portals guide](https://dupple.com/learn/best-internal-developer-portals).

Prices below were verified on each vendor's own pricing page on September 25, 2026.

## Top Picks

Based on features, real-world fit, and value for money.

Best Internal Developer Portals in 2026: 9 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Backstage | Free and open source; self-hosted, so the real cost is the engineer who runs and upgrades it. | Platform teams with 2 or more engineers who want a catalog with no vendor lock-in or license fee |

Port | Free up to 15 seats; Basic $30/seat/mo, Standard $40/seat/mo, both billed annually; Enterprise: custom. | Teams that want Backstage's catalog idea without operating Backstage themselves |

Cortex | Quote-only per-seat pricing that scales with org size; publishes no list price. | Larger engineering orgs that want scorecards and maturity tracking and already budget for a sales cycle |

OpsLevel | Standard and Enterprise are both quote-only ('Get pricing'); publishes no list price. | Teams that want AI-assisted catalog enrichment and are comfortable requesting a quote |

Roadie | Teams $24/dev/mo, 50-150 developers, existing subscribers only; Growth and Enterprise Context: custom. | Teams that want Backstage without operating it, if they can still get a quote |

Configure8 | Free up to 10 scorecard/action users; Enterprise $22.89/mo or $228/yr per user, about $19/mo annual. | Small platform teams, under about 50 engineers, that want cloud cost tracking built into the catalog |

Atlassian Compass | Free up to 3 full users; Standard $7.67/user/mo, Premium $23.96/user/mo; unlimited basic users on every plan. | Teams already paying for Jira and Confluence who want a catalog in the same admin console |

Harness IDP | Bundled inside the Enterprise plan only, one of four Operations modules; no standalone price, contact sales. | Teams already buying or evaluating Harness for CI/CD who want the portal in the same contract |

Humanitec | Teams $2,199/mo, 5 users; Pro $5,499/mo, 50 users; Enterprise and self-hosted: custom quote. | Platform teams building golden paths that provision real infrastructure, not just a catalog |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 5 of 9 do not publish a comparable monthly price and are left out rather than estimated.

1

### Backstage

Top Pick

Best for: Platform teams with 2 or more engineers who want a catalog with no vendor lock-in or license fee

PricingFree and open source; self-hosted, so the real cost is the engineer who runs and upgrades it.

+No license fee at any team size, and no seat cap or entity cap to negotiate around

+CNCF Incubating status keeps a growing open source plugin ecosystem outside any single vendor

+Full control of hosting, so catalog and ownership data never leave infrastructure you already run

−No vendor support line: upgrades, plugin maintenance and uptime are your platform team's job

−Most teams need at least one dedicated engineer to keep it running, a real salary, not a subscription line

Visit Backstage →

2

### Port

Best for: Teams that want Backstage's catalog idea without operating Backstage themselves

PricingFree up to 15 seats; Basic $30/seat/mo, Standard $40/seat/mo, both billed annually; Enterprise: custom.

+The Free plan covers 15 seats, 10K entities and 400 automation runs a month with no card required

+Basic and Standard both include commercial support and a 99.8% uptime SLA, not just community help

+Every plan gets the full platform (Context Lake, actions, AI agents, scorecards); features are not tiered

−Basic caps at 50 seats and Standard at 200, so a company past 200 developers is already quoting Enterprise

−Standard's per-seat rate scales to $4,000/mo for a 100-person team before any add-on

Visit Port →

3

### Cortex

Best for: Larger engineering orgs that want scorecards and maturity tracking and already budget for a sales cycle

PricingQuote-only per-seat pricing that scales with org size; publishes no list price.

+Per-seat pricing scales with org size, so a large buyer can negotiate volume the way they would with any enterprise tool

+No published tier limits to hit and re-negotiate around, since the whole plan is custom from the start

+Built specifically around scorecards and engineering maturity, a narrower and deeper focus than a general catalog

−No self-serve trial and no published starting price, so you cannot budget for it without talking to sales first

−A small team cannot test the product without becoming a sales lead

Visit Cortex →

4

### OpsLevel

Best for: Teams that want AI-assisted catalog enrichment and are comfortable requesting a quote

PricingStandard and Enterprise are both quote-only ('Get pricing'); publishes no list price.

+Standard already includes SSO, AI-powered catalog enrichment and doc summaries, features some rivals reserve for a top tier

+Up to 50 users on Standard covers most mid-size platform teams before Enterprise is even a conversation

+Enterprise adds on-prem and single-tenancy options for teams that cannot use a shared SaaS instance

−Two tiers, zero published prices; you cannot compare it to Port or Compass without requesting a quote from each

−No self-hosted or open-source path if the budget for a SaaS contract disappears

Visit OpsLevel →

5

### Roadie

Best for: Teams that want Backstage without operating it, if they can still get a quote

PricingTeams $24/dev/mo, 50-150 developers, existing subscribers only; Growth and Enterprise Context: custom.

+Runs the actual open source Backstage catalog, so plugins and upgrades follow the same project, not a proprietary fork

+SSO and a self-hosted-grade feature set (TechDocs, API specs, self-service templates) come standard, on top of the open source Backstage plugin directory

+Billing counts only contributing developers who write code tracked in the catalog, so non-coding team members get access at no charge

−The published Teams price and the custom Growth price are both marked existing subscribers only on Roadie's own pricing page

−New visitors are currently pointed to a waitlist rather than a signup flow, so plan for a sales conversation, not a card swipe

Visit Roadie →

6

### Configure8

Best for: Small platform teams, under about 50 engineers, that want cloud cost tracking built into the catalog

PricingFree up to 10 scorecard/action users; Enterprise $22.89/mo or $228/yr per user, about $19/mo annual.

+Enterprise's annual rate (about 17% off the monthly rate) publishes an actual number, unlike Cortex, OpsLevel or Harness IDP

+Cloud cost tracking is native, not a bolt-on: the Free tier tracks up to $7,500/mo, Enterprise is unlimited

+The Free plan allows unlimited integrations and catalog users up to a data cap, with no credit card required

−Under roughly 50 engineers, Configure8 states a $1,000/mo minimum, which can price out a very small team

−The Free tier has no API access, RBAC, audit logs or SSO, so it will not pass most security reviews

Visit Configure8 →

7

### Atlassian Compass

Best for: Teams already paying for Jira and Confluence who want a catalog in the same admin console

PricingFree up to 3 full users; Standard $7.67/user/mo, Premium $23.96/user/mo; unlimited basic users on every plan.

+Standard is the lowest named per-seat price in this category, with a 14-day free trial and no card required

+Unlimited basic users on every plan, including Free, means only the people who edit scorecards count against the price

+Premium adds a 99.9% uptime SLA and 24/7 support with a 1-hour response time, a jump few competitors match at that price

−Free caps at just 3 full users, so most teams outgrow it within the first week

−Full value depends on already living in Jira and Confluence; it is not built as a standalone catalog

Visit Atlassian Compass →

8

### Harness IDP

Best for: Teams already buying or evaluating Harness for CI/CD who want the portal in the same contract

PricingBundled inside the Enterprise plan only, one of four Operations modules; no standalone price, contact sales.

+Sits next to CI/CD, feature flags and cloud cost management in one contract, so a service's pipeline and its catalog entry live in the same platform

+Enterprise-grade features apply across every bundled module, not a stripped-down add-on tier

+One vendor relationship instead of stitching a separate catalog tool onto an existing Harness deployment

−No standalone IDP plan or published price; you buy the whole Harness Enterprise bundle to get it

−Not a fit if you are not already using or evaluating Harness for CI/CD, since the portal cannot be purchased alone

Visit Harness IDP →

9

### Humanitec

Best for: Platform teams building golden paths that provision real infrastructure, not just a catalog

PricingTeams $2,199/mo, 5 users; Pro $5,499/mo, 50 users; Enterprise and self-hosted: custom quote.

+Teams and Pro both include a free trial with no credit card, unusual at this price point

+Pro's unlimited environments and 50 included users suit a platform team standardizing golden paths across many services

+A self-hosted plan exists for regulated teams that cannot run workloads through a third-party SaaS control plane

−Its entry price is the steepest in this list by a wide margin for just 5 users

−It is an infrastructure orchestrator with a portal on top, not a lightweight catalog, overkill if all you need is a service directory

Visit Humanitec →

## What it is

An internal developer portal (IDP) is a single interface over your service catalog, docs, templates and ownership data, usually built as a UI layer on top of systems you already run: CI/CD, Kubernetes, and source control.

It answers three questions on one page: what exists, who owns it, and how do I get a new one.

That is a different job from a wiki, which stores prose that goes stale, and from an API documentation tool, which describes one API rather than every service.

An IDP is also not a CI/CD platform: it is the front door to one, with a scorecard or template layered on top so a self-service action does not require reading a runbook first.

## Why it matters

The pricing model splits three ways, and it changes who signs the contract. Backstage is free software, so its real cost is the platform engineer who maintains it, typically the reason teams adopt Backstage only once they have two or more people to run it.

Port, Compass and Configure8 price per seat or per user, so the bill scales with headcount and a procurement team can forecast it. Cortex, OpsLevel and Harness IDP are quote-only, so the number depends on the sales call.

Watch the fine print on managed Backstage.

Roadie's published $24/dev/mo Teams plan and its custom Growth plan are both marked existing subscribers only on its own pricing page, and the site's navigation currently points new visitors to a waitlist rather than a signup form.

A team evaluating Roadie today should expect a custom quote, not the printed sticker.

Catalog enrichment is also splitting on AI: [OpsLevel](https://toolradar.com/tools/opslevel) and [Cortex](https://toolradar.com/tools/cortex) both build AI-assisted descriptions and doc summaries into their pricing tiers, a feature set that overlaps with the tools on the AI for DevOps guide.

Devshot data: 28,000 developers read this newsletter, with a 30% average open rate, from the audience file updated September 20, 2026.

How we compared: nine internal developer portal platforms with a pricing page a buyer can read today, checked on September 25, 2026, ranked on published price, seat model, and what a self-hosted alternative costs to skip.

## Key features to look for

License and hosting model
Open source and self-hosted (Backstage) versus managed SaaS (Port, Cortex, OpsLevel, Roadie, Configure8, Compass, Harness IDP) versus a platform orchestrator you can also self-host (Humanitec). This decides whether the bill is a subscription or a headcount.

Seat-based vs quote-only pricing
Port, Compass and Configure8 publish a per-seat or per-user rate you can build a budget from. Cortex, OpsLevel, Harness IDP and Humanitec's Enterprise tier require a sales call before you see a number.

What counts as a seat
Some platforms bill every developer who touches the catalog. Atlassian Compass splits paying full users from unlimited free basic users who only browse, so the seat count and the developer count are not the same number.

Catalog vs golden paths
A catalog that lists services and owners is the baseline every tool here provides. Humanitec and Harness IDP go further and provision the infrastructure itself, which is a large part of why they cost more.

Entity and automation-run caps
Port's Free and Basic plans cap both catalog entities and automation runs a month. Configure8 caps cloud cost tracking at $7,500/mo on its free tier. Hit either ceiling and the next tier is the only way past it.

## Pricing

Every price above was read on the vendor's own pricing page on September 25, 2026, in USD. Backstage is open source with no vendor price to check. Cortex, OpsLevel and Harness IDP publish no list price for any tier, so budget for a quote rather than a sticker.

Port's Basic and Standard rows are billed annually per seat, with no monthly option published. Configure8's Enterprise plan shows a monthly rate ($22.89/user/mo) and an annual rate ($228/user/yr, about $19/mo), a 17% discount for prepaying, plus a stated $1,000/mo minimum for teams under roughly 50 engineers.

Atlassian Compass prices Standard and Premium per user per month, with no annual option published yet.

Roadie is the outlier: its Teams and Growth plans both carry an existing subscribers only label, and the site steers new visitors to a waitlist. Treat the printed Teams price as historical, not a price a new buyer can order today.

Humanitec is the only platform here with a currency toggle, showing both EUR and USD; the figures used are the USD column, with a stated 10% discount for annual billing not printed as a dollar amount.

Plan | Price | Best for |

Backstage | Free | Open source, self-hosted; CNCF Incubating project, community support only |

Port Free | Free | Up to 15 seats, 10K entities, 400 automation runs a month |

Port Basic | $30/seat/mo | Up to 50 seats, 50K entities, commercial support, 99.8% uptime SLA; billed annually |

Port Standard | $40/seat/mo | Up to 200 seats, 250K entities, SSO, 5 workspaces; billed annually |

Port Enterprise | Custom quote | Unlimited seats, 1M+ entities, SCIM, private link, 99.9% uptime SLA |

Cortex | Custom quote | Per-seat pricing scaled to org size; no published rate |

OpsLevel Standard | Custom quote | Up to 50 users, unlimited components, AI-powered catalog features, SSO |

OpsLevel Enterprise | Custom quote | Unlimited users, custom integrations, dedicated success manager |

Roadie Teams | $24/dev/mo | 50-150 developers; existing subscribers only, new buyers see a custom quote |

Roadie Growth | Custom quote | 100+ developers; existing subscribers only |

Configure8 Free | Free | Unlimited catalog users to a cap; 10 scorecard/action users; cloud cost tracking to $7,500/mo |

Configure8 Enterprise | $228/user/yr | About $19/mo paid annually; unlimited users, RBAC, audit logs, SSO; $1,000/mo team minimum |

Harness IDP | Custom quote | One of four Operations modules inside Harness Enterprise; no standalone plan |

Compass Free | Free | Up to 3 full users, unlimited basic users, unlimited services |

Compass Standard | $7.67/user/mo | Up to 35,000 full users, 50 custom fields, 1-year data retention |

Compass Premium | $23.96/user/mo | 99.9% uptime SLA, 2-year retention, 24/7 support, IP allowlisting |

Humanitec Teams | $2,199/mo | 5 users, up to 2 projects with up to 5 environments each, email support |

Humanitec Pro | $5,499/mo | 50 users, unlimited environments, RBAC, sandbox organizations |

Humanitec Enterprise / Self-Hosted | Custom quote | SAML, audit logs, dedicated architect, or fully self-hosted deployment |

Mistakes to avoid

×Comparing Backstage's $0 price tag to Port's or Compass's per-seat rate without pricing the engineer who has to run and upgrade Backstage; the software is free, the operating cost is not.

×Quoting Roadie's Teams plan as a price a new customer can order today, when the plan is marked existing subscribers only and the site points new visitors to a waitlist.

×Signing an Enterprise-only contract with Cortex, OpsLevel or Harness IDP before a named-price option (Port, Compass, Configure8) has been priced for the same headcount.

×Counting every developer as a paid seat on Atlassian Compass, when unlimited basic users are free and only full users who edit scorecards count against the per-user price.

×Buying Humanitec or Harness IDP for a service catalog alone, when both are infrastructure orchestrators priced well above catalog-only tools like Port or Configure8.

Expert tips

→Start with Backstage if a platform engineer already exists to run it: no license fee, no seat cap, and the plugin ecosystem behind most of the vendors on this list.

→Ask Cortex, OpsLevel and Harness IDP for a number at the same headcount you would use to price Port or Compass, so the quote-only vendors are not competing against an outdated internal estimate.

→If the team mostly needs alerting and pipeline visibility alongside the catalog, price the <a href="/reviews/best-monitoring-tools">monitoring tools</a> and a Git host from the <a href="/reviews/best-git-clients">Git clients guide</a> alongside whichever portal you pick, rather than expecting one tool to cover all three.

## The bottom line

Backstage is the right default if a platform engineer already exists to run it: no license fee, no seat cap, and the plugin ecosystem behind most of the vendors on this list.

Port is the best managed alternative, free to 15 seats and priced per seat once you outgrow that, with commercial support and an SLA that neither Backstage nor the quote-only tools publish.

Pick Atlassian Compass if the team already lives in Jira and Confluence and wants a catalog at its published entry rate.

Pick Configure8 for built-in cloud cost tracking at a named enterprise price.

Go to Cortex or OpsLevel only once a sales call is acceptable, and reserve Harness IDP or Humanitec for teams that need a bundled CI/CD contract or real infrastructure provisioning, not just a catalog.

Cite this: Devshot, "Best Internal Developer Portals in 2026", September 2026.

## Frequently asked questions

What is the best internal developer portal in 2026?

Backstage, if a platform engineer already runs it, because it is free, open source, and the CNCF project most of this category is built on or measured against. Port is the best managed alternative, free up to 15 seats and priced per seat after that. Atlassian Compass is the cheapest named per-seat price for teams already on Jira and Confluence.

Is there a free internal developer portal?

Yes. Backstage is free and open source at any team size, though you operate and host it yourself, similar to running your own Terraform or IaC tooling instead of buying a managed platform. Port's Free plan covers up to 15 seats, 10K entities and 400 automation runs a month with no card required. Atlassian Compass is free for up to 3 full users with unlimited basic, read-only users.

How much does Backstage cost to run?

Backstage has no license fee. The real cost is engineering time: most teams need at least one platform engineer to install, maintain and upgrade it, plus the infrastructure it runs on. That is why managed alternatives like Port, Roadie or Compass exist, they trade a per-seat bill for that operating cost.

Port vs Backstage: which should I pick?

Backstage if you want to own the catalog with no license fee and already have engineering time to run it. Port if you want a managed version of the same idea: free up to 15 seats, then $30/seat/mo on Basic or $40/seat/mo on Standard, both billed annually, with commercial support and a 99.8% uptime SLA that self-hosted Backstage does not come with out of the box.

Is Roadie still accepting new customers?

Roadie's own pricing page marks both its Teams ($24/dev/mo) and Growth (custom) plans existing subscribers only, and its site currently points new visitors to a waitlist rather than a signup form. Treat the published price as historical and expect a custom quote if you are evaluating Roadie today.

What does Humanitec cost for a growing platform team?

Humanitec's Teams plan is $2,199/mo for 5 users and one project with up to 5 environments. Pro is $5,499/mo for 50 users and unlimited environments. Both include a free trial with no credit card. Enterprise and the self-hosted option are quote-only.

Do I need a dedicated IDP if I already use Atlassian Compass or Jira?

Atlassian Compass is itself an internal developer portal, priced from its published Standard rate, and it is the natural first stop if the team already pays for Jira and Confluence. A separate tool like Port, Cortex or Backstage is worth it once you need deeper self-service actions, AI-assisted enrichment, or a catalog not tied to Atlassian's admin console.

Which internal developer portal has the cheapest published price?

Atlassian Compass Standard is the lowest named per-seat price. Configure8's Enterprise plan is close behind on its annual rate. Backstage is free but unpriced because it is self-hosted software, not a vendor plan.

Related guides

Ci Cd ToolsKubernetes ToolsTerraform Iac ToolsMonitoring ToolsAi For DevopsApi Documentation ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Log Management Tools in 2026

URL: https://devshot.co/reviews/best-log-management-tools
Type: review
Published: 2026-08-28
Updated: 2026-09-25
Summary: The log management platforms worth running in 2026, compared on what they charge per gigabyte, how long they keep your data…

Expert Guide

## The Best Log Management Tools in 2026

One of these gives you 500GB a month free. Another gives you 3GB. Both call it a free tier.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

Axiom has the most generous free tier in the category by two orders of magnitude: 500GB a month, permanently, with Cloud at $25 including 1TB. Better Stack is the cleanest bundle pricing, from $30 a month for 40GB. Grafana Cloud gives 50GB of logs free and fits if you already run Grafana. New Relic includes 100GB a month and charges $0.40 per GB after.

Key facts5 tools compared: Axiom, Better Stack, Grafana Cloud, New Relic, Datadog

- Updated: September 25, 2026
- Top pick: Axiom (best for: Teams that want to stop rationing their logs)
- Top pick price as of September 25, 2026: Axiom: From $25/mo (Cloud, platform fee); free Personal plan ($0/month, permanent)

- 5 tools compared: Axiom, Better Stack, Grafana Cloud, New Relic, Datadog
- Better Stack (best for: Teams that want a predictable bundle rather than a usage meter): From $30/mo (Nano, Europe); free plan (3GB logs, retained 3 days)
- Grafana Cloud (best for: Anyone already running Grafana dashboards): Free tier (50GB logs, 14-day retention); paid tiers scale by usage
- New Relic (best for: Teams that want one platform for logs, APM and everything else): Free tier (100GB ingest/month); overage $0.40/GB; Standard, Pro, Enterprise scale up

Log management is priced by the gigabyte, which means your bill is set by how chatty your services are rather than by how many people use the tool.

That is an uncomfortable relationship: the cheapest way to cut the invoice is to log less, which is the opposite of what you want during an incident.

So the number that matters is the included volume, and the spread here is extraordinary.

Axiom's free plan takes 500GB a month. Better Stack's free plan takes 3GB. Same category, same word, a 166-fold difference.

## Top Picks

Based on features, real-world fit, and value for money.

Best Log Management Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Axiom | From $25/mo (Cloud, platform fee); free Personal plan ($0/month, permanent) | Teams that want to stop rationing their logs |

Better Stack | From $30/mo (Nano, Europe); free plan (3GB logs, retained 3 days) | Teams that want a predictable bundle rather than a usage meter |

[Grafana Cloud](https://toolradar.com/tools/grafana-cloud) | Free tier (50GB logs, 14-day retention); paid tiers scale by usage | Anyone already running Grafana dashboards |

[New Relic](https://toolradar.com/tools/new-relic) | Free tier (100GB ingest/month); overage $0.40/GB; Standard, Pro, Enterprise scale up | Teams that want one platform for logs, APM and everything else |

[Datadog](https://toolradar.com/tools/datadog) | $0.10 per GB ingested, plus $1.70 per million events indexed (15-day retention, annual); free trial | Large estates that want the deepest integration coverage |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

1

### Axiom

Top Pick

Best for: Teams that want to stop rationing their logs

PricingFrom $25/mo (Cloud, platform fee); free Personal plan ($0/month, permanent)

+500GB a month free, permanently, with no credit card

+Cloud at $25 includes a full terabyte of ingest

+Query compute is stated explicitly rather than hidden

−Smaller ecosystem than the incumbents

−Query compute is a separate allowance to track

Visit Axiom →

2

### Better Stack

Best for: Teams that want a predictable bundle rather than a usage meter

PricingFrom $30/mo (Nano, Europe); free plan (3GB logs, retained 3 days)

+Genuinely transparent pricing, ingest and retention both published

+Bundles make the monthly bill predictable

+Uptime monitoring and incident management in the same product

−Free tier is only 3GB retained for 3 days

−Regional pricing means Singapore costs roughly triple Europe

Visit Better Stack →

3

### Grafana Cloud

Best for: Anyone already running Grafana dashboards

PricingFree tier (50GB logs, 14-day retention); paid tiers scale by usage

+50GB of logs on the free tier, well above most rivals

+Same query language and dashboards as self-hosted Grafana

+Metrics, logs and traces genuinely unified rather than bolted together

−14-day retention on free, shorter than an average incident review cycle

−LogQL takes time if you have never used it

Visit Grafana Cloud →

4

### New Relic

Best for: Teams that want one platform for logs, APM and everything else

PricingFree tier (100GB ingest/month); overage $0.40/GB; Standard, Pro, Enterprise scale up

+100GB a month free covers a lot of small production estates

+APM, logs, traces and browser monitoring in one place

+Published $0.40 per GB overage, so the downside is calculable

−Per-user pricing on top of ingest once you leave the free tier

−The breadth is overwhelming if you only want logs

Visit New Relic →

5

### Datadog

Best for: Large estates that want the deepest integration coverage

Pricing$0.10 per GB ingested, plus $1.70 per million events indexed (15-day retention, annual); free trial

+Integration coverage nothing else here matches

+Excellent correlation between logs, traces and infrastructure

+The safe institutional choice

−Ingest and indexing are billed separately, so the total is hard to predict

−Regularly the most expensive option in a bake-off

Visit Datadog →

## What it is

A log management platform ingests structured and unstructured logs from your services, indexes them so you can search across sources, retains them for a period, and alerts on patterns.

Modern ones bundle traces and metrics alongside, which is where the observability label comes from.

Pricing has three moving parts and vendors emphasise whichever flatters them: ingest per gigabyte, retention per gigabyte per month, and sometimes query compute.

A cheap ingest rate with expensive retention is a common shape, and it only shows up on the second invoice.

## Why it matters

Every serious incident review ends at the same question: what did the logs say. If your retention window is shorter than your detection time, the answer is nothing, and the tool has failed at the one job it was bought for.

The second reason is behavioural.

Teams on aggressive per-gigabyte pricing start sampling their logs, then dropping debug levels, then removing log lines. Each step is rational and the cumulative effect is that the expensive observability platform observes less than the free one would have.

## Key features to look for

Included ingest volume
How many gigabytes a month the plan takes before overage. The single number that decides your bill, and the one the pricing page states least clearly.

Retention window
How long logs stay searchable. Thirty days is common; free tiers often drop to three. If it is shorter than your time to detect a problem, the tool cannot answer the question you will ask it.

Overage rate
What each extra gigabyte costs once you pass the allowance. New Relic publishes $0.40 per GB; Better Stack runs $0.10 to $0.35 depending on region. This is what your bill actually moves on.

Query cost
Whether searching your own data costs extra. Some platforms meter query compute separately, which turns a long incident investigation into a line item.

Structured search
Querying by field rather than grepping strings. The difference between finding the failing request in seconds and reading pages of output.

Traces and metrics alongside
Whether the same platform holds the other two pillars. Consolidating usually costs less than three vendors, but check the bundle's per-signal allowances.

## Pricing

Log management is priced on ingest per gigabyte, retention per gigabyte per month, and sometimes query compute, so the included volume sets the bill more than the platform fee.

Axiom and Better Stack publish full rate cards, checked 28 August 2026: Axiom Personal is $0 a month and Cloud is a $25 platform fee, and Better Stack bundles start at $30 a month in Europe. Grafana paid tiers scale by usage, and Datadog bills ingest per GB plus $1.70 per million events indexed for 15 days on annual billing.

New Relic publishes a free allowance and $0.40 per GB of overage, while Standard, Pro and Enterprise scale up without listed prices. The cheapest credible entry with real volume is Axiom's permanent free plan.

Costs jump at Better Stack's $500 Tera bundle, on higher US and Singapore rates, past the included gigabytes, and when retention or query compute is billed apart from ingest.

Plan | Price | Best for |

Axiom Personal | $0/month, permanent | 500GB/mo data loading, 10 GB-hours query compute, 25GB storage |

Axiom Cloud | $25/month platform fee | 1TB/mo loading, 100 GB-hours query, 100GB storage |

Axiom Cloud usage | Usage-based | After the included Cloud loading, query and storage |

Better Stack Free | Free | 3GB logs and 3GB traces retained 3 days, 30GB metrics |

Better Stack Nano | $30/mo (Europe) | 40GB each; US and Singapore cost more |

Better Stack Tera | $500/mo | 700GB, with 30-day retention on bundles |

Better Stack ingest | $0.10 to $0.35/GB | Pay-as-you-go ingest, priced by region |

Better Stack retention | $0.05 to $0.18/GB/mo | Pay-as-you-go retention per GB per month |

Grafana Cloud Free | Free | 10,000 metric series, 50GB logs, 50GB traces, 14-day retention |

Grafana Cloud paid | Scales by usage | Paid tiers above the free allowance |

New Relic Free | Free | 100GB ingest per month |

New Relic Standard | Scales up | Paid tier above the free allowance |

New Relic Pro | Scales up | Named paid tier, price scales up |

New Relic Enterprise | Scales up | Named paid tier, price scales up |

New Relic overage | $0.40/GB | Each extra gigabyte past included ingest |

Datadog Log Management | $1.70 per million events indexed | Plus a per-GB ingest fee; 15-day retention, billed annually |

Mistakes to avoid

×Comparing free tiers without reading the retention. Better Stack's free plan keeps logs for 3 days and Grafana Cloud's for 14. A 3-day window cannot answer a question raised in a Monday incident review about Friday night.

×Ignoring the region multiplier. Better Stack's Nano bundle is $30 in Europe and $105 in Singapore for the same 40GB. Vendors rarely surface this in a comparison table and it can triple the bill.

×Cutting log volume to cut the bill. It works, and it is the wrong lever: you are buying the tool precisely for the moment when you need the line you deleted. Change tiers or change vendors instead.

Expert tips

→Measure a week of real ingest before choosing. Most teams are wrong about their own volume by a factor of two, and every price here is a function of that number.

→Price the overage, not the plan. New Relic's $0.40 per GB and Better Stack's $0.10 to $0.35 tell you what a bad month costs, which the headline never does.

→Check whether querying costs extra. Axiom states query compute as its own allowance, which is honest and worth budgeting; a platform that meters it silently turns a long investigation into an invoice.

## The bottom line

Axiom is the standout on economics: 500GB a month free permanently, and $25 for a terabyte.

If your logs are the thing you keep rationing, that changes the calculation more than any feature here.

Better Stack is the pick when you want a bundle you can forecast, from $30 a month, and it publishes the clearest rate card of the group.

Grafana Cloud is the obvious choice if you already live in Grafana, with 50GB free. New Relic makes sense if you want logs and APM together, and Datadog when integration breadth outweighs cost.

## Frequently asked questions

Which log tool has the best free tier?

Axiom, by a very large margin: 500GB a month of data loading, permanently, with 25GB of storage and no credit card. Grafana Cloud is second with 50GB of logs at 14-day retention. New Relic includes 100GB a month. Better Stack's free tier is 3GB retained for 3 days. All checked 28 August 2026.

How is log management priced?

Almost always per gigabyte, split between ingest and retention, and sometimes query compute on top. Better Stack publishes $0.10 to $0.35 per GB ingested and $0.05 to $0.18 per GB per month retained, varying by region. New Relic publishes $0.40 per GB of overage. Datadog adds a second meter on top of ingest, charging per million log events indexed, which makes it hard to compare directly.

How long should we retain logs?

Long enough to cover the gap between something happening and someone asking about it, which in most teams is longer than they think. Thirty days is the usual paid default and a reasonable target. Free tiers at 3 or 14 days are fine for development and will not survive a real incident review.

Should logs, metrics and traces live in one platform?

Usually yes, because correlating across three vendors during an incident wastes the time you have least of. The caveat is that bundles allocate per signal, so check that the logs allowance is not being subsidised by metrics you do not send.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Grafana Cloud pricing](https://grafana.com/pricing)checked Sep 2026
- [New Relic pricing](https://newrelic.com/pricing)checked Sep 2026
- [Datadog pricing](https://datadoghq.com/pricing)checked Sep 2026

Related guides

Monitoring ToolsError Tracking ToolsServerless PlatformsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Monitoring Tools in 2026

URL: https://devshot.co/reviews/best-monitoring-tools
Type: review
Published: 2026-09-02
Updated: 2026-09-25
Summary: The best monitoring tools in 2026, compared on pricing model rather than dashboards. Datadog, Grafana Cloud, New Relic, Sentry and Prometheus…

Expert Guide

## The Best Monitoring Tools in 2026

Observability platforms compared on the thing that decides the bill: how each one charges for the data you send it.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

Pick by billing model, because that is what you will argue about in six months. Datadog is the most complete and the easiest to overspend on: infrastructure from $15 per host per month on annual, with logs at $0.10 per ingested GB plus $1.70 per million indexed events. Grafana Cloud has the most generous free tier here (10k metric series, 50 GB of logs) and usage-based Pro pricing, and it is the natural choice if you want the option of self-hosting later. New Relic charges by data ingested plus full-platform users, which suits small teams with a lot of telemetry and punishes large ones. Sentry is not a full observability platform and is the fastest way to find out why your application is failing, free for one developer and $26/month on Team. Prometheus is free and open source, and costs you an engineer instead.

Key facts5 tools compared: Datadog, Grafana Cloud, New Relic, Sentry, Prometheus

- Updated: September 25, 2026
- Top pick: Datadog (best for: Teams that want one platform for everything and can govern their own spend)
- Top pick price as of September 25, 2026: Datadog: From $15/host/mo (Infrastructure, billed annually); free for 5 hosts

- 5 tools compared: Datadog, Grafana Cloud, New Relic, Sentry, Prometheus
- Grafana Cloud (best for: Teams who want open standards and a credible exit): From $19/mo plus $6.50 per 1,000 metric series (Pro); free tier
- New Relic (best for: Small teams sending a lot of telemetry): $0.40/GB ingested after 100 GB free a month; 1 full user free
- Sentry (best for: Finding out why the application is broken, fastest): From $26/mo (Team, billed annually); free Developer plan ($0)

Every platform on this list will show you a graph of CPU. That is not the decision.

The decision is what happens when a noisy service starts emitting ten times the logs it did last month, because that is when observability stops being a tool and becomes a line item somebody asks you to justify.

Monitoring vendors compete on dashboards and sell on ingest.

Per-host pricing looks predictable until you count containers. Per-GB pricing looks fair until a debug flag ships to production on a Friday. The comparison below leads with the billing model of each platform, because features converge and pricing does not.

## Top Picks

Based on features, real-world fit, and value for money.

Best Monitoring Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Datadog](https://toolradar.com/tools/datadog) | From $15/host/mo (Infrastructure, billed annually); free for 5 hosts | Teams that want one platform for everything and can govern their own spend |

[Grafana Cloud](https://toolradar.com/tools/grafana-cloud) | From $19/mo plus $6.50 per 1,000 metric series (Pro); free tier | Teams who want open standards and a credible exit |

[New Relic](https://toolradar.com/tools/new-relic) | $0.40/GB ingested after 100 GB free a month; 1 full user free | Small teams sending a lot of telemetry |

[Sentry](https://toolradar.com/tools/sentry) | From $26/mo (Team, billed annually); free Developer plan ($0) | Finding out why the application is broken, fastest |

[Prometheus](https://toolradar.com/tools/prometheus) | Free (Apache 2.0); no licence cost, you pay to run it | Teams with platform engineers and a reason to keep data in-house |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Datadog

Top Pick

Best for: Teams that want one platform for everything and can govern their own spend

PricingFrom $15/host/mo (Infrastructure, billed annually); free for 5 hosts

+Genuinely covers infrastructure, APM, logs, RUM and synthetics in one place with consistent tagging

+The integration catalogue means most of your stack reports without custom work

+Splitting log ingest from indexing lets a disciplined team keep everything and index selectively

−Per-host pricing interacts badly with containers and autoscaling, where host counts move independently of value

−The number of separately billed products makes forecasting genuinely hard, and overspend stories are common

Visit Datadog →

2

### Grafana Cloud

Best for: Teams who want open standards and a credible exit

PricingFrom $19/mo plus $6.50 per 1,000 metric series (Pro); free tier

+The free tier is large enough to run a small production system on properly, not just to evaluate

+Built on Prometheus, Loki and Tempo, so your instrumentation and queries are portable if you self-host later

+Usage-based pricing tracks the data you actually send rather than the machines you happen to run

−Assembling metrics, logs and traces into one coherent view takes more setup than an all-in-one

−Metric series pricing means a careless high-cardinality label is still an expensive mistake

Visit Grafana Cloud →

3

### New Relic

Best for: Small teams sending a lot of telemetry

Pricing$0.40/GB ingested after 100 GB free a month; 1 full user free

+100 GB of monthly ingest free is a substantial allowance for a small production estate

+Pricing does not care how many hosts or containers you run, which suits autoscaling well

+One platform across APM, infrastructure and logs without per-product SKUs

−Per-user pricing for full-platform access gets expensive as the number of engineers who need to investigate grows

−The two-axis model, data plus users, takes real modelling before you can predict a bill

Visit New Relic →

4

### Sentry

Best for: Finding out why the application is broken, fastest

PricingFrom $26/mo (Team, billed annually); free Developer plan ($0)

+Groups errors intelligently and shows the stack trace, release and user context that actually resolves a bug

+Team at $26/month with unlimited users is the best value here for a small engineering team

+Session replay turns an unreproducible bug report into something you can watch

−Not infrastructure monitoring; you will still need something for hosts and metrics

−Event quotas are easy to exhaust when a single loop starts erroring, so sampling matters

Visit Sentry →

5

### Prometheus

Best for: Teams with platform engineers and a reason to keep data in-house

PricingFree (Apache 2.0); no licence cost, you pay to run it

+No vendor bill and no per-host or per-GB anxiety, which changes how teams instrument

+The query language and exposition format are the industry standard, so the skills transfer everywhere

+Data stays in your infrastructure, which settles most compliance conversations before they start

−Long-term storage and high availability need extra components, and that is real ongoing engineering work

−Metrics only: you still need something for logs and traces, so it is part of a stack rather than a product

Visit Prometheus →

## What it is

Monitoring is the practice of collecting signals from running systems and alerting when something is wrong.

The signals are conventionally three: metrics (numbers over time, cheap to store, poor at explaining), logs (text events, expensive to store, best at explaining) and traces (the path of one request through your services).

Observability is the broader claim that with enough of those you can answer questions you did not anticipate. In pricing terms only one distinction matters: whether a vendor charges for the machines you run, the data you send, or the people who log in.

## Why it matters

Two costs, and teams usually only budget for one. The obvious one is the invoice, and it is the most common source of surprise spend in an engineering budget because it scales with incidents rather than with headcount.

The less obvious one is alert quality: a platform that makes it easy to alert on everything will, and a team that is paged nightly for non-issues stops reading pages. That failure is more expensive than the subscription, because it is the one that lets a real outage run for an hour. Choose the tool your team will tune, not the one with the most integrations.

## Key features to look for

Billing modelEssential
Per host, per ingested GB, per user, or a mix. This decides your bill far more than feature choice, and the right answer depends on whether you have many machines, much data, or many engineers.

Log ingestion versus indexingEssential
Several vendors charge separately to ingest logs and to make them searchable. Ingest-cheap, index-expensive pricing rewards teams that are disciplined about what they index and punishes everyone else.

Retention and its cost
How long data is queryable, and what longer retention costs. Most incident investigation happens within 48 hours, so paying for a year of hot storage is usually waste.

Cardinality limits
How the platform behaves when you add a high-cardinality label such as user ID. This is the most common cause of a metrics bill exploding without any traffic change.

Alerting and routingEssential
Whether alerts can be expressed as conditions over time rather than single thresholds, and how they reach the on-call engineer. A tool that only does thresholds guarantees noisy pages.

Open standards support
OpenTelemetry ingestion, and whether your instrumentation travels if you leave. This is the difference between switching vendors and re-instrumenting an estate.

## Pricing

Vendors here charge for hosts, for the data you send, for the people who log in, or for a mix of those. Datadog, Grafana Cloud, New Relic and Sentry publish rates. Sentry Enterprise is the only custom quote, and Prometheus is free under the Apache 2.0 licence.

The cheapest credible paid entry is Sentry Team at $26/month billed annually, and Datadog infrastructure starts at $15 per host per month on annual billing. Grafana Cloud Pro is a $19 monthly platform fee plus usage: $6.50 per 1,000 metric series, and logs or traces at $0.55 per GB across processing, writing and retention.

Costs jump on Datadog APM at $31 per host, on log indexing at $1.70 per million events, and on New Relic full-platform users at $99 per user on Standard or $349 on Pro with an annual commitment.

Plan | Price | Best for |

Datadog Free | Free, 5 hosts | Free tier covers 5 hosts |

Datadog Infrastructure | $15/host/mo, annual | On-demand rates run higher than annual |

Datadog APM | $31 per host | APM rate per host |

Datadog Log ingest | $0.10 per ingested GB | Log Management charge to ingest |

Datadog Log indexing | $1.70 per million indexed events | Charge to make logs searchable |

Grafana Cloud Free | Free | 10,000 metric series, 50 GB logs, 50 GB traces, 14-day retention |

Grafana Cloud Pro metrics | $6.50/1,000 metric series | Plus a $19 monthly platform fee; 10,000 series included |

Grafana Cloud Pro logs | $0.55/GB | $0.05 process, $0.40 write, $0.10 retain per GB; 50 GB included |

New Relic Free | Free | 100 GB of ingest per month, one full user |

New Relic Ingest | $0.40/GB | After the free 100 GB; $0.60 on the Data Plus option |

New Relic Full-platform user | $99 to $349/user/mo | Standard: $10 first user, $99 each after; Pro $349 annual |

Sentry Developer | $0 | 1 user, 5,000 errors/month, 5M spans, 50 replays |

Sentry Team | $26/mo, billed annually | Unlimited users and 50,000 errors |

Sentry Business | $80/mo, billed annually | Unlimited dashboards and anomaly detection |

Sentry Enterprise | Custom quote | Enterprise plan at a custom price |

Prometheus | Free | You pay in storage, compute and engineering |

Mistakes to avoid

×Choosing on the dashboard demo. Every vendor demos well on a healthy system. Ask instead what happens to the bill when one service starts logging ten times more, because that is the scenario you will actually meet.

×Adding a high-cardinality label to a metric. User ID, request ID or session ID as a label multiplies your metric series and, on usage-based pricing, your invoice. This is the most common cause of a monitoring bill doubling with no change in traffic.

×Alerting on causes instead of symptoms. Pages for high CPU train the team to ignore pages. Alert on what the user experiences, latency and error rate, and use everything else to investigate once you are awake.

Expert tips

→Sample traces, keep all errors. Full trace collection is expensive and mostly redundant on healthy requests. A low sample rate for successes plus everything for failures gives you the same diagnostic power at a fraction of the cost.

→Set a billing alert on the observability platform itself. The irony is deliberate: the tool that watches everything else is the one whose spend nobody watches.

→Instrument with OpenTelemetry rather than a vendor agent where you can. It costs a little more effort now and turns a future migration into a configuration change instead of a re-instrumentation project.

## The bottom line

For an application team that wants one thing that pays for itself immediately, Sentry at $26/month on Team is the easiest recommendation here, though you will need something alongside it for infrastructure.

For a full platform, the choice is between Datadog and Grafana Cloud, and it comes down to temperament: Datadog if you want everything integrated and will govern the spend, Grafana Cloud if you want open standards, a genuinely usable free tier, and the ability to leave.

New Relic is the value pick when you have a lot of telemetry and few people who need to query it, and the wrong pick when that ratio inverts. Prometheus is free to license and expensive to run, which is a good trade only if you already have the engineers.

Whichever you choose, decide your log retention and your sampling rate on day one, because those two settings determine most of the invoice.

## Frequently asked questions

What is the difference between monitoring and observability?

Monitoring answers questions you knew to ask, through dashboards and alerts you set up in advance. Observability is the claim that with rich enough telemetry you can answer questions you did not anticipate, which usually means traces and high-cardinality data alongside metrics. In practice vendors use the words interchangeably, and the distinction matters less to your bill than how each one charges for data.

Why do monitoring bills grow so unpredictably?

Because they scale with incidents and mistakes rather than with your business. A retry loop that starts erroring, a debug log level shipped to production, or a new metric label containing user IDs can multiply your data volume overnight with no change in traffic. Per-host pricing has the same problem in reverse, where an autoscaling event quietly triples your host count.

Should I self-host Prometheus or pay for a managed platform?

Self-host if you have platform engineers, a compliance reason to keep data in-house, or a scale where vendor pricing has become genuinely painful. Otherwise pay, because the licence is the cheap part: long-term storage, high availability and upgrades are ongoing work, and that engineer time is usually worth more than the subscription you avoided.

How long should I retain logs?

Most incident investigation happens within 48 hours, so hot, queryable retention of one to two weeks covers the overwhelming majority of real use. Anything you keep beyond that is usually for compliance or trend analysis, and both are better served by cheap cold storage than by keeping everything indexed at full price.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Datadog pricing](https://datadoghq.com/pricing)checked Sep 2026
- [Grafana Cloud pricing](https://grafana.com/pricing)checked Sep 2026
- [New Relic pricing](https://newrelic.com/pricing)checked Sep 2026
- [Sentry pricing](https://sentry.io/pricing)checked Sep 2026
- [Prometheus pricing](https://prometheus.io)checked Sep 2026

Related guides

Error Tracking ToolsAi For DevopsPostgres HostingSentry ReviewFree Uptime Monitoring 2026: Free Plan Limits ComparedDatadog Alternatives 2026: Cheaper Monitoring Compared at 10 HostsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Postgres Hosting in 2026

URL: https://devshot.co/reviews/best-postgres-hosting
Type: review
Published: 2026-08-18
Updated: 2026-09-25
Summary: The best managed Postgres hosting in 2026, compared on pricing model, cold starts, branching and what happens when you outgrow the free tier.

Expert Guide

## The Best Postgres Hosting in 2026

Where to run managed PostgreSQL in 2026, from scale-to-zero serverless to a plain always-on instance, compared on what actually shows up on the invoice.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

For a side project or a preview environment, Neon is the default: it scales compute to zero, its free tier gives you 100 CU-hours and 0.5 GB per project, and database branching makes per-pull-request databases trivial. If you want auth, file storage and realtime bundled with the database, Supabase at $25/month on Pro is the fastest path from empty repo to shipped product. If the database is one piece of a service you are already deploying, Railway keeps it in the same project for $5/month on Hobby. For a production database you intend to keep for years, Crunchy Bridge starts at $9/month with backups and connection pooling included, and Amazon RDS remains the answer when compliance and VPC peering decide for you.

Key facts5 tools compared: Neon, Supabase, Railway, Crunchy Bridge, Amazon RDS

- Updated: September 25, 2026
- Top pick: Neon (best for: Preview environments, spiky traffic and anyone who wants a database per pull request)
- Top pick price as of September 25, 2026: Neon: From $0.106/CU-hour (Launch, usage-based); free plan (100 CU-hours, 0.5 GB)

- 5 tools compared: Neon, Supabase, Railway, Crunchy Bridge, Amazon RDS
- Supabase (best for: Shipping a product fast when you need auth, storage and realtime alongside the database): From $25/month (Pro); free plan (500 MB database, 50,000 MAU)
- Railway (best for: Teams already deploying their app on Railway who want the database in the same project): From $5/month (Hobby); free plan with $1 of credits
- Crunchy Bridge (best for: A production database you plan to keep for years, run by people who take Postgres seriously): From $9/month (Hobby-0); storage $0.10 per GB, prorated

Managed Postgres stopped being one product a while ago. The name on the tin is the same, and so is the wire protocol, but a serverless Postgres that suspends after five minutes of silence and a provisioned instance that never sleeps solve genuinely different problems, and they fail in different ways.

Picking on price per gigabyte is how teams end up migrating twice.

The useful question is not which host is fastest. It is which billing model matches your traffic shape, and how much operational work the host takes off your hands.

A database that costs nothing while nobody uses it is a gift for a preview environment and a liability for an API that has to answer in 40 milliseconds at 3am. Below, five hosts that cover the realistic range, with the prices they actually publish.

## Top Picks

Based on features, real-world fit, and value for money.

Best Postgres Hosting in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Neon](https://toolradar.com/tools/neon) | From $0.106/CU-hour (Launch, usage-based); free plan (100 CU-hours, 0.5 GB) | Preview environments, spiky traffic and anyone who wants a database per pull request |

[Supabase](https://toolradar.com/tools/supabase) | From $25/month (Pro); free plan (500 MB database, 50,000 MAU) | Shipping a product fast when you need auth, storage and realtime alongside the database |

[Railway](https://toolradar.com/tools/railway) | From $5/month (Hobby); free plan with $1 of credits | Teams already deploying their app on Railway who want the database in the same project |

Crunchy Bridge | From $9/month (Hobby-0); storage $0.10 per GB, prorated | A production database you plan to keep for years, run by people who take Postgres seriously |

Amazon RDS | Usage-based: per instance-hour plus storage and I/O; 12-month free tier for new accounts | Organisations that need private networking, compliance, or an existing AWS commitment |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Neon

Top Pick

Best for: Preview environments, spiky traffic and anyone who wants a database per pull request

PricingFrom $0.106/CU-hour (Launch, usage-based); free plan (100 CU-hours, 0.5 GB)

+Branching is copy-on-write, so a branch of a 50 GB database is instant and nearly free

+Compute scales to zero, which makes dozens of preview databases cost close to nothing

+Autoscaling up to 16 CU (64 GB RAM) on Launch, so the free tier is not a dead end

−Cold starts on a suspended branch are real, and noticeable on user-facing requests

−Usage-based billing is harder to forecast than a flat instance, especially with many branches

Visit Neon →

2

### Supabase

Best for: Shipping a product fast when you need auth, storage and realtime alongside the database

PricingFrom $25/month (Pro); free plan (500 MB database, 50,000 MAU)

+You get auth, storage and realtime subscriptions without wiring three more services

+It is genuinely plain Postgres underneath, so nothing traps you at the SQL layer

+Row-level security is a first-class citizen rather than something you bolt on

−The bundle is only a bargain if you use the bundle; for a database alone it is not the cheapest

−The free tier pauses inactive projects, which surprises people using it for a staging environment

Visit Supabase →

3

### Railway

Best for: Teams already deploying their app on Railway who want the database in the same project

PricingFrom $5/month (Hobby); free plan with $1 of credits

+The database sits in the same private network as your services, so no public egress and no extra config

+Per-second billing on memory and CPU, so a small database really does cost a few dollars

+Hobby scales up to 48 vCPU and 48 GB per service, which is far more headroom than the price suggests

−Credits-plus-usage pricing takes a month of real traffic before you can predict the bill

−It is a deployment platform first, so the Postgres-specific tooling is thinner than a dedicated host

Visit Railway →

4

### Crunchy Bridge

Best for: A production database you plan to keep for years, run by people who take Postgres seriously

PricingFrom $9/month (Hobby-0); storage $0.10 per GB, prorated

+Backups and connection pooling are included rather than line items you discover later

+Storage is billed by actual use at $0.10/GB, so you are not buying a volume you half fill

+Runs on your choice of cloud, which matters when the rest of your stack already lives somewhere

−No scale to zero, so an idle database still costs the monthly instance price

−No branching, so preview environments need the old dump-and-restore approach

Visit Crunchy Bridge →

5

### Amazon RDS

Best for: Organisations that need private networking, compliance, or an existing AWS commitment

PricingUsage-based: per instance-hour plus storage and I/O; 12-month free tier for new accounts

+Multi-AZ failover, read replicas and point-in-time recovery are mature and well understood

+Sits inside your VPC, which is often a hard requirement rather than a preference

+Reserved instances cut the cost substantially on a workload you know will stay

−The most operational work of the five: parameter groups, subnet groups and upgrade windows are yours

−The bill is genuinely hard to predict, because storage, I/O, backups and egress are all separate lines

Visit Amazon RDS →

## What it is

Managed Postgres hosting means someone else runs the server, the backups, the failover and the version upgrades, and you get a connection string. What separates the options in 2026 is the compute model. Serverless hosts like Neon separate storage from compute and suspend the compute when idle, so you pay for the seconds you query.

Provisioned hosts like Crunchy Bridge and Amazon RDS give you a machine that stays up, which you pay for whether or not anyone connects. Platform hosts like Supabase and Railway wrap Postgres in a broader product, so the database arrives with an API layer, auth, or the rest of your services already next to it.

## Why it matters

The choice shows up in two places you will feel. The first is the bill, and it rarely scales the way you assume: usage-based compute is dramatically cheaper for spiky or low traffic and dramatically worse for a steady always-on workload, because you are paying a premium rate for every second instead of a flat rate for a machine.

The second is cold starts. A suspended serverless database wakes in a few hundred milliseconds, which is invisible in a dashboard and very visible on a checkout page. Neither is a defect. They are trade-offs, and picking the wrong side is the single most common reason teams migrate a database they did not need to migrate.

## Key features to look for

Pricing modelEssential
Usage-based compute billed by the CU-hour or the second, versus a flat monthly instance. Usage-based wins on spiky and low traffic, flat wins the moment the database is genuinely always busy.

Scale to zero and cold startsEssential
Whether idle compute suspends, and how long it takes to wake. Free for preview environments, a latency tax on anything user-facing.

Database branching
Copy-on-write branches of your data, so every pull request gets its own database with production-shaped rows instead of an empty schema and a seed script.

Connection poolingEssential
Postgres handles a few hundred connections badly and serverless functions open thousands. A built-in pooler (PgBouncer or equivalent) is the difference between working and mysteriously not working under load.

Backups and point-in-time recoveryEssential
How far back you can rewind, and whether it is included or billed separately. Check the retention window on the tier you will actually be on, not the enterprise one.

Extensions and version control
Whether you can install pgvector, PostGIS or TimescaleDB, and whether you choose when the major version upgrade happens rather than being told.

## Pricing

Neon, Supabase, Railway and Crunchy Bridge publish prices. Amazon RDS does not: rates vary by instance class and region, billed on-demand per instance-hour plus storage and I/O, with reserved instances for committed workloads, priced in the AWS calculator.

The lowest published paid price is Railway Hobby at $5/month, and Crunchy Bridge Hobby-0 starts at $9/month with backups and pooling included. Neon, Supabase and Railway each have a free plan. Costs jump at Supabase Team ($599/month), when Neon moves from Launch compute to Scale, and when Railway usage runs past the included credits.

Plan | Price | Best for |

Neon Free | $0/month | 100 CU-hours and 0.5 GB per project |

Neon Launch | $0.106/CU-hour, $0.35/GB-month | Usage-based compute, plus storage |

Neon Scale | $0.222/CU-hour, $0.35/GB-month | Higher compute rate, same storage price |

Supabase Free | $0/month | 500 MB database, 1 GB file storage, 50,000 MAU |

Supabase Pro | $25/month, then $0.125/GB | 8 GB disk included, then extra storage |

Supabase Team | $599/month | Team plan above Pro |

Railway Free | $0/month with $1 credits | Free plan with usage credits |

Railway Hobby | $5/month including $5 credits | Hobby plan with included credits |

Railway Pro | $20/month, includes $20 credits | Billed per workspace |

Railway memory | $0.00000386/GB/second | Memory per second after credits |

Railway vCPU | $0.00000772/vCPU/second | CPU per second after credits |

Railway volumes | $0.00000006/GB/second | Volume storage per second after credits |

Crunchy Bridge Hobby-0 | From $9/month | 2 cores, 0.5 GB memory; machine, hosting, backups, connection pooling included |

Crunchy Bridge storage | $0.10 per GB | Billed separately, prorated by use |

Amazon RDS | Per instance-hour | Varies by instance class and region; price it in the AWS calculator |

Mistakes to avoid

×Choosing serverless for a database that is never idle. Scale to zero saves nothing when traffic is constant, and the per-second rate is higher than the equivalent flat instance. Look at your traffic graph before your pricing page.

×Ignoring connection pooling until production. Postgres degrades badly past a few hundred connections, and serverless functions open one per invocation. Every host on this list offers a pooler; use the pooled connection string from day one rather than the direct one.

×Treating the free tier as a staging environment. Several free tiers suspend or pause inactive projects, which means your staging database is asleep exactly when a colleague tries to demo something.

Expert tips

→Price your real workload, not the headline. Take a week of query volume and idle time, then run it through the usage rates above. The answer flips between serverless and provisioned far more often than people expect.

→If you use branching, put a limit on branch lifetime. Copy-on-write is cheap per branch and expensive at forty abandoned branches, and nobody deletes them by hand.

→Check which extensions you need before you commit. pgvector, PostGIS and TimescaleDB availability varies by host and by tier, and discovering the gap after migrating is an expensive afternoon.

## The bottom line

There is no best Postgres host, only a best match for your traffic shape. Spiky, bursty or preview-heavy work belongs on Neon, whose branching turns per-pull-request databases from a chore into a default. If the database is the start of a product rather than a component, Supabase at $25/month on Pro saves you weeks of wiring auth and storage.

Railway makes sense precisely when you are already there. And for a database that has to be up, backed up and boring for the next three years, Crunchy Bridge from $9/month or Amazon RDS inside your own VPC are the grown-up answers. Decide on the billing model first; the feature list follows from it.

## Frequently asked questions

Is serverless Postgres actually cheaper?

Only for workloads with real idle time. Serverless bills a premium rate per second of compute and charges nothing while suspended, so a preview environment used two hours a day is dramatically cheaper. An API answering queries around the clock is usually cheaper on a flat instance, because you pay the lower rate continuously instead of the higher rate continuously.

What is database branching and do I need it?

Branching creates a copy-on-write clone of your database, so a branch of a large database is near-instant and costs only the changed pages. It is genuinely useful if you run preview environments per pull request and want production-shaped data in them. If your team ships from one staging database, it is a feature you will not use.

How bad are cold starts on a suspended database?

On a modern serverless Postgres a suspended compute resumes in a few hundred milliseconds. That is invisible on an internal dashboard or a cron job and clearly visible on a page a customer is waiting on. If a cold start would land on a user-facing request, either keep a minimum compute running or use a provisioned instance.

Can I move between these hosts later?

Yes, and that is the point of choosing Postgres. All five run real PostgreSQL, so pg_dump and pg_restore or logical replication will move you. The friction is never the data, it is the surrounding features: Supabase auth and row-level security policies, or Neon branches, do not travel. Keep the parts you would have to rewrite small if you expect to move.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Neon pricing](https://neon.tech/pricing)checked Sep 2026
- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026

Related guides

Ai For DevopsSelf Hosting ToolsTerraform Iac ToolsFree Database HostingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Serverless Platforms in 2026

URL: https://devshot.co/reviews/best-serverless-platforms
Type: review
Published: 2026-08-27
Updated: 2026-09-25
Summary: The serverless and edge platforms worth deploying to in 2026, compared on their pricing model, their free tiers, and where the bill actually comes from.

Expert Guide

## The Best Serverless Platforms in 2026

The per-seat platforms bill you for your team. The usage-based ones bill you for your traffic. Pick the one whose surprise you can absorb.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

Cloudflare Workers is the cheapest way to run code at the edge, $5 a month minimum with 10 million requests included and no egress charges. Vercel is the best experience for Next.js and bills $20 per user. Netlify now sells credits instead of seats: Pro is one flat monthly fee for the whole team. Fly.io and Render are the picks when you want a long-running process rather than a function, from roughly $2 and $7 a month respectively.

Key facts5 tools compared: Cloudflare Workers, Vercel, Netlify, Fly.io, Render

- Updated: September 25, 2026
- Top pick: Cloudflare Workers (best for: Edge logic, APIs and anything where bandwidth is the risk)
- Top pick price as of September 25, 2026: Cloudflare Workers: From $5/mo minimum per account; free tier (100,000 requests/day)

- 5 tools compared: Cloudflare Workers, Vercel, Netlify, Fly.io, Render
- Vercel (best for: Next.js teams that want the best deployment experience available): From $20 per user/mo (Pro); Hobby free forever; Enterprise custom
- Netlify (best for: Front ends that are not tied to one framework): From $9/mo (Personal); Pro from $20/mo with unlimited members; free plan
- Fly.io (best for: Long-running processes close to users, without Kubernetes): From roughly $1.94-$2.02/mo (shared-cpu-1x, 256MB)

Serverless stopped meaning one thing years ago. It now covers edge functions that run in milliseconds, full application platforms that deploy a framework, and container hosts that keep a process alive. They compete for the same deployments and bill in completely different shapes.

That billing shape is the decision.

Per-seat platforms scale with your team, usage-based platforms scale with your traffic, and the wrong choice is only visible once one of those grows.

## Top Picks

Based on features, real-world fit, and value for money.

Best Serverless Platforms in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Cloudflare Workers | From $5/mo minimum per account; free tier (100,000 requests/day) | Edge logic, APIs and anything where bandwidth is the risk |

[Vercel](https://toolradar.com/tools/vercel) | From $20 per user/mo (Pro); Hobby free forever; Enterprise custom | Next.js teams that want the best deployment experience available |

[Netlify](https://toolradar.com/tools/netlify) | From $9/mo (Personal); Pro from $20/mo with unlimited members; free plan | Front ends that are not tied to one framework |

Fly.io | From roughly $1.94-$2.02/mo (shared-cpu-1x, 256MB) | Long-running processes close to users, without Kubernetes |

[Render](https://toolradar.com/tools/render) | From about $7/mo per instance; free for static sites and small services | Teams that want Heroku's simplicity with current pricing |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Cloudflare Workers

Top Pick

Best for: Edge logic, APIs and anything where bandwidth is the risk

PricingFrom $5/mo minimum per account; free tier (100,000 requests/day)

+No egress charges at all, which removes the least predictable line on the bill

+$5 a month covers 10 million requests

+Runs at the edge by default, so latency is a non-issue

−CPU time per invocation is capped, so heavy compute does not fit

−Its own runtime rather than plain Node, so some libraries do not work

Visit Cloudflare Workers →

2

### Vercel

Best for: Next.js teams that want the best deployment experience available

PricingFrom $20 per user/mo (Pro); Hobby free forever; Enterprise custom

+Best-in-class developer experience, particularly with Next.js

+Preview deployments per pull request that reviewers actually use

+Generous personal tier for side projects

−Per-user pricing grows with headcount rather than usage

−Usage charges sit on top of the seat price

Visit Vercel →

3

### Netlify

Best for: Front ends that are not tied to one framework

PricingFrom $9/mo (Personal); Pro from $20/mo with unlimited members; free plan

+Framework-agnostic, so it does not push you toward one stack

+Mature build plugin ecosystem

+Straightforward for static and hybrid sites

−Credits cover bandwidth, compute and deploys, so a traffic spike drains the monthly pool

−The free plan stops at 300 credits, and a zero balance pauses the team's projects

Visit Netlify →

4

### Fly.io

Best for: Long-running processes close to users, without Kubernetes

PricingFrom roughly $1.94-$2.02/mo (shared-cpu-1x, 256MB)

+Genuinely cheap for small always-on workloads, from roughly $2 a month

+Real containers, so no runtime compatibility surprises

+Multi-region deployment without an orchestration layer

−Pay-per-resource means the bill takes attention rather than a plan choice

−More operational surface than a push-to-deploy platform

Visit Fly.io →

5

### Render

Best for: Teams that want Heroku's simplicity with current pricing

PricingFrom about $7/mo per instance; free for static sites and small services

+Simplest path from repository to running service

+Managed Postgres, cron and background workers included in the model

+Predictable per-instance pricing

−Free instances sleep, which surprises people the first time

−Fewer regions than the edge platforms

Visit Render →

## What it is

A serverless platform runs your code without you provisioning a server.

In practice there are three families: edge runtimes that execute small functions close to the user, application platforms that build and host a framework with previews and CDN included, and managed container hosts that run an ordinary long-lived process without you touching Kubernetes.

The families overlap enough that a single project could ship on any of them, which is why the pricing model tends to decide it rather than the technology.

## Why it matters

Egress is where serverless bills go wrong. Cloudflare charges nothing for bandwidth on Workers, which removes the single most unpredictable line from the invoice. Platforms that meter bandwidth can hand you a bill shaped by one popular post.

The second trap is per-seat pricing on a large team.

Vercel Pro at $20 per developer seat per month is perfectly reasonable for five people and an argument at fifty, regardless of how much traffic you serve. Netlify moved the other way: its credit-based Pro plan includes unlimited members, so headcount no longer moves that bill.

## Key features to look for

Pricing model
Per-seat, per-request, or per-instance. This decides whether your bill tracks your team, your traffic or your uptime, and it is the only choice here that is hard to reverse later.

Egress charges
Whether bandwidth is metered. Cloudflare charges none on Workers; most others meter it, and it is the line most likely to produce a bill nobody predicted.

Cold starts and runtime limits
How long a function may run and how quickly it wakes. Edge runtimes start almost instantly but cap CPU time; container hosts have no cap but keep a process alive and charge for it.

Long-running processes
Whether the platform can host a websocket server or a background worker at all. Edge functions cannot; Fly and Render are built for exactly this.

Preview deployments
A URL per pull request. Standard on the application platforms and the main reason teams accept per-seat pricing.

Free tier shape
What you can run at zero. Cloudflare gives 100,000 requests a day, Vercel and Netlify give a personal tier, Render covers static sites and small services.

## Pricing

Cloudflare, Vercel, Netlify, Fly.io and Render all publish an entry price, and Cloudflare, Vercel, Netlify and Render publish a free tier. Vercel Enterprise and Netlify Enterprise are the only custom quotes. The cheapest credible paid entry is Fly.io, at roughly $1.94 to $2.02 a month for a shared-cpu-1x with 256MB in some regions.

Cloudflare's paid minimum is $5 a month, and Render's instances start at about $7 a month. Costs jump with headcount on Vercel, where Pro is $20 per developer seat per month with usage on top; Netlify Pro charges one flat monthly fee for unlimited members and meters usage in credits.

The bill also jumps on Cloudflare's per-million overages and on Fly.io support and Kubernetes, which sit apart from the small machine price.

Plan | Price | Best for |

Cloudflare Workers Free | Free | 100,000 requests/day, 10ms CPU per invocation |

Cloudflare Workers Paid | $5/mo minimum per account | 10 million requests, no egress or bandwidth charges |

Cloudflare Workers requests | $0.30 per additional million | After the 10 million requests included |

Cloudflare Workers CPU | $0.02 per additional million | After 30 million CPU-milliseconds included |

Vercel Hobby | Free forever | Hobby plan, free forever |

Vercel Pro | $20 per user per month | Matching usage credit included, usage billed on top |

Vercel Enterprise | Custom quote | Enterprise plan at a custom price |

Netlify Free | Free | 300 usage credits a month, hard limit |

Netlify Pro | From $20/mo | Unlimited members, 3,000 credits a month and up |

Netlify Enterprise | Custom quote | Enterprise plan at a custom price |

Fly.io shared-cpu-1x | Roughly $1.94-$2.02/mo | 256MB machine in some regions |

Fly.io volumes | $0.15/GB/mo | Volume storage, per GB per month |

Fly.io Standard support | $29/mo | Standard support plan |

Fly.io Premium support | $199/mo | Premium support plan |

Fly.io Enterprise support | From $2,500/mo | Enterprise support plan |

Fly.io Kubernetes | $75/mo per cluster | Fly Kubernetes, billed per cluster |

Render free | Free | Static sites and small services |

Render instances | From about $7/mo | Paid instances, per month |

Render Pro | $25/mo + compute | Flat workspace fee with unlimited seats |

Mistakes to avoid

×Picking a per-seat platform for a large team serving little traffic. At Vercel Pro's per-seat price, a twenty-person team pays $400 a month before a single request, which a usage-based platform would bill at a fraction of.

×Ignoring egress until the invoice. Bandwidth is metered almost everywhere and is the line that moves when something of yours gets shared widely. Cloudflare not charging for it is a real structural difference, not a marketing point.

×Putting a long-running process on an edge runtime. Workers cap CPU per invocation by design. A websocket server or a queue consumer belongs on Fly or Render, and finding that out in production is expensive.

Expert tips

→Work out whether your cost driver is people or traffic before comparing anything. That single answer eliminates half this list immediately.

→Run the free tiers in parallel for a week with the same project. Cloudflare's 100,000 requests a day and Render's free services are enough to compare real deploy times rather than documented ones.

→Read the overage rate, not the plan price. Cloudflare's $0.30 per additional million requests is the kind of number that decides your bill at scale, and it sits well below the headline.

## The bottom line

If your code fits an edge runtime, Cloudflare Workers is the most economical serious option available: $5 a month, 10 million requests, and no bandwidth charge at all.

If you ship Next.js and value the workflow more than the invoice, Vercel is worth its $20 per seat, with Netlify as the framework-agnostic alternative whose Pro plan does not charge per member.

For anything that has to stay running, Fly.io is remarkably cheap at roughly $2 a month for a small machine, and Render is the easiest on-ramp if you want managed databases and cron in the same place.

## Frequently asked questions

What does serverless actually cost?

It depends entirely on the model. Cloudflare Workers starts at $5 a month for 10 million requests with no egress charge. Vercel charges $20 per developer seat with usage on top, while Netlify Pro is one flat monthly fee for unlimited members, with usage billed in credits. Fly bills per machine, from roughly $2 a month for a small instance, and Render's instances start around $7. Figures checked 26 to 27 August 2026.

Which serverless platform has the best free tier?

Cloudflare's is the most useful for real work: 100,000 requests a day, every day, with no bandwidth charge. Vercel and Netlify's personal tiers are generous for side projects but not licensed for commercial use in the way teams often assume. Render's free services sleep when idle, which is fine for a demo and not for an API.

Can I run a websocket server on serverless?

Not on an edge function runtime, which caps CPU time per invocation by design. You want a container host: Fly.io and Render both run long-lived processes normally, and Fly's per-machine pricing makes a small always-on service genuinely cheap.

Is per-seat or usage-based pricing better?

Per-seat is predictable and punishes headcount; usage-based is cheap while you are small and unpredictable when something takes off. A five-person team serving heavy traffic usually wins on per-seat; a twenty-person team serving modest traffic almost always loses on it.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Vercel pricing](https://vercel.com/pricing)checked Sep 2026
- [Netlify pricing](https://netlify.com/pricing)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026

Related guides

Ci Cd ToolsMonitoring ToolsPostgres HostingFly Io ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Terraform & IaC Tools in 2026

URL: https://devshot.co/reviews/best-terraform-iac-tools
Type: review
Published: 2026-07-09
Updated: 2026-09-25
Summary: The infrastructure-as-code tools engineers actually run in 2026, from Terraform and its open fork to Pulumi and orchestration platforms, ranked.

Expert Guide

## The Best Terraform & IaC Tools in 2026

Declare your infrastructure in code, preview every change, and stop clicking through cloud consoles. The tools worth standardizing on, ranked.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 6 tools compared

TL;DR

The best infrastructure-as-code tools in 2026 are Terraform as the industry standard with the deepest provider ecosystem, OpenTofu as the fully open-source drop-in fork, Pulumi if you would rather write infrastructure in a real programming language, AWS CDK for teams all-in on AWS, and Terragrunt or Spacelift once you need to run all of it at scale. Your cloud, your language preference, and your feelings about Terraform's license decide most of the choice.

Key facts6 tools compared: Terraform, OpenTofu, Pulumi, AWS CDK, Terragrunt, Spacelift

- Updated: September 25, 2026
- Top pick: Terraform (best for: The default choice with the widest provider support)
- Top pick price as of September 25, 2026: Terraform: Free CLI (BSL license); HCP Terraform from free tier

- 6 tools compared: Terraform, OpenTofu, Pulumi, AWS CDK, Terragrunt, Spacelift
- OpenTofu (best for: Teams that want a truly open Terraform): Free (open source)
- Pulumi (best for: Engineers who prefer real programming languages): Free for 1 user; Essentials from $40/mo with usage credits
- AWS CDK (best for: Teams building only on AWS): Free (open source)

Infrastructure as code replaces clicking through cloud consoles with files you can review, version, and reapply. The market shifted in 2023 when Terraform changed its license, which spawned OpenTofu, a truly open fork now under the Linux Foundation.

Around that core sit alternatives that use real programming languages, cloud-specific frameworks, and platforms that run your IaC in CI with policy and drift detection. We ranked on ecosystem breadth, how safe the plan-and-apply loop feels, and total cost. Here are six worth standardizing on.

## Top Picks

Based on features, real-world fit, and value for money.

Best Terraform & IaC Tools in 2026: 6 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Terraform](https://toolradar.com/tools/terraform) | Free CLI (BSL license); HCP Terraform from free tier | The default choice with the widest provider support |

[OpenTofu](https://toolradar.com/tools/opentofu) | Free (open source) | Teams that want a truly open Terraform |

[Pulumi](https://toolradar.com/tools/pulumi) | Free for 1 user; Essentials from $40/mo with usage credits | Engineers who prefer real programming languages |

[AWS CDK](https://toolradar.com/tools/aws-cdk) | Free (open source) | Teams building only on AWS |

[Terragrunt](https://toolradar.com/tools/terragrunt) | Free (open source) | Keeping large Terraform setups DRY |

[Spacelift](https://toolradar.com/tools/spacelift) | Free (2 users, 1 public worker); Starter+ $20,000/yr; Business and Enterprise by quote | Running IaC in CI with policy and drift detection |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Terraform

Top Pick

Best for: The default choice with the widest provider support

PricingFree CLI (BSL license); HCP Terraform from free tier

+Thousands of providers and modules

+Huge community and easy hiring

+Mature tooling and documentation

−Now under the Business Source License, not fully open

−HCL is less flexible than a real programming language

Visit Terraform →

2

### OpenTofu

Best for: Teams that want a truly open Terraform

PricingFree (open source)

+Fully open source under the MPL

+Compatible with existing Terraform configs and providers

+Backed by the Linux Foundation and major vendors

−Slightly behind Terraform on the newest features

−Smaller mindshare, though growing fast

Visit OpenTofu →

3

### Pulumi

Best for: Engineers who prefer real programming languages

PricingFree for 1 user; Essentials from $40/mo with usage credits

+Use a real language with full IDE support

+Strong for developers who dislike HCL

+Multi-cloud with a solid free individual tier

−Smaller ecosystem than Terraform

−Real code can hide complexity if teams are undisciplined

Visit Pulumi →

4

### AWS CDK

Best for: Teams building only on AWS

PricingFree (open source)

+Excellent, opinionated AWS constructs

+Real languages that synthesize to CloudFormation

+Free and backed by AWS

−AWS only, no multi-cloud

−Inherits CloudFormation's slower deploys and limits

Visit AWS CDK →

5

### Terragrunt

Best for: Keeping large Terraform setups DRY

PricingFree (open source)

+Cuts copy-paste across environments

+Works with both Terraform and OpenTofu

+Free and battle-tested by Gruntwork

−Adds another layer to learn and debug

−Only useful once your setup is genuinely large

Visit Terragrunt →

6

### Spacelift

Best for: Running IaC in CI with policy and drift detection

PricingFree (2 users, 1 public worker); Starter+ $20,000/yr; Business and Enterprise by quote

+Policy-as-code and drift detection built in

+Works across Terraform, OpenTofu, and Pulumi

+Usable free tier for small teams

−Paid tiers jump to enterprise pricing fast

−Overkill for a single small project

Visit Spacelift →

## What it is

Infrastructure as code means defining your servers, networks, databases, and cloud services in text files instead of provisioning them by hand. A tool reads that definition, compares it to what already exists, and makes the minimum set of changes to match.

Because the definition is code, you can commit it to Git, review changes in a pull request, and recreate an entire environment from scratch. Terraform popularized the declarative HCL approach; Pulumi and AWS CDK let you do the same in TypeScript, Python, or Go.

## Why it matters

Manual infrastructure does not scale and does not survive turnover. Someone clicks together a working setup, forgets what they did, and six months later nobody can reproduce it or knows why a security group is open.

IaC makes infrastructure reviewable, repeatable, and auditable, and it lets you preview the exact impact of a change before it touches production.

The risk lives in state and blast radius: a bad apply or a leaked state file can take down or expose real systems, which is why the plan step and remote state locking matter as much as the language.

## Key features to look for

Declarative configurationEssential
You describe the desired end state and the tool computes the changes needed to get there, instead of scripting each step by hand.

Plan and preview before applyEssential
A dry run that shows exactly what will be created, changed, or destroyed, so you catch a mistake before it hits production.

State management and lockingEssential
A record of what the tool has provisioned, stored remotely with locking so two engineers cannot corrupt it with concurrent applies.

Provider and ecosystem breadth
The number of clouds and services the tool can manage. Terraform and OpenTofu lead here with thousands of providers.

Modules and reusability
Package infrastructure into reusable components so you define a pattern once and stamp it out across environments.

Drift detection and remote execution
Detect when someone changed infrastructure by hand and run applies in CI with policy checks, which orchestration platforms add on top.

Mistakes to avoid

×Committing state files or secrets to Git. State often contains plaintext secrets, so it belongs in remote backend storage with encryption, never in the repo.

×Running apply without reading the plan. The plan output is the whole safety mechanism, and skimming it is how a destroy slips through into production.

×Skipping remote state and locking on a team. Two people applying against local state at once corrupts it, and recovering a corrupted state file is a bad afternoon.

Expert tips

→Set up a remote backend with state locking on day one, before more than one person touches the code.

→Pin your provider and module versions. An unpinned upgrade can rewrite resources you never meant to change.

→Structure code into small, reusable modules per environment so a change to staging cannot accidentally hit production.

## The bottom line

If you want the safest default and the easiest hiring, Terraform is still the standard, and OpenTofu is the drop-in choice if the license change bothers you or your legal team.

Developers who would rather write infrastructure in a real language should try Pulumi, and AWS-only shops get the cleanest experience from AWS CDK. Add Terragrunt or Spacelift only once your setup is large enough that repetition and drift are real problems, not before.

## Frequently asked questions

Terraform or OpenTofu, which should I use?

OpenTofu is a drop-in fork that reads your existing Terraform configs and providers, so migration is low-risk. Choose Terraform for the very latest features and the largest ecosystem, and OpenTofu if you want a fully open-source license under the Linux Foundation.

Is Terraform still free?

The CLI is free to use, but since 2023 it is under the Business Source License rather than a fully open one, which restricts competing commercial products. For unrestricted open source, OpenTofu is the alternative.

Pulumi or Terraform for a new project?

Pick Terraform or OpenTofu if you want the biggest ecosystem and a declarative language your whole team can read. Pick Pulumi if your team is strong in TypeScript, Python, or Go and wants loops, functions, and tests in their infrastructure code.

Do I need Terragrunt or Spacelift?

Not at first. Terragrunt earns its place when you are copy-pasting the same config across many environments, and Spacelift when you need policy checks, drift detection, and controlled applies in CI. A single small project does not need either.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Terraform pricing](https://terraform.io/pricing)checked Sep 2026
- [OpenTofu pricing](https://opentofu.org)checked Sep 2026
- [Pulumi pricing](https://pulumi.com/pricing)checked Sep 2026
- [AWS CDK pricing](https://aws.amazon.com/cdk)checked Sep 2026
- [Terragrunt pricing](https://terragrunt.gruntwork.io)checked Sep 2026
- [Spacelift pricing](https://spacelift.io/pricing)checked Sep 2026

Related guides

Self Hosting ToolsDocker Desktop AlternativesAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Text to Speech APIs in 2026

URL: https://devshot.co/reviews/best-text-to-speech-apis
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Eight text-to-speech APIs compared on the dollar cost per 1 million characters, what a free tier actually covers…

Expert Guide

## The Best Text to Speech APIs in 2026

Eight APIs compared on the dollar cost per 1 million characters, what the free tier actually includes, and which ones are built for real-time voice agents.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

The best text-to-speech API for most developers in 2026 is ElevenLabs, from $6/mo for 30,000 credits, or about $0.10 per 1,000 characters (roughly $100 per 1M) pay-as-you-go. OpenAI's tts-1 is the cheapest big-cloud option at $15 per 1M characters, and Google Cloud's Standard and WaveNet voices are $4 per 1M characters with 4 million free every month.

Pick Azure AI Speech when you need a custom brand voice built in, at the same flat rate as OpenAI. Pick Deepgram Aura or Cartesia from $5/mo when latency in a live voice agent matters more than the cheapest character. Amazon Polly matches that floor on Standard voices if the app already lives on AWS.

Key facts8 tools compared: ElevenLabs, OpenAI TTS, Google Cloud Text-to-Speech, Azure AI Speech, Amazon…

- Updated: September 25, 2026
- Top pick: ElevenLabs (best for: Teams that need the most natural-sounding voice and will pay a premium for it)
- Top pick price as of September 25, 2026: ElevenLabs: From $6/mo (Starter, 30K credits); about $0.10/1K chars ($100/1M) API v3, $0.05/1K v3 Conversational.

- 8 tools compared: ElevenLabs, OpenAI TTS, Google Cloud Text-to-Speech, Azure AI Speech, Amazon Polly, Deepgram Aura, Cartesia, Hume Octave
- OpenAI TTS (best for: Teams already on the OpenAI API who want one flat rate with no separate voice contract): tts-1 $15/1M characters; tts-1-hd $30/1M; gpt-4o-mini-tts from $0.60/1M input plus $12/1M output tokens.
- Google Cloud Text-to-Speech (best for: High-volume batch narration where the free tier and the flat rate matter more than the newest voice): Standard/WaveNet $4/1M characters (4M free/mo); Neural2 $16/1M (1M free); Chirp 3: HD $30/1M (1M free).
- Azure AI Speech (best for: Regulated or enterprise teams that need a trained brand voice inside the Azure billing account): Neural $15/1M characters (0.5M free/mo); custom Professional Voice synthesis $24/1M, Neural HD $48/1M.

Every text-to-speech API turns a string into audio. They do not bill the same unit, and that is the whole decision. OpenAI and Google Cloud charge per character. Cartesia and Hume charge in credits that convert to streaming minutes.

ElevenLabs charges credits too, but a character can cost half a credit or a full one depending on the model.

A comparison that lines a per-character rate up against a per-credit rate as if they were the same kind of dollar is how a team picks the wrong API for a million-user app.

Toolradar data: the [directory homepage](https://toolradar.com/) lists 10,796+ tools across 401+ categories, checked 25 September 2026.

Its [free AI voice guide](https://toolradar.com/best/free/ai-voice) separately evaluated 15 AI voice tools to build that ranking, checked 25 September 2026, so open [ElevenLabs](https://toolradar.com/tools/elevenlabs), [Deepgram](https://toolradar.com/tools/deepgram), [Cartesia](https://toolradar.com/tools/cartesia), and [Hume](https://toolradar.com/tools/hume-ai) there for the wider AI voice category and use the ladders below for the API rate.

How we ranked: eight APIs, each read on the vendor's own pricing page on 25 September 2026.

The order follows the published dollar rate for a production workload, what a character or credit includes, and whether the fast path to a working voice agent needs a sales call or a card.

A voice cloning app or a dubbing tool is a different buy than a raw API.

That comparison lives in the [free text-to-speech tools guide](https://toolradar.com/guides/best-free-text-to-speech-tools) on Toolradar.

## Top Picks

Based on features, real-world fit, and value for money.

Best Text to Speech APIs in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[ElevenLabs](https://toolradar.com/tools/elevenlabs) | From $6/mo (Starter, 30K credits); about $0.10/1K chars ($100/1M) API v3, $0.05/1K v3 Conversational. | Teams that need the most natural-sounding voice and will pay a premium for it |

OpenAI TTS | tts-1 $15/1M characters; tts-1-hd $30/1M; gpt-4o-mini-tts from $0.60/1M input plus $12/1M output tokens. | Teams already on the OpenAI API who want one flat rate with no separate voice contract |

Google Cloud Text-to-Speech | Standard/WaveNet $4/1M characters (4M free/mo); Neural2 $16/1M (1M free); Chirp 3: HD $30/1M (1M free). | High-volume batch narration where the free tier and the flat rate matter more than the newest voice |

Azure AI Speech | Neural $15/1M characters (0.5M free/mo); custom Professional Voice synthesis $24/1M, Neural HD $48/1M. | Regulated or enterprise teams that need a trained brand voice inside the Azure billing account |

Amazon Polly | Standard $4/1M chars (5M free/mo); Neural $16/1M (1M free, 12mo); Generative $0.03/1K, Long-Form $100/1M. | AWS-native apps that want the cheapest floor rate and can live with older-sounding Standard voices |

Deepgram Aura | Aura-2 pay-as-you-go $0.030/1K chars; Growth plan $0.027/1K from $4,000+/year; $200 free credit. | Voice agents that need low, predictable latency and already use Deepgram for speech-to-text |

Cartesia | Free $0/mo (~27 TTS min); Pro $5/mo (~133 min); Startup $49/mo (~1,667 min); Scale $299/mo (~10,667 min). | Voice agents built around Cartesia's Sonic model, where sub-second latency is the buying reason |

Hume Octave | Free $0/mo (10K chars); Starter $3/mo (30K chars); Creator $14/mo ($7 first month, 140K chars); Pro $70/mo. | Apps where emotional tone and delivery instructions matter more than the lowest per-character rate |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 4 of 8 do not publish a comparable monthly price and are left out rather than estimated.

1

### ElevenLabs

Top Pick

Best for: Teams that need the most natural-sounding voice and will pay a premium for it

PricingFrom $6/mo (Starter, 30K credits); about $0.10/1K chars ($100/1M) API v3, $0.05/1K v3 Conversational.

+Instant and professional voice cloning, in 29 languages, with commercial rights starting on the Starter plan

+Credit rollover up to three months of quota softens a spiky month

+The v3 Conversational model halves the per-character cost for chat-style output

−The free plan is personal use only, so a production key means a paid tier from day one

−A character can cost 0.5 to 1 credit depending on the model, so the effective per-1M-character rate is not fixed

Visit ElevenLabs →

2

### OpenAI TTS

Best for: Teams already on the OpenAI API who want one flat rate with no separate voice contract

Pricingtts-1 $15/1M characters; tts-1-hd $30/1M; gpt-4o-mini-tts from $0.60/1M input plus $12/1M output tokens.

+tts-1 undercuts Azure's and ElevenLabs' equivalent tiers on a flat per-character basis

+gpt-4o-mini-tts takes a text prompt to steer tone and pacing, not just a voice ID

+One API key covers text, audio, and the rest of OpenAI's models, so there is no separate voice vendor to bill

−No free tier for the audio endpoints, unlike Google Cloud or Amazon Polly

−No voice cloning; the catalog is a fixed set of preset voices

Visit OpenAI TTS →

3

### Google Cloud Text-to-Speech

Best for: High-volume batch narration where the free tier and the flat rate matter more than the newest voice

PricingStandard/WaveNet $4/1M characters (4M free/mo); Neural2 $16/1M (1M free); Chirp 3: HD $30/1M (1M free).

+4 million free Standard or WaveNet characters every month, renewing, not a one-time trial

+Five voice tiers span the cheapest to the priciest rate on this whole list, letting a team pick quality against cost per feature

+Gemini-TTS models add prompt-steered delivery, billed in tokens, for teams already on Vertex AI

−Studio voices at $160 per 1M characters are the most expensive tier on this whole list

−No native voice cloning; Instant Custom Voice ($60 per 1M characters) is a narrower feature, not full cloning

Visit Google Cloud Text-to-Speech →

4

### Azure AI Speech

Best for: Regulated or enterprise teams that need a trained brand voice inside the Azure billing account

PricingNeural $15/1M characters (0.5M free/mo); custom Professional Voice synthesis $24/1M, Neural HD $48/1M.

+Neural voices match OpenAI's tts-1 rate, with 100+ languages out of the box

+Professional Voice fine-tuning trains a brand voice for $52 per compute hour, up to $936 per training run

+Prepaid commitment tiers cut the rate to $7.50 per 1M characters at 2,000M characters a month

−Custom voice access requires an application and Microsoft's responsible-use review before you can train one

−Endpoint hosting for a trained voice is a separate running cost, $4.04 per model per hour

Visit Azure AI Speech →

5

### Amazon Polly

Best for: AWS-native apps that want the cheapest floor rate and can live with older-sounding Standard voices

PricingStandard $4/1M chars (5M free/mo); Neural $16/1M (1M free, 12mo); Generative $0.03/1K, Long-Form $100/1M.

+Standard voices are the cheapest floor rate on this list, with 5 million free characters a month, ongoing

+Four voice tiers (Standard, Neural, Generative, Long-Form) map cleanly to a quality-versus-cost decision

+Bills inside an existing AWS account, so no second vendor relationship for a team already there

−Neural, Generative, and Long-Form's free allowances expire after the first 12 months on the account

−Long-Form voices are the second most expensive tier on this list, after Google Studio

Visit Amazon Polly →

6

### Deepgram Aura

Best for: Voice agents that need low, predictable latency and already use Deepgram for speech-to-text

PricingAura-2 pay-as-you-go $0.030/1K chars; Growth plan $0.027/1K from $4,000+/year; $200 free credit.

+A single flat rate per model (Aura-1 or Aura-2), no credit conversion or model-dependent multiplier

+$200 in free credit for new pay-as-you-go accounts covers a real pilot before a card is required

+Pairs naturally with Deepgram's speech-to-text API for a full voice-agent pipeline on one bill

−Aura-2 costs twice OpenAI's tts-1 rate for the newer, better-rated voice model

−The discounted Growth rate needs a $4,000-plus annual prepaid commitment, not a monthly toggle

Visit Deepgram Aura →

7

### Cartesia

Best for: Voice agents built around Cartesia's Sonic model, where sub-second latency is the buying reason

PricingFree $0/mo (~27 TTS min); Pro $5/mo (~133 min); Startup $49/mo (~1,667 min); Scale $299/mo (~10,667 min).

+A usable free tier (about 27 minutes a month) to prototype an agent before paying anything

+Pro is the cheapest paid entry point on this list for a low-volume live agent

+Instant voice cloning is included from the Pro tier, not gated behind a higher plan

−Pricing is in minutes and credits, so it does not translate cleanly to a per-1M-character forecast for a batch job

−The free and Pro tiers cap concurrent requests at 2 and 3, a real limit for a multi-caller phone line

Visit Cartesia →

8

### Hume Octave

Best for: Apps where emotional tone and delivery instructions matter more than the lowest per-character rate

PricingFree $0/mo (10K chars); Starter $3/mo (30K chars); Creator $14/mo ($7 first month, 140K chars); Pro $70/mo.

+Octave takes acting instructions in plain language ('sound relieved but tired'), not just a voice ID

+A genuine free tier at 10,000 characters a month to test delivery before paying

+Overage rate drops from $0.15 to $0.05 per 1,000 characters as plans scale to Business

−Creator's listed $7/mo is a first-month price; the page states the regular rate is $14/mo

−Character allowances are smaller per dollar than Google Cloud's or Amazon Polly's flat tiers at the low end

Visit Hume Octave →

## What it is

A text-to-speech API takes a string, a voice ID, and some settings, and returns audio, usually in a few hundred milliseconds for a short clip.

The bill almost always follows the input text, not the audio length, which is why a script full of SSML tags and punctuation costs more than the same words written plainly.

The category splits by billing shape.

Per-character APIs (OpenAI, Google Cloud, Azure, Amazon Polly) post a flat rate per 1M characters and are the easiest to forecast.

Credit-based APIs (ElevenLabs, Cartesia, Hume) sell a monthly bundle where a character or a minute consumes a variable number of credits depending on the voice model, so the effective per-character cost moves with the model you pick.

Deepgram sits in between: a flat rate per 1,000 characters with a discounted annual commitment. A voice agent built on any of these usually sits behind an API gateway and a backend that logs every call for billing.

## Why it matters

A 10 million character month, roughly a long audiobook catalog or a year of a busy voice agent, is a different invoice on every API. On Google Cloud's Standard voices it is $40, most of it inside the 4 million free characters.

On OpenAI's tts-1 it is $150. On Azure's custom Professional Voice it is $240.

On ElevenLabs' pay-as-you-go v3 rate it is close to $1,000, the price of the voice most consumer apps actually ship because it is the one that does not sound synthetic.

Devshot data: 28,000 developer readers get this brief, and the audience file updated 20 September 2026 puts the open rate at 30%.

They forecast an API the way they forecast a cloud bill, so a credit system that quietly halves or doubles the per-character cost by model is the month-two surprise. Later pricing changes land in the Devshot brief for the next check.

Latency is the other axis.

A chatbot reading a paragraph can tolerate a full round trip. A phone agent or a live game NPC cannot, and that is a separate buy from a batch narration job.

The monitoring tools guide, the Datadog review, and the error tracking tools guide cover the dashboards that catch a voice API's slow p99 or a failed synthesis call before a user does.

## Key features to look for

The billing unitEssential
Characters, tokens, or credits that convert to minutes. Three APIs on this list bill per character, two bill in credits with a model-dependent rate, and OpenAI's newer audio models bill in tokens. That unit decides whether a quote is comparable at all.

What the free tier actually coversEssential
Google Cloud gives 4 million Standard characters free every month, ongoing. Amazon Polly's neural free tier runs out after 12 months. Not every free plan clears a production key without an upgrade.

Latency and streamingEssential
Cartesia and Deepgram are built for sub-second, streamed audio for live voice agents. OpenAI and Google Cloud's batch-oriented voices are fine for narration and slower for a live phone call.

Voice cloning and custom voices
ElevenLabs and Azure sell a trained brand voice, at very different prices and approval processes. Amazon Polly and Deepgram ship a fixed voice catalog with no cloning.

Enterprise commitment tiers
Azure and Deepgram both cut their per-unit rate for a prepaid annual commitment, a different purchase from the pay-as-you-go card most teams start on.

## Pricing

Compare the flat-rate APIs first, because they convert cleanly to a per-1M-character number. Google Cloud's Standard and WaveNet voices are the cheapest ongoing free tier here.

OpenAI's tts-1 and Azure's Neural voices land on the same flat rate, and tts-1-hd doubles it for higher fidelity.

Amazon Polly's Standard tier matches Google Cloud's floor, but its Neural and Long-Form tiers climb well past it, the second-most expensive rate on this page after Google's Studio voices.

Deepgram's Aura-2 pay-as-you-go rate drops further on a Growth plan that needs a $4,000-plus annual prepay.

The credit-based APIs do not convert as cleanly. ElevenLabs' API rate works out to roughly ten times OpenAI's tts-1 rate on the v3 model, or half that on v3 Conversational, while its monthly plans run $6 to $990.

Cartesia and Hume bill in minutes and characters respectively inside a monthly credit bundle: Cartesia from $5/mo for about 133 TTS minutes, Hume from $3/mo for 30,000 characters, with Hume's Creator tier at $14/mo regular ($7 the first month).

Azure's custom Professional Voice adds a separate synthesis rate plus $52 per compute hour to train the model, verified on each vendor's pricing page, September 2026.

Plan | Price | Best for |

ElevenLabs Free | $0/mo | 10,000 credits/month, personal use only |

ElevenLabs Starter | $6/mo | 30,000 credits, commercial use rights |

ElevenLabs Creator | $22/mo | Regular rate; $11 the first month, 121,000 credits |

ElevenLabs Pro | $99/mo | About 600,000 credits |

ElevenLabs Scale | $299/mo | About 1,800,000 credits |

ElevenLabs API pay-as-you-go, v3 | $0.10/1K characters | About $100 per 1M characters |

ElevenLabs API pay-as-you-go, v3 Conversational | $0.05/1K characters | About $50 per 1M characters |

OpenAI tts-1 | $15/1M characters | No free tier for audio endpoints |

OpenAI tts-1-hd | $30/1M characters | Higher-fidelity model, same billing unit |

OpenAI gpt-4o-mini-tts | $0.60/1M input, $12/1M output tokens | Token-billed, takes a delivery instruction prompt |

Google Cloud Standard / WaveNet | $4/1M characters | 4 million characters free every month |

Google Cloud Neural2 / Polyglot | $16/1M characters | 1 million characters free every month |

Google Cloud Chirp 3: HD | $30/1M characters | 1 million characters free every month |

Google Cloud Instant Custom Voice | $60/1M characters | No free tier |

Google Cloud Studio | $160/1M characters | 1 million characters free every month |

Azure Neural / Neural HD Flash | $15/1M characters | 0.5 million characters free every month |

Azure custom Professional Voice synthesis | $24/1M characters | Requires application and responsible-use review |

Azure custom Professional Voice, Neural HD | $48/1M characters | Higher-fidelity custom voice output |

Azure Professional Voice training | $52/compute hour | Up to $936 per training run |

Azure Neural 1 commitment, 80M characters | $12/1M characters | $960/mo prepaid commitment tier |

Azure Neural 1 commitment, 2,000M characters | $7.50/1M characters | $15,000/mo prepaid commitment tier |

Amazon Polly Standard | $4/1M characters | 5 million characters free every month |

Amazon Polly Neural | $16/1M characters | 1 million free/month, first 12 months only |

Amazon Polly Generative | $30/1M characters | 100,000 free/month, first 12 months only |

Amazon Polly Long-Form | $100/1M characters | 500,000 free/month, first 12 months only |

Deepgram Aura-2 pay-as-you-go | $0.030/1K characters | Flat per-character rate, no credit conversion |

Deepgram Aura-2 Growth | $0.027/1K characters | Discounted rate; needs $4,000+/year prepaid |

Cartesia Free | $0/mo | About 27 TTS minutes, 2 concurrent requests |

Cartesia Pro | $5/mo | About 133 TTS minutes; instant voice cloning |

Cartesia Startup | $49/mo | About 1,667 TTS minutes; professional voice cloning |

Cartesia Scale | $299/mo | About 10,667 TTS minutes, 15 concurrent requests |

Hume Octave Free | $0/mo | 10,000 characters (about 10 minutes) |

Hume Octave Starter | $3/mo | 30,000 characters; page lists this as the current rate |

Hume Octave Creator | $14/mo | Regular rate; $7 the first month, 140,000 characters |

Hume Octave Pro | $70/mo | 1,000,000 characters, $0.12/1,000 overage |

Hume Octave Business | $500/mo | 10,000,000 characters, $0.05/1,000 overage |

Mistakes to avoid

×Comparing ElevenLabs' monthly plan price against OpenAI's per-character rate as if they were the same kind of number. Convert ElevenLabs to its published $0.10 per 1,000 character API rate before putting it next to a flat-rate API.

×Budgeting Amazon Polly's Neural, Generative, or Long-Form free tier past 12 months. Only Standard voices keep an ongoing free allowance; the others drop to the paid rate on the account's first anniversary.

×Picking the cheapest per-character rate for a live phone agent. Deepgram and Cartesia are built and priced for streaming latency; a batch-oriented API at a lower headline rate can add hundreds of milliseconds a call needs.

Expert tips

→Run the same 60-second script through two or three APIs before committing. Character count and voice model both move the real bill, and a demo script rarely matches production text.

→If the app trains or hosts a custom voice, price Azure's training ($52/compute hour, up to $936 per training run) and hosting ($4.04/model/hour) separately from the per-character synthesis rate. They land on the same invoice but are billed differently.

→For a voice agent, pull the latency and uptime numbers into whatever already watches the rest of the stack; the <a href="/reviews/best-api-testing-tools">API testing tools guide</a> covers catching a slow TTS call before a caller hangs up. A batch narration job that runs on demand fits a <a href="/reviews/best-serverless-platforms">serverless platform</a> better than a server kept running for occasional requests.

## The bottom line

Default to ElevenLabs when voice quality is the product, and budget its API rate well above its entry plan's sticker once converted to pay-as-you-go.

Default to OpenAI's tts-1 when the team is already on the OpenAI API and wants one flat, forecastable bill.

Use Google Cloud for high-volume batch narration; its free tier and flat rate beat every other API here on cost.

Use Azure when a trained brand voice has to clear an enterprise procurement review, and Amazon Polly when the app already lives on AWS and Standard voices are good enough.

Use Deepgram Aura or Cartesia for a live voice agent where latency, not the lowest per-character rate, is the buying reason.

Use Hume Octave when the product needs prompted emotional delivery and can live with a narrower voice catalog. Cite this: Devshot, "Best Text to Speech APIs in 2026", September 2026.

## Frequently asked questions

What is the best text-to-speech API for developers in 2026?

ElevenLabs is the default for voice quality, from $6/mo for 30,000 credits or about $0.10 per 1,000 characters on the API, checked on ElevenLabs' pricing page in September 2026. OpenAI's tts-1 is the cheapest flat-rate alternative among the big clouds. Google Cloud's Standard and WaveNet voices are the cheapest overall, with 4 million free characters every month.

How much does a text-to-speech API cost per 1 million characters?

On the pages checked in September 2026: Google Cloud Standard and Amazon Polly Standard share the cheapest per-character rate on this list. OpenAI's tts-1 and Azure's Neural voices land at the same mid-tier rate, Deepgram's Aura-2 sits a step above that, and Google Cloud Studio voices are the most expensive at $160 per 1M. ElevenLabs' credit-based API works out to roughly $50 per 1M characters on the fastest model, and about double that on the standard v3 model.

Is there a free text-to-speech API?

Google Cloud gives 4 million Standard or WaveNet characters free every month, ongoing, the largest recurring free tier here. Amazon Polly's Standard tier gives 5 million free characters a month, also ongoing, but its Neural, Generative, and Long-Form free tiers expire after 12 months. Cartesia and Hume both have a genuine free plan with no card required; ElevenLabs' free plan is capped at 10,000 credits.

ElevenLabs vs OpenAI TTS: which should a developer use?

ElevenLabs when voice quality and cloning matter more than cost: it is the voice most consumer apps ship, at roughly ten times OpenAI's flat per-character rate on the v3 API. OpenAI's tts-1 when one flat rate and one API key for text and audio matter more than the single best-sounding voice. Neither offers a production-ready free tier: ElevenLabs' free credits do not carry commercial rights, and OpenAI's audio endpoints have no free allowance at all.

Which text-to-speech API is best for real-time voice agents?

Deepgram Aura and Cartesia are both built for streamed, low-latency output rather than batch narration. Deepgram bills a flat $0.030 per 1,000 characters on Aura-2 pay-as-you-go, with a discounted Growth plan starting at a $4,000-plus annual prepay. Cartesia bills in minutes instead of characters, from a free tier (about 27 minutes) through $299/mo (about 10,667 minutes), and includes instant voice cloning from its $5/mo Pro plan.

Can I clone a voice with these APIs?

ElevenLabs and Cartesia both offer instant voice cloning on their entry paid plans, and ElevenLabs adds professional cloning on higher tiers. Azure sells a full custom voice pipeline through Professional Voice fine-tuning, at $24 to $48 per 1M characters for synthesis plus training and hosting fees, but it requires an application and a responsible-use review before access opens. OpenAI, Google Cloud's Instant Custom Voice aside, Amazon Polly, and Deepgram ship fixed voice catalogs with no general cloning feature.

Do these APIs bill by character or by audio minute?

Most of this list bills by input character: OpenAI, Google Cloud, Azure, Amazon Polly, Deepgram, and Hume all post a rate per 1,000 or 1 million characters. Cartesia is the exception, billing in credits that convert to TTS minutes rather than characters, which is why its plans do not translate directly into a per-1M-character number the way the others do.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [ElevenLabs pricing](https://elevenlabs.io/pricing)checked Sep 2026

Related guides

Api Testing ToolsMonitoring ToolsApi GatewaysAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Vibe Coding Tools in 2026

URL: https://devshot.co/reviews/best-vibe-coding-tools
Type: review
Published: 2026-08-22
Updated: 2026-09-25
Summary: The best vibe coding tools in 2026, compared on credit economics, code ownership, where each one hits a wall and what it costs to graduate.

Expert Guide

## The Best Vibe Coding Tools in 2026

Prompt-to-app tools that get you a working prototype in an afternoon, compared on what they cost once you start debugging rather than building.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 6 tools compared

TL;DR

Lovable is the friendliest way from a sentence to a deployed app, and its free tier gives you 5 build credits a day up to 30 a month. Bolt at $25/month is the most generous on raw allowance, starting at 10 million tokens a month with unused tokens rolling over. v0 is the pick if the output has to land in a real Next.js codebase, at $30 per user per month on Plus. Replit on Core is the only one that is a genuine development environment as well as a generator, which matters the moment you need a database and a background job. And Cursor, free on Hobby and $20/month on Pro, is where you go when the vibes run out, which they will. The honest framing: every tool here is excellent at the first 80% and the same wall waits at the end of all of them.

Key facts6 tools compared: Lovable, Bolt, v0, Replit, Cursor, Whacka

- Updated: September 25, 2026
- Top pick: Lovable (best for: Non-developers and founders getting a real, deployed app from a description)
- Top pick price as of September 25, 2026: Lovable: Free: 5 build credits/day (up to 30/month), plus 20 Cloud credits; Pro from $25/mo

- 6 tools compared: Lovable, Bolt, v0, Replit, Cursor, Whacka
- Bolt (best for: Anyone who wants the most build allowance per dollar): From $25/mo (Pro, billed monthly); free ($0) with 300K tokens/day and 1M/month
- v0 (best for: Teams whose output has to drop into an existing Next.js and React codebase): From $30/user/mo (Plus); free ($0/mo) with $5 credits and 7 messages/day
- Replit (best for: Projects that outgrow a prototype and need a real environment around them): From $20/mo (Core), or $18/mo billed annually; free Starter plan

Vibe coding is the practice of describing what you want and shipping whatever the model produces without reading much of it closely.

It works far better than it has any right to, and it fails in a specific, predictable way: the first version arrives in minutes, and then you ask for one change the model did not anticipate and spend two hours in a loop where each fix breaks something else.

That loop is also where the money goes, which is why credit economics matter more than the monthly price.

Every tool on this list bills by credits, tokens or messages, and generating a new feature is cheap while debugging a broken one is expensive. A $25 plan that covers a week of building can evaporate in an afternoon of arguing with the model about why the login button stopped working. Below, what each one is genuinely good at, and where it stops.

## Top Picks

Based on features, real-world fit, and value for money.

Best Vibe Coding Tools in 2026: 6 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Lovable](https://toolradar.com/tools/lovable) | Free: 5 build credits/day (up to 30/month), plus 20 Cloud credits; Pro from $25/mo | Non-developers and founders getting a real, deployed app from a description |

[Bolt](https://toolradar.com/tools/bolt) | From $25/mo (Pro, billed monthly); free ($0) with 300K tokens/day and 1M/month | Anyone who wants the most build allowance per dollar |

[v0](https://toolradar.com/tools/v0) | From $30/user/mo (Plus); free ($0/mo) with $5 credits and 7 messages/day | Teams whose output has to drop into an existing Next.js and React codebase |

[Replit](https://toolradar.com/tools/replit) | From $20/mo (Core), or $18/mo billed annually; free Starter plan | Projects that outgrow a prototype and need a real environment around them |

[Cursor](https://toolradar.com/tools/cursor) | From $20/mo (Pro); free Hobby plan, no card required | The moment you need to read and change the code yourself |

Whacka | Freemium; free plan available | Non-coders shipping a working app fast |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Lovable

Top Pick

Best for: Non-developers and founders getting a real, deployed app from a description

PricingFree: 5 build credits/day (up to 30/month), plus 20 Cloud credits; Pro from $25/mo

+Genuinely good at going from one paragraph to a deployed, working application

+Handles database and auth provisioning rather than leaving you with a front end only

+The free daily credit grant is enough to evaluate it properly before paying

−Paid pricing is not published in a readable form, which is a poor look in a category where everyone else lists numbers

−Credit consumption during debugging is where the free tier runs out fastest

Visit Lovable →

2

### Bolt

Best for: Anyone who wants the most build allowance per dollar

PricingFrom $25/mo (Pro, billed monthly); free ($0) with 300K tokens/day and 1M/month

+10M tokens a month on Pro at $25 is the best raw allowance here, and unused tokens roll over

+The free tier is unusually usable: 1M tokens a month, hosting and unlimited databases

+Runs a real toolchain in the browser, so the output is conventional code rather than an abstraction

−Token-based metering is harder to reason about than message counts until you have a month of history

−Free tier puts Bolt branding on your sites, so it is not a route to a client deliverable

Visit Bolt →

3

### v0

Best for: Teams whose output has to drop into an existing Next.js and React codebase

PricingFrom $30/user/mo (Plus); free ($0/mo) with $5 credits and 7 messages/day

+The output is conventional, readable React and Tailwind rather than a proprietary structure

+Slots straight into an existing Vercel and Next.js workflow with no translation step

+Business tier opts out of training by default, which matters for client work

−The 7 message per day cap on free is restrictive, and $5 of credits does not go far

−Strongest on interfaces; you will still assemble the backend yourself

Visit v0 →

4

### Replit

Best for: Projects that outgrow a prototype and need a real environment around them

PricingFrom $20/mo (Core), or $18/mo billed annually; free Starter plan

+A real terminal, filesystem and database, which means you can fix by hand what the agent cannot

+Database rollbacks up to 28 days on Pro, genuinely valuable when an agent damages your schema

+Parallel agents on Core and Pro, so a long task does not block a short one

−Credits are denominated in dollars, so heavy agent use can quietly exceed the plan value

−More concepts to learn than a pure prompt-to-app tool, which is the point but also a cost

Visit Replit →

5

### Cursor

Best for: The moment you need to read and change the code yourself

PricingFrom $20/mo (Pro); free Hobby plan, no card required

+Works on your own repository, so nothing is trapped in a hosted environment

+The agent is strong enough to keep most of the speed benefit while you retain control

+A free Hobby tier means the graduation step costs nothing to try

−It is an editor, so it assumes you can read code; it will not host or deploy for you

−Overlaps with the broader AI coding assistant category, where the competition is fiercer

Visit Cursor →

6

### Whacka

Best for: Non-coders shipping a working app fast

PricingFreemium; free plan available

+No code required

+Web and mobile

−Not for complex production systems

Visit Whacka →

## What it is

These tools sit between a chatbot and an IDE. You describe an application in prose, the model writes the code, and the platform runs it, hosts it and shows you the result in a browser. The category splits on what you get at the end.

Some are closed generators: they produce a working app inside their own environment and getting the code out is an afterthought. Others are code-first: they emit a real repository in a conventional framework, and the generator is a starting point rather than the destination.

That distinction decides whether you own an asset or rent a prototype, and it is almost never on the pricing page.

## Why it matters

The value is real and so is the ceiling. For a landing page, an internal tool, a prototype to put in front of a customer on Thursday, these tools collapse a week into an hour, and that is not hype. What they do not do is remove the need to understand the code, they defer it.

The moment something breaks in a way the model cannot see, you are debugging software you did not write and do not understand, which is strictly harder than debugging your own. Treat the output as a first draft by a fast junior developer: enormously useful, and not something you deploy to paying customers without reading. The tools that let you take the code with you are the ones that survive that realisation.

## Key features to look for

Credit and token economicsEssential
How usage is metered, and whether unused allowance rolls over. Debugging consumes far more than building, so the allowance that matters is the one on a bad day, not a good one.

Code export and ownershipEssential
Whether you can pull a real repository out and run it elsewhere. This is the difference between a prototype you rent and a codebase you own, and it decides how much of the work survives.

Framework and output quality
What the tool actually emits. Conventional Next.js or React that a developer can read beats a proprietary abstraction, however elegant the preview looks.

Backend, database and authEssential
Whether the tool can provision persistence and authentication or only generates a front end. Most real requests need a database within the first hour.

Deployment and custom domains
Hosting included, and whether you can point your own domain at it. Fine for a prototype, worth checking before it becomes production.

Debugging affordancesEssential
Access to logs, the ability to edit files directly, and whether you can drop into a terminal. The escape hatch you will need on day two.

## Pricing

Lovable, Bolt, v0, Replit and Cursor publish monthly prices, and every tool here has a free or freemium entry. Whacka is listed only as freemium, and Enterprise on Bolt, v0, Replit and Cursor is a custom quote. Cursor lists Pro+ at $60 and Ultra at $200 a month.

The cheapest published paid price is Replit Core at $18/month billed annually, with Replit Core monthly and Cursor Pro next, while Lovable Pro and Bolt Pro are $25/month billed monthly. Costs jump to $30 per user per month on v0 Plus, $40 per user per month on Cursor Teams, and $100 a month on v0 Business and Replit Pro.

Plan | Price | Best for |

Lovable Free | Free, plus 20 Cloud credits/mo | 5 build credits a day, up to 30 a month |

Lovable Pro | $25/mo | 100 monthly credits on top of the daily grants, with rollover and custom domains |

Lovable Business | $50/mo | Adds a team workspace, role-based access and internal-only publishing |

Bolt Free | Free ($0) | 300K tokens/day, 1M/month, 10MB uploads, hosting, unlimited databases |

Bolt Pro | $25/mo, billed monthly | Starts at 10M tokens a month, unused tokens roll over |

Bolt Teams | $30 per month and member | Teams plan |

Bolt Enterprise | Custom quote | Enterprise plan |

v0 Free | Free ($0/mo) | $5 of included monthly credits, 7 messages a day |

v0 Plus | $30 per user/month | $30 of included credits, plus $2 free daily on login |

v0 Business | $100 per user/month | Same credit grant, training opt-out by default |

v0 Enterprise | Custom quote | Enterprise plan |

Replit Starter | Free | Daily Agent credits, built-in database, 1 published project |

Replit Core | $20/mo, or $18/mo billed annually | Monthly credit toward premium models, Plan mode and unlimited workspaces |

Replit Pro | $100/mo, or $90/mo billed annually | $100 of credits, 10 parallel agents, 28-day database rollbacks |

Replit Enterprise | Custom quote | Enterprise plan |

Cursor Hobby | Free | Limited Agent requests, no card required |

Cursor Pro | $20/mo | Individual Pro plan |

Cursor Pro+ | $60/mo | Cursor's recommended tier for daily agent users |

Cursor Ultra | $200/mo | Top individual tier for agent power users |

Cursor Teams | From $40 per user/month | Teams plan starts at this rate |

Cursor Enterprise | Custom quote | Enterprise plan |

Whacka | Freemium | Free plan available |

Mistakes to avoid

×Judging a plan by the build phase. Generating features is cheap; debugging is expensive, because each failed attempt spends allowance and adds context. Look at what a bad afternoon costs, which is why Bolt's token rollover and Replit's dollar-denominated credits are worth understanding before you subscribe.

×Shipping generated code to paying customers without reading it. The output is a competent first draft, not reviewed software. Authentication, authorisation and anything touching payments deserve a human pass, and that is the one corner where the time saved is not real.

×Choosing a tool you cannot export from. If the code cannot leave, then every hour of prompting builds an asset you do not own. Check the export path on day one, not in month three when you have something worth keeping.

Expert tips

→Commit early and often, even if you are not reading the code. The single most useful habit in vibe coding is having a known-good state to return to when a fix cascades, and it costs nothing.

→Describe the data model before the interface. Tools that provision a database from a vague prompt tend to guess badly, and a wrong schema is the most expensive thing to change later.

→When you hit the second failed fix on the same bug, stop prompting and read the code. The loop rarely resolves itself, and this is the point where a free Cursor Hobby account saves more time than another twenty messages.

## The bottom line

Pick by what you intend to do with the result. If you want a working app from a description and you do not write code, start on Lovable's free tier and see how far 30 monthly credits take you. If you expect to iterate heavily, Bolt at $25 a month gives the most room, and the token rollover is a genuine advantage on debugging-heavy weeks.

If the output must live in a real Next.js codebase your team maintains, v0 at $30 per user per month emits code they will not resent. If the project is going to become software rather than stay a prototype, Replit at $20 on Core gives you a terminal and a database to fall back on.

And keep a free Cursor Hobby account open regardless, because the wall arrives for everyone, and the people who ship are the ones who stop prompting and start reading at the right moment.

## Frequently asked questions

What does vibe coding actually mean?

It describes building software by prompting a model and accepting the output without closely reviewing it, trusting the result because it appears to work. The term caught on because it names something people were already doing. It is genuinely productive for prototypes, internal tools and landing pages, and genuinely risky for anything handling authentication, payments or personal data.

Can I use these tools without knowing how to code?

Yes, up to a point, and Lovable is the best starting place. What you cannot avoid is the debugging wall: when the model produces something broken in a way it cannot diagnose, no amount of rephrasing fixes it, and you need someone who can read the code. Plan for that moment rather than being surprised by it.

Do I own the code these tools generate?

Generally yes, but check the export path rather than the licence page. The practical question is whether you can pull a conventional repository out and run it elsewhere. v0 and Bolt emit standard code, Replit gives you a real filesystem, and closed generators make leaving harder in practice even when the terms allow it.

Why do the credits run out so fast?

Because debugging costs several times what building costs. Every failed attempt sends the growing conversation back to the model, so the fifth fix on one bug is far more expensive than the first feature. This is why allowance structure matters: Bolt rolls unused tokens over, Replit denominates credits in dollars so overage is visible, and v0's free tier caps you at 7 messages a day precisely because a debugging loop would drain it in one sitting.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Lovable pricing](https://lovable.dev/pricing)checked Sep 2026
- [Bolt pricing](https://bolt.new/pricing)checked Sep 2026
- [v0 pricing](https://v0.dev/pricing)checked Sep 2026
- [Replit pricing](https://replit.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsAi App Development PlatformsPostgres HostingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Best Web Scraping Tools for 2026

URL: https://devshot.co/reviews/best-web-scraping-tools
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: Bright Data, ThorData, and Browse AI compared for developers who need reliable web scraping and data extraction in 2026.

Expert Guide

## Best Web Scraping Tools for 2026

A developer's guide to the scraping and data extraction platforms that actually hold up at scale, from raw proxies to no-code robots.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 6 tools compared

TL;DR

Bright Data is the pick for most engineering teams that need to scrape reliably at scale, with proxies, a scraper API, a headless browser, and ready-made datasets under one roof.

ThorData is the cheaper proxy-plus-API alternative when budget matters more than breadth, and Browse AI wins for no-code extraction and page monitoring. Volza is a specialist for global import-export trade data, not general scraping.

Key facts6 tools compared: Bright Data, Firecrawl, ThorData, Context, Browse AI, Volza

- Updated: September 25, 2026
- Top pick: Bright Data (best for: Engineering teams scraping at scale)
- Top pick price as of September 25, 2026: Bright Data: Free 5K records/mo; Scraper API $1.50 per 1K records; residential proxies from $5/GB; custom enterprise plans

- 6 tools compared: Bright Data, Firecrawl, ThorData, Context, Browse AI, Volza
- Firecrawl (best for: Turning sites into LLM-ready data): Free 1k credits; Hobby $16/mo annual; 10% off via partner link
- ThorData (best for: Budget-conscious teams needing proxies plus APIs): Residential proxies from about $2/GB; Scraper API from $1 per 1K results; free trials on all products
- Context (best for: Web data plus brand intelligence in one API): Free 1,000 credits/mo; paid from $19/mo; 25% off for 6 months via partner link

Web scraping breaks in two places: the sites you target fight back with bot detection, and the volume you need outgrows a hobby script. The tool you pick decides whether you spend your time writing extraction logic or fighting CAPTCHAs and IP bans.

This guide ranks four options by who they actually fit. Some are raw infrastructure for engineers, one is no-code for teams without a scraping stack, and one is a trade-data specialist. Match the tool to your real job, not the flashiest feature list.

## Top Picks

Based on features, real-world fit, and value for money.

Best Web Scraping & Data Extraction Tools in 2026: 6 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Bright Data](https://toolradar.com/tools/bright-data) | Free 5K records/mo; Scraper API $1.50 per 1K records; residential proxies from $5/GB; custom enterprise plans | Engineering teams scraping at scale |

Firecrawl | Free 1k credits; Hobby $16/mo annual; 10% off via partner link | Turning sites into LLM-ready data |

[ThorData](https://toolradar.com/tools/thordata) | Residential proxies from about $2/GB; Scraper API from $1 per 1K results; free trials on all products | Budget-conscious teams needing proxies plus APIs |

Context | Free 1,000 credits/mo; paid from $19/mo; 25% off for 6 months via partner link | Web data plus brand intelligence in one API |

[Browse AI](https://toolradar.com/tools/browse-ai) | Free plan (50 credits/mo); Personal from $19/mo billed annually; Professional from $69/mo annually; Premium from $500/mo | No-code extraction and site monitoring |

Volza | Free limited search; paid annual plans with pricing through sales (contact for a quote) | Global import-export trade data |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 6 do not publish a comparable monthly price and are left out rather than estimated.

1

### Bright Data

Top Pick

Best for: Engineering teams scraping at scale

PricingFree 5K records/mo; Scraper API $1.50 per 1K records; residential proxies from $5/GB; custom enterprise plans

+Large residential proxy network with high success rates on tough anti-bot sites

+Over 1,000 pre-built scrapers plus pay-only-for-successful-results billing

+Datasets option lets you buy data for common sources instead of collecting it

−Pricing spans many products and gets complex, and can climb quickly at volume

−Overkill with a real learning curve for small or one-off jobs

Visit Bright Data →

2

### Firecrawl

Best for: Turning sites into LLM-ready data

PricingFree 1k credits; Hobby $16/mo annual; 10% off via partner link

+Handles JavaScript, proxies, anti-bot

+Open source with SDKs

−Usage-based costs add up at scale

Visit Firecrawl →

3

### ThorData

Best for: Budget-conscious teams needing proxies plus APIs

PricingResidential proxies from about $2/GB; Scraper API from $1 per 1K results; free trials on all products

+Cheaper per-GB residential proxies than most tier-one vendors

+Covers the core kit: residential, mobile, ISP, and datacenter proxies plus scraper and SERP APIs

+Simple per-GB and per-1K pricing that is easy to forecast

−Smaller IP pool and shorter track record than Bright Data or Oxylabs

−Fewer pre-built scrapers and less mature tooling and documentation

Visit ThorData →

4

### Context

Best for: Web data plus brand intelligence in one API

PricingFree 1,000 credits/mo; paid from $19/mo; 25% off for 6 months via partner link

+Combines scraping and company data

+Handles rendering and anti-bot

−Newer entrant

Visit Context →

5

### Browse AI

Best for: No-code extraction and site monitoring

PricingFree plan (50 credits/mo); Personal from $19/mo billed annually; Professional from $69/mo annually; Premium from $500/mo

+Record a scraping robot by clicking through a page, no code required

+Built-in change monitoring and alerts for tracking prices or listings

+Generous free plan plus integrations with Google Sheets, Zapier, and more

−Credit system gets expensive as row counts and run frequency grow

−Less control than raw code when sites change layout or add hard anti-bot defenses

Visit Browse AI →

6

### Volza

Best for: Global import-export trade data

PricingFree limited search; paid annual plans with pricing through sales (contact for a quote)

+Covers billions of import-export shipment records across 200+ countries

+Search by product, HS code, buyer, or supplier to find leads and competitors

+Finished data means no scraping infrastructure to build or maintain

−Solves a narrow use case and is useless if you need to scrape arbitrary websites

−Pricing is opaque and pushed through sales rather than published clearly

Visit Volza →

## What it is

Web scraping and data extraction tools collect structured data from websites that do not offer a clean API. At the low end that means a script plus a proxy to avoid getting blocked.

At the high end it means managed infrastructure: rotating residential IPs, headless browsers that render JavaScript, anti-bot bypass, and pre-built scrapers for sites like Amazon and LinkedIn. Some vendors also sell finished datasets so you skip collection entirely.

## Why it matters

Choose badly and the cost shows up later. A cheap proxy pool with low success rates means half your jobs fail silently and your data has holes. A no-code tool hits a wall the moment you need 10 million pages. An enterprise platform bills you for scale you do not use yet.

Scraping also carries legal and compliance risk, so the vendor's stance on terms of service and personal data matters as much as speed. The right fit saves both engineering hours and a surprise invoice.

## Key features to look for

Anti-bot and proxy handlingEssential
The core of any serious scraper: rotating residential IPs, CAPTCHA solving, and JavaScript rendering that get you past the defenses on hard targets like LinkedIn, Amazon, and Google.

Success rate and reliabilityEssential
A high, consistent success rate matters more than a low headline price. Failed requests mean retries, missing rows, and silent data gaps you find out about too late.

Scale and pricing modelEssential
How the tool bills, per GB, per 1,000 records, or per credit, and whether it stays affordable from a few thousand pages to tens of millions without a painful tier jump.

Ease of use vs control
No-code robots suit teams without a scraping stack; a raw API and proxies give developers full control over parsing and pipelines. Pick the level of abstraction you need.

Pre-built scrapers and datasets
Ready-made scrapers for common sites and finished datasets let you skip building and maintaining extraction logic for sources like ecommerce, social, and business directories.

Compliance and legal safeguards
The vendor's stance on site terms, personal data, and know-your-customer checks. This limits your legal exposure, especially at scale or with regulated information.

Mistakes to avoid

×Buying enterprise scraping infrastructure for a job a $19 no-code tool would handle. Match the tool to your real volume, not the volume you imagine you will hit someday.

×Comparing only price per GB and ignoring success rate. A cheap proxy that fails 40% of requests costs more in retries, missing data, and debugging time than a pricier one.

×Assuming every tool here does the same thing. Volza sells finished trade data, Browse AI is no-code, and Bright Data is raw infrastructure. They are not interchangeable.

Expert tips

→Start on the free tier. Bright Data, Browse AI, and ThorData all let you test real jobs before paying, so validate success rates on your actual targets first.

→Estimate your monthly volume in records or GB before comparing plans. The cheapest headline rate is rarely the cheapest bill once your volume is real.

→Check the vendor's compliance stance on terms of service and personal data early, especially if you scrape at scale or touch anything regulated.

## The bottom line

For most engineering teams, Bright Data is the safe default. It scrapes the hardest sites, scales without you rebuilding anything, and offers datasets when you would rather buy than collect. If budget is the deciding factor and your targets are not the toughest, ThorData covers the same core needs for less.

Teams without a scraping stack, or anyone who just needs to monitor a few pages, should start with Browse AI's free plan. And if what you actually need is global trade data rather than web scraping, Volza is the specialist to look at. Pick by the job, not the brand.

## Frequently asked questions

What is the best web scraping tool for developers in 2026?

For most engineering teams it is Bright Data, because it combines proxies, a scraper API, a headless browser, and ready-made datasets in one platform and handles tough anti-bot sites. ThorData is the cheaper alternative if you mainly need proxies and basic scraping APIs.

Do I need to write code to scrape websites?

No. Browse AI lets you record a scraper by clicking through a page and turns it into a spreadsheet or API. Developers who want full control still prefer code plus a proxy provider like Bright Data or ThorData, but no-code is fine for smaller extraction and monitoring jobs.

Is web scraping legal?

Scraping public data is generally allowed in many places, but it depends on the site's terms, the data type, and your jurisdiction. Personal data and logged-in content raise the risk. Vendors like Bright Data run compliance checks, but you should still get legal advice for large or sensitive projects.

How is Volza different from the other tools?

Volza is not a web scraper. It is a database of global import-export shipment records you search by product, HS code, or company. If you need trade intelligence, sales leads, or supplier data, it fits. If you need to extract data from arbitrary websites, pick Bright Data, ThorData, or Browse AI.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Bright Data pricing](https://brightdata.com/pricing)checked Sep 2026
- [ThorData pricing](https://www.thordata.com)
- [Browse AI pricing](https://browse.ai/pricing)checked Sep 2026

Related guides

Ai App Development PlatformsDocker Desktop AlternativesAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Bright Data Review

URL: https://devshot.co/reviews/bright-data-review
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: A developer-focused Bright Data review: real pricing per GB and per 1,000 requests, where costs spike, and the alternatives worth testing first.

Review

## Bright Data Review

The most complete web data platform on the market, and among the priciest. Built for engineering teams that scrape defended sites at scale.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Bright Data is the most complete web data platform you can buy, and you pay for it. Residential proxies run about $4 to $8 per GB on pay-as-you-go, easing toward $3.50/GB on monthly commitments, with Web Unlocker and SERP API billed per 1,000 requests and a free trial to start.

The strength is coverage and reliability: a 400M+ IP pool, high success rates on defended sites, and modules for proxies, scraping, and prebuilt datasets under one roof. The catch is cost and complexity, plus mandatory KYC. If you want most of that power for less, Oxylabs is the closest rival worth pricing against it.

Key factsBest for: Engineering teams scraping defended sites at scale

- Updated: September 25, 2026
- Best for: Engineering teams scraping defended sites at scale
- Price as of September 25, 2026: From about $4 to $8/GB (residential, pay as you go); free trial, no free plan

- The most complete web data platform, and among the most expensive.
- Founded: 2014
- Headquarters: Netanya, Israel
- Customers: 20,000+ worldwide
- Free plan: No, free trial available

Pros

- Largest residential pool (400M+ IPs) with the best success rates on hard, defended targets
- One vendor spans raw proxies, unlocking APIs, scraping browsers, and prebuilt datasets
- Pay-per-success billing on Web Unlocker and a free trial to validate before committing

Cons

- Residential bandwidth gets expensive fast on JavaScript-heavy sites
- Mandatory KYC and compliance review can delay your first project
- Mixed billing units (per GB, per IP, per 1,000) and a broad console mean a steeper learning curve

Founded2014

HeadquartersNetanya, Israel

Customers20,000+ worldwide

Free planNo, free trial available

If you have ever tried to scrape a site behind Cloudflare, PerimeterX, or Akamai, you already know the problem Bright Data solves. Getting a single page is easy. Getting a million pages, reliably, from IPs that do not get instantly blocked, is a full infrastructure problem.

Bright Data is the largest and most established vendor built to solve exactly that, and it has grown from a pure proxy provider into a full web data stack.

The real question for an engineering team is not whether Bright Data works.

It does. The question is whether the coverage and success rates justify the price, and whether you would be better served by a cheaper proxy pool or a simpler per-request API.

This review is grounded in the live pricing and product pages, and it weighs Bright Data against the vendors developers actually cross-shop before signing off on a data budget.

## What is Bright Data?

Bright Data is an Israel-based web data platform, one of the oldest and largest vendors in the proxy and scraping market. At its core sit three proxy networks: residential (400M+ IPs across 195 countries), ISP proxies (static residential IPs on datacenter hardware), and datacenter proxies for raw speed and volume.

Every network supports country, city, state, ZIP, and ASN targeting, plus sticky or rotating sessions over HTTP/S and SOCKS5.

On top of that infrastructure Bright Data layers managed data products. The Web Unlocker takes a URL and returns clean HTML while handling CAPTCHAs, bot managers like Akamai, and proxy rotation for you. The SERP API returns structured results from Google, Bing, and other engines as JSON or HTML.

The Scraping Browser, or Browser API, hands you a remote stealth browser compatible with Puppeteer and Playwright for JavaScript-heavy targets. Above those, the Web Scraper APIs and AI Scraper Studio build structured extraction pipelines, and the Dataset Marketplace sells 5B+ prebuilt records across 250+ domains.

The point is that you can start with a raw proxy and move all the way up to bought-and-refreshed datasets without switching vendors.

## How Bright Data works

Setup starts in the control panel, where you create a zone for each product you want: residential proxies, Web Unlocker, SERP API, and so on. Each zone gets its own credentials and its own usage meter. For proxies you get a host, port, username, and password to drop into your HTTP client or scraping framework.

For the Web Unlocker and SERP API you send a request to a single endpoint and get structured data back, which is the part most developers actually want.

Integration is genuinely broad. There are code examples for Python, Node.js, Java, C#, and shell, native compatibility with Puppeteer, Playwright, and Selenium, and a Scraping Browser that plugs into existing headless scripts with a connection-string change.

The dashboard exposes per-zone logs, request statistics, and cost breakdowns, which matters when bandwidth is the meter.

The rough edges are real. Bright Data requires KYC verification, and some use cases or geos stay gated until you pass a compliance review, which can slow a first project by hours or days. Billing spans per GB, per IP, and per 1,000 requests depending on the product, so forecasting spend takes effort.

And the sheer breadth of the platform gives the console a learning curve that a single-purpose API does not.

## Bright Data key features

Residential Proxy NetworkEssential
A pool of 400M+ ethically sourced residential IPs across 195 countries, with targeting down to city, state, ZIP, and ASN level. Sticky or rotating sessions and a stated 99.95% success rate make it the reason most teams pick Bright Data for defended sites.

Web Unlocker APIEssential
Send a URL, get clean HTML back. It automatically solves reCAPTCHA, hCaptcha, Cloudflare Turnstile, and 30+ others, imitates real browser fingerprints, and you pay only for successful requests, billed per 1,000. A 98% success rate is claimed on hard targets.

Scraping Browser (Browser API)
A remote, stealth browser compatible with Puppeteer, Playwright, and Selenium. It runs the full page, executes JavaScript, and handles blocking at the browser layer, so existing headless scripts work with a connection-string change instead of a rewrite.

Web Scraper APIs & AI Scraper Studio
Prebuilt and custom scraping endpoints for popular sites that return structured JSON without you maintaining selectors or proxies. AI Scraper Studio turns a target into a repeatable pipeline, useful when page layouts change often and break brittle scrapers.

Dataset Marketplace
A catalog of 5B+ prebuilt records across 250+ domains, refreshed on a schedule with automated quality checks. For teams that need data more than infrastructure, buying a maintained dataset can replace building and running a scraper entirely.

SERP API
On-demand search results from Google, Bing, and other engines, returned as JSON or HTML with geo-targeting per query. Billed per 1,000 requests, it removes the CAPTCHA and layout-parsing pain of scraping search pages yourself, a common rank-tracking need.

## Bright Data pricing

Bright Data, Oxylabs, Decodo, Apify, ScraperAPI, and ThorData each publish an entry price on the core plan. The lowest residential rate is ThorData at about $2/GB on its smallest 1 GB pack, with a free trial, and Decodo is next at $3.75/GB on monthly plans, slightly more pay as you go.

Apify starts at $19/mo with a free plan, and ScraperAPI starts at $49/mo with a free credit allowance.

Bright Data's pay-as-you-go residential band is about $4 to $8/GB, and the bill jumps when defended, JavaScript-heavy pages burn bandwidth. Monthly commitments cut that unit rate toward $3.50/GB, and Oxylabs runs from $6/GB on its $30 Starter plan down to $2.50/GB at 1TB.

Web Unlocker, the SERP API, datasets, and Oxylabs scraper products are billed separately, on top of the per-GB rate.

Plan | Price | Best for |

Bright Data Residential pay as you go | About $4 to $8/GB, pay as you go | Entry bandwidth rate before a monthly commit |

Bright Data Residential commitment | About $3.50/GB, monthly plans | About $499/month for around 141 GB |

Bright Data ISP | About $1.80 per IP | Static residential IPs, billed per IP |

Bright Data Datacenter | About $1.40 per IP | Cheaper high-volume access, billed per IP |

Bright Data Web Unlocker | About $1.50 per 1,000 | Pay as you go, plus 5,000 free requests a month |

Bright Data SERP API | Per 1,000 requests | Structured search results, billed per 1,000 |

Bright Data Datasets | Per record or dataset | Prebuilt data, priced apart from proxies |

Oxylabs Residential | $4/GB, 125GB at $500/mo | Advanced plan rate; Starter is 5GB at $6/GB |

Oxylabs Residential 1TB | About $2.50/GB at 1TB | Volume residential rate on large plans |

Oxylabs Web Scraper API | Priced separately | Managed scraping, not included in proxy rates |

Oxylabs Web Unblocker | Priced separately | Unblocking billed apart from the proxy plans |

Decodo Residential pay as you go | $4/GB | Pay as you go, with a 3-day free trial |

Decodo Residential larger plans | $2.75 to $2/GB | 100GB monthly plan down to 1TB enterprise plan |

Apify Starter | $19/mo | Monthly plan, billed in compute units |

Apify Scale | $199/mo | Mid-tier plan plus proxy usage on top |

Apify Business | $999/mo | Highest listed plan, compute units plus proxies |

Apify Residential | About $8/GB | Proxy usage billed on top of the plan |

ScraperAPI Hobby | $49/mo | 100K credits, proxies included in the credit |

ScraperAPI Startup | $149/mo | 1M credits a month, flat per credit |

ScraperAPI Business | $299/mo | 3M credits a month, proxies included |

ThorData Residential | About $2/GB at 1 GB | Cheaper datacenter and ISP options, free trial |

## Bright Data pros and cons

### What we like

- Largest residential pool (400M+ IPs) with the best success rates on hard, defended targets
- One vendor spans raw proxies, unlocking APIs, scraping browsers, and prebuilt datasets
- Pay-per-success billing on Web Unlocker and a free trial to validate before committing

### What could be better

- Residential bandwidth gets expensive fast on JavaScript-heavy sites
- Mandatory KYC and compliance review can delay your first project
- Mixed billing units (per GB, per IP, per 1,000) and a broad console mean a steeper learning curve

## Who Bright Data is for

Bright Data is a strong fit for engineering and data teams scraping defended, high-value targets at scale: price monitoring across marketplaces, travel and real estate aggregation, brand protection, or feeding an AI pipeline with fresh web data.

If your bottleneck is getting blocked, not budget, Bright Data is usually the vendor that gets the data. The compliance posture and KYC also reassure larger companies that need a defensible, auditable source.

It is the wrong tool for a hobby project, a one-off scrape, or a startup counting every dollar. The per-GB residential rates and the console learning curve are overkill when you just need a few thousand pages a month.

A solo developer or small team on a tight budget is usually better served by ScraperAPI's flat per-credit pricing or Decodo's cheaper proxy pool. Teams that mainly need finished data, not infrastructure, should compare the Dataset Marketplace against building anything at all.

## Best Bright Data alternatives

If Bright Data is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Bright Data | Engineering teams scraping defended sites at scale | From about $4 to $8/GB (residential, pay as you go) | Visit → |

Oxylabs | Enterprises wanting a premium Bright Data rival | From $6/GB (Starter, 5GB for $30/mo), cheaper per GB on larger plans | Visit → |

Decodo (formerly Smartproxy) | Small and mid-size teams wanting cheaper proxies | From $3.75/GB (residential, monthly plan) | Visit → |

Apify | Developers who want ready-made scrapers and a platform | From $19/mo (Starter) | Visit → |

ScraperAPI | Solo devs and startups wanting simple per-request scraping | From $49/mo (Hobby, 100K credits) | Visit → |

ThorData | Budget-conscious teams needing residential proxies | From about $2/GB residential (1 GB pack), under $1/GB at 350GB | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 6 do not publish a comparable monthly price and are left out rather than estimated.

Oxylabs

The closest premium competitor, with comparable scale and a similar price tag.

Visit →

Decodo (formerly Smartproxy)

A lighter, cheaper proxy pool for teams that do not need the full stack.

Visit →

Apify

A scraping platform and actor marketplace, not just a proxy vendor.

Visit →

ScraperAPI

A single endpoint that handles proxies and CAPTCHAs, priced by credits.

Visit →

ThorData

A lower-cost proxy provider covering the core residential and datacenter needs.

Visit →

## The bottom line

Bright Data earns its reputation. If your job is to pull large volumes of data from sites that actively fight scrapers, it is the vendor most likely to succeed, thanks to the biggest residential pool, high success rates, and a full stack from raw proxies to bought datasets.

For price monitoring, AI training data, and enterprise-grade collection where reliability beats cost, it is the safe default.

The caveat is money and overhead. Residential bandwidth adds up quickly, and the platform's breadth plus mandatory KYC is more than a small project needs.

If budget is the constraint, look at Decodo for cheaper proxies or ScraperAPI for simple per-request scraping. If you want a premium platform with comparable scale, price it against Oxylabs. And if you would rather run prebuilt scrapers than manage infrastructure, Apify is the better home.

Trial Bright Data on your actual targets, measure GB per thousand pages, then decide with real numbers.

## Frequently asked questions

How much does Bright Data cost?

Residential proxies run about $4 to $8 per GB on pay-as-you-go, dropping toward $3.50/GB on monthly commitments like $499/month for roughly 141 GB. ISP proxies are about $1.80 per IP, datacenter around $1.40 per IP, and Web Unlocker and SERP API bill per 1,000 requests. There is a free trial plus a 5,000-request free tier on Web Unlocker.

Is Bright Data worth it for developers?

If you scrape defended sites at scale, yes. Its success rates and IP coverage are the best in the market, and pay-per-success billing on the Web Unlocker limits waste. For hobby projects or a few thousand pages a month, the per-GB rates and console complexity are overkill, and a simpler tool like ScraperAPI will cost far less.

Does Bright Data have a free plan or trial?

There is no permanent free plan, but there is a free trial, a matching sign-up bonus on your first deposit up to $500, and a free tier of 5,000 requests a month on the Web Unlocker. That is enough to validate success rates and measure your real bandwidth per thousand pages before you commit to a paid plan.

What are the best Bright Data alternatives?

Oxylabs is the closest premium rival, with similar scale and pricing. Decodo, formerly Smartproxy, is the cheaper mid-market proxy option. ScraperAPI is the simplest per-request choice for small teams, Apify is better if you want prebuilt scrapers and orchestration, and ThorData is a budget proxy provider worth a look.

Residential vs datacenter proxies: which should I use?

Datacenter proxies (around $1.40 per IP) are fast and cheap, and fine for sites with light defenses. Residential and ISP proxies use real consumer or ISP IPs that are far harder to block, which is why they cost more and are billed by bandwidth. Start on datacenter, and only move to residential for targets that block you.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Bright Data pricing](https://brightdata.com/pricing)checked Sep 2026
- [Oxylabs pricing](https://oxylabs.io/pricing)checked Sep 2026
- [Decodo pricing](https://decodo.com)checked Sep 2026
- [Apify pricing](https://apify.com/pricing)checked Sep 2026
- [ScraperAPI pricing](https://www.scraperapi.com/pricing)checked Sep 2026
- [ThorData pricing](https://www.thordata.com)

Related guides

Web Scraping ToolsDify ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Claude Code Review

URL: https://devshot.co/reviews/claude-code-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Claude Code review 2026: which paid seat covers a week of agent work, how the shared usage pool bills, and when Cursor, Copilot, Codex…

Review

## Claude Code Review

Worth it when the agent should live in the terminal and the chat subscription you already pay. A weak default when the job is unlimited autocomplete in the IDE you already use.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Claude Code is worth it in 2026 when you want an agent on the repo you already have open, and you will share one usage pool with Claude chat. Annual Pro is $17/mo with Claude Code included, and Max 5x multiplies that session allowance by five. The free plan leaves Claude Code off, verified on Anthropic's pricing page in September 2026.

The terminal, VS Code, JetBrains, the desktop app, and the browser all draw from that same pool, so a long coding session can pause chat too. For autocomplete that never touches the credit meter, start with the GitHub Copilot review.

Key factsBest for: Engineers who want a coding agent in the terminal, VS Code…

- Updated: September 25, 2026
- Best for: Engineers who want a coding agent in the terminal, VS Code, or JetBrains.
- Price as of September 25, 2026: From $20/mo on Pro; annual Pro is $17/mo; not on the free plan.

- Anthropic's repo agent, included with paid Claude plans and absent from the free plan.
- Vendor: Anthropic
- Free plan: No Claude Code
- Paid entry: Pro, monthly or annual
- Alternatives covered: Cursor, GitHub Copilot, Codex, Devin

Pros

- Pro, Max, Team, and Enterprise all include it, so you do not buy a second coding SKU.
- The same CLAUDE.md, MCP servers, and settings follow the CLI, the IDE plugins, desktop, and the browser.
- Team limits are per member, so one heavy user does not drain the rest of the company.

Cons

- Chat and the agent share one five-hour window, and Anthropic publishes no fixed prompt count.
- An ANTHROPIC_API_KEY in the environment bills the Console at API rates instead of the subscription.
- Max has no annual discount, and Enterprise adds API usage on top of the seat.

VendorAnthropic

Free planNo Claude Code

Paid entryPro, monthly or annual

Best forTerminal and IDE agents

Claude Code is not a second subscription sitting next to Claude. It is included on every paid Claude plan, and it is absent on the free one, so the buying question is which seat covers a real week of agent work. A five-hour window plus a weekly cap is the limit you will hit.

Anthropic does not publish a fixed prompt count, because a migration and a one-line fix do not cost the same.

Toolradar data: the [September 2026 AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 153 tools. [Claude Code](https://toolradar.com/tools/claude-code) and [Cursor](https://toolradar.com/tools/cursor) both have directory pages there.

The longer head-to-head is [Claude Code versus Cursor](https://toolradar.com/blog/claude-code-vs-cursor-2026) on that site.

How we compared: pricing and product pages for Claude, Cursor, GitHub Copilot, Codex, and Devin, read on September 24, 2026, in USD. A hands-on setup guide lives at how to use Claude Code.

## What is Claude Code?

Claude Code is Anthropic's coding agent for the repo you already have open. It reads the repository, edits files, runs commands, and can open a pull request, so the seat pays for work on the codebase rather than another chat tab.

Anthropic's pricing comparison marks it unavailable on Free and available on Pro, both Max tiers, Team, and Enterprise, including self-serve Enterprise, so a free login cannot turn the agent on.

You can run it in the terminal, in VS Code (the extension also installs inside Cursor), and in JetBrains. It also runs in the Claude desktop app, the browser, and the iOS and Android apps. You do not have to switch editors to use it, which is the point of that spread.

A CLAUDE.md file in the project root is read at the start of every session. If the repo already carries an AGENTS.md for another agent, Claude Code can read that file on its own or beside CLAUDE.md, so existing agent rules still apply.

The product asks before it writes a file or runs a command, which is the control you want on a shared repo. Anthropic says the terminal session talks to the model without a remote code index. Set ANTHROPIC_API_KEY and the CLI skips the subscription login and asks you to approve that key.

The session then bills the Console at API rates instead of the seat. Teleport, which pulls a web session into the terminal, requires a claude.ai subscription rather than that key, so a Console-only account cannot make that handoff.

Individual plans let you opt out of model training, while Team and Enterprise do not train on your content by default, which is what to clear with counsel before a pilot. SSO and SCIM stay off Pro and Max, so a personal seat cannot meet an identity rule.

They start on Team for SSO, and on Enterprise for SCIM, audit logs, and a HIPAA-ready offering, so none of that sits on a personal seat.

## How Claude Code works

Install the CLI, open a project, and run claude, then describe the change in plain language. The agent searches the tree, edits several files, and can run the test command you already use, so the check is the one the repo already trusts.

Hooks can format after an edit or lint before a commit, and skills wrap a repeatable prompt, such as a pull-request review, so the team does not retype it.

Limits reset on a rolling five-hour window, and paid plans add a weekly cap on top. Chat on the web, the desktop app, and Claude Code share that allowance, so a long agent session can pause chat until the reset.

When the bar is empty you wait for the reset, move to a higher plan, or turn on usage credits, which continue at standard API rates. There is no published message quota to budget against, so you plan the week rather than a message count.

Cloud routines keep running after the laptop sleeps, so a task can finish with the lid closed. You create them on the web, in the desktop app, or with /schedule in the CLI.

For review comments on the pull request after the agent opens it, the AI code review list is the next decision, and CodeRabbit is the dedicated reviewer rather than a second coding agent.

## Claude Code key features

One pool across chat and the agentEssential
Terminal work and claude.ai chat draw from the same five-hour window, with a weekly cap on paid plans, so a heavy coding block can pause chat. A migration and a one-line fix do not draw the same share of that window.

Terminal, IDE, desktop, and webEssential
The CLI, a VS Code extension that also installs in Cursor, a JetBrains plugin, the desktop app, and claude.ai/code share CLAUDE.md, settings, and MCP servers. One written standard then follows you onto every surface you open. JetBrains still needs the CLI installed beside the plugin, which is an extra step if that is the IDE the team uses.

Permission checks and a local sessionEssential
Claude Code asks before it changes files or runs a command. That pause is the brake you want on a shared branch. Anthropic says the terminal path does not require a remote code index. An ANTHROPIC_API_KEY in the environment bills the Console instead of the subscription login.

CLAUDE.md, skills, hooks, and subagents
CLAUDE.md sets standards for every session, and an existing AGENTS.md can be read too, so you do not start from a blank rules file. Skills package a shared workflow, hooks run a shell command around an edit, and a lead agent can split work across subagents.

Opus 5.5 fast mode on usage credits
Fast mode for Opus 5.5 is a research preview at 2.5 times the usual speed, priced at twice the standard Opus 5.5 token rates. The speedup is a surcharge on those token rates. Subscription users reach it through usage credits, while consumption plans can use it directly.

Routines and phone handoff
Routines run in the cloud when the computer is off, including on a GitHub event. Teleport and the desktop handoff require a claude.ai subscription, which leaves API-key users out of that path. A Slack mention of @Claude can send a bug report back as a pull request.

## Claude Code pricing

Claude Code has no price of its own, so you pay for a Claude plan and the agent is either included or absent. Pro is the individual entry, and Max is monthly only, in a 5x tier and a 20x tier, each measured against Pro inside the five-hour window.

A heavier week costs more per month, because you are buying a multiple of that same window. Annual billing exists for Pro and for Team, not for Max, and the annual Pro charge is $200 up front.

Team needs at least two members and stops at 150 seats, so one person cannot buy SSO here. Standard seats include 1.25 times Pro's per-session allowance, and Premium seats include 6.25 times Pro, which is five times a Standard seat. Premium is for the people who actually run long sessions.

Limits are per person, so one developer burning the week does not empty a teammate's bar. Usage credits, if an admin turns them on, let that person continue at API rates. Monthly Standard and ChatGPT Business, billed monthly, are each $25 per user, the comparison if the alternative is a ChatGPT team seat.

Premium monthly is $125 per user, and the annual Premium seat matches the Max 5x sticker, per member, so you pay that multiple once per person.

Enterprise bills differently from a flat individual seat, and the invoice grows with usage. The pricing FAQ lists a seat fee plus usage at API rates, billed annually, with SCIM, audit logs, custom retention, and role-based access.

Anthropic's own cost guide, on the Claude Code costs page, puts a typical enterprise deployment around $13 per developer per active day and $150 to $250 per developer per month. It also says costs stay under $30 per active day for 90% of users. That range is their figure, from pilots, not a quote you can paste into a budget.

On the API, Opus 5.5 lists $4 per million input tokens, and output is five times that input rate, so a long reply costs more than the prompt. US-only inference is 1.1 times the input and output rates, and fast mode doubles those Opus 5.5 token rates.

Cursor's docs recommend Pro Plus, at $60/mo, for daily agent use, the plan to compare when the agent runs all day. Teams adds $0.25 per million tokens on third-party models, including Auto routes that land on one, so the seat is not the whole team bill.

Copilot bills extra AI credits at $0.01 each, and it does not meter inline completions on paid plans. A completions-heavy day stays off the meter that agent tasks spend.

Plan | Price | Best for |

Claude Free | No-cost | Claude Code is marked off. Chat, artifacts, and a small project cap remain. |

Claude Pro, monthly | $20/mo | Includes Claude Code. Shares the five-hour window with chat. |

Claude Pro, annual | $17/mo | Anthropic's stated monthly equivalent of $200 billed once. |

Claude Max 5x | $100/mo | 5x Pro per five-hour session. Monthly only. Priority at peak times. |

Claude Max 20x | $200/mo | 20x Pro per session, plus higher output limits. Monthly only. |

Claude Team Standard | $25/user/mo | Monthly. Minimum two seats. 1.25x Pro, limits per member, not pooled. |

Claude Team Premium | $125/user/mo | Monthly. 6.25x Pro per session. Mix with Standard seats. |

Claude Enterprise | $20/seat/mo plus API | Annual. Seat fee plus API usage. SCIM, audit logs, HIPAA-ready. |

Cursor Pro | Matches monthly Pro | Hobby is free. Two usage pools: Cursor models and other models. |

Cursor Pro Plus | $60/mo | The plan Cursor recommends for daily agent users. |

Cursor Ultra | Matches Max 20x | Top individual plan. Teams Standard is a separate per-user rate. |

Cursor Teams Standard | $40/user/mo | SSO. Premium seats are $120/user/mo for 5x the agent limits. |

GitHub Copilot Pro | $10/mo | 1,500 AI credits. Completions on paid plans stay outside that meter. |

GitHub Copilot Pro+ | $39/mo | 7,000 AI credits. Free plan stops at 2,000 completions a month. |

GitHub Copilot Max | $100/mo | 20,000 AI credits. Business is $19 per granted seat. |

GitHub Copilot Enterprise | $39/user/mo | 3,900 credits per user, pooled. Not sold for Enterprise Server. |

Codex on ChatGPT Plus | Matches monthly Pro | CLI, IDE, web, and iOS. No annual Plus or Pro billing. |

Codex on ChatGPT Pro | From $100/mo | 5x or 20x Plus. New sign-ups to the 20x tier paused Sept 10, 2026. |

Codex Go | $8/mo | Light coding tasks. API-key use has no cloud review or Slack. |

Devin Pro | Matches monthly Pro | Frontier models, Devin Cloud, extra usage at API rates. |

Devin Max | Matches Cursor Ultra | Higher quotas. SWE-2 Free in Desktop and CLI through Oct 10, 2026. |

Devin Teams | $80/mo minimum | $40/mo per full seat, so two seats meet the minimum. Flex seats skip that charge. |

## Claude Code pros and cons

### What we like

- Pro, Max, Team, and Enterprise all include it, so you do not buy a second coding SKU.
- The same CLAUDE.md, MCP servers, and settings follow the CLI, the IDE plugins, desktop, and the browser.
- Team limits are per member, so one heavy user does not drain the rest of the company.

### What could be better

- Chat and the agent share one five-hour window, and Anthropic publishes no fixed prompt count.
- An ANTHROPIC_API_KEY in the environment bills the Console at API rates instead of the subscription.
- Max has no annual discount, and Enterprise adds API usage on top of the seat.

## Who Claude Code is for

Claude Code fits a developer who already lives in the terminal or who wants the agent beside VS Code or JetBrains without adopting a second editor. Indie hackers and small teams get the cleanest bill on Pro, because one login covers chat and the agent.

Engineering leads who need SSO should start at Team, and they should assign Premium only to the people who will actually run long sessions, since the limit does not pool.

It is the wrong default when the daily habit is Tab-to-accept completions. Copilot keeps those completions off the credit meter on every paid plan. Cursor's Tab model is the reason people stay in that editor, and the Cursor review is the editor comparison.

The Cursor versus Copilot note covers that pair without the terminal agent in the middle.

Skip a personal Max plan when the company will pay for a seat. Enterprise's seat fee looks small until API usage lands on the same invoice, which is why Anthropic publishes a per-active-day average instead of a flat seat.

If the bug is a failing test rather than a missing feature, the debugging tools list is a narrower buy than another coding subscription.

Devshot's daily brief is where a pause like Codex's September signup stop belongs. Subscribe free if you want that note in the inbox.

## Best Claude Code alternatives

If Claude Code is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Claude Code | Engineers who want a coding agent in the terminal, VS Code, or JetBrains. | From $20/mo on Pro | Visit → |

Cursor | Developers who want Tab completions and an agent inside a VS Code fork. | Pro at the monthly individual rate | Visit → |

GitHub Copilot | Teams that want IDE completions plus a CLI, on a GitHub seat they already manage. | From $10/mo (Pro) | Visit → |

Codex | Developers who already pay for ChatGPT and want a CLI and IDE agent. | Included with ChatGPT Plus | Visit → |

Devin | Engineers who want Tab completions, cloud agents, and an editor of their own. | Pro matches the monthly individual rate | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 5 do not publish a comparable monthly price and are left out rather than estimated.

Cursor

An AI editor with its own model pool plus third-party models, and a free Hobby plan.

Visit →

GitHub Copilot

The IDE assistant whose paid plans leave inline completions outside the AI-credit meter.

Visit →

Codex

OpenAI's coding agent on ChatGPT plans, with a 20x tier closed to new buyers.

Visit →

Devin

Cognition's agent editor, with a free quota, a Pro plan, and per-seat Teams billing above an $80 monthly minimum.

Visit →

## The bottom line

Pay for Pro if the agent is the reason you open a terminal, and pay annually if you know you will still be using it in month twelve. Move to Max 5x when the five-hour window is the thing that stops a workday, and treat Max 20x as a personal power plan rather than a team design.

Put a company on Team when you need SSO and per-person limits, and on Enterprise only after someone has modeled API usage on top of the seat.

Choose Cursor when Tab completions inside a dedicated editor are the habit you will not give up. Choose GitHub Copilot when completions should stay off a meter and the seat should live in GitHub.

Choose [Codex](https://toolradar.com/tools/openai-codex) when the ChatGPT login should be the agent, and check whether the 20x tier is still paused before you promise it to someone. Choose [Devin](https://toolradar.com/tools/devin) when you want that editor shape and can live with a team fee plus seats.

The wider set of assistants is the AI coding assistants list, and the chat-model version of this choice is ChatGPT versus Claude for coding. Subscribe free if you want the next plan change in the Devshot brief.

Cite this: Devshot, "Claude Code Review 2026", September 2026.

## Frequently asked questions

Is Claude Code worth it in 2026?

Yes, when you want a coding agent in the terminal, VS Code, or JetBrains and you can share one usage pool with Claude chat. Annual Pro includes Claude Code, Max 5x multiplies that Pro session allowance by five inside each five-hour window, and the higher Max tier multiplies it by twenty. It is the wrong buy when you mainly want unlimited autocomplete, which is Copilot's paid-plan behavior, or when you want Tab inside a dedicated editor, which is Cursor. The free Claude plan does not include Claude Code.

How much does Claude Code cost?

There is no separate Claude Code price, so the chat plan is the coding plan. Monthly Pro is $20 and includes it, and annual Pro is $17 a month, with $200 billed up front. Max 5x is $100 a month and Max 20x costs twice that, both monthly only, at 5x and 20x the Pro session allowance. Team Standard is $25 per user monthly, Premium is $125 per user monthly, and Enterprise is a seat fee plus API usage, billed annually. Prices were verified on Anthropic's pricing pages in September 2026.

Is there a free Claude Code plan?

No, the pricing table marks Claude Code off on the free plan and on for every paid plan. The free plan still covers chat, web search, artifacts, and up to five projects. A Console account can run Claude Code on API credits instead, and an ANTHROPIC_API_KEY in the environment takes that path even if you also pay for Pro. That is a metered bill, not a free tier.

What happens when Claude Code hits its limit?

The allowance is a rolling five-hour window, plus a weekly cap on paid plans, shared with Claude chat. Anthropic does not publish a fixed number of prompts, so you cannot budget a sprint as a message count. When you hit the limit you wait for the reset, move to a higher plan, or turn on usage credits and keep working at standard API rates. On Team, the cap is per member, so a teammate who still has allowance is unaffected. Pro and Max can show a monthly spend limit for those credits, or Unlimited if you never set one.

How does Claude Code compare with Cursor?

Claude Code keeps your current editor and bills through a Claude subscription, so you keep the editor you already use. Cursor is its own editor, with Pro at the same monthly individual rate as Claude Pro, Pro Plus at $60 a month for daily agent use, and Ultra as the top individual plan. Cursor includes unlimited Tab completions on those paid individual plans, and Claude Code does not offer that autocomplete. Many people run the Claude Code extension inside Cursor, which means paying for both if both pools get heavy use.

How does Claude Code compare with Codex?

Both are agents you can run from a CLI without switching editors, so the choice is which chat subscription owns the agent. Codex comes with ChatGPT Plus at the same monthly rate as Claude Pro, and with Pro at 5x or 20x that Plus allowance. OpenAI paused new sign-ups and upgrades to the 20x Pro tier on September 10, 2026. Claude's higher Max tier was still listed as a monthly plan on the same day. Codex on an API key drops cloud features such as GitHub review and Slack. Claude Code on an API key bills token rates and skips the subscription login.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026

Related guides

Cursor ReviewGithub Copilot ReviewAi Coding AssistantsAi For Code ReviewAi For DebuggingCoderabbit ReviewVibe Coding ToolsAi App Development PlatformsOpenCode vs Claude Code 2026: Cost, Models and Which to UseCodex vs Claude Code 2026: Plans, Limits and Real CostsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# CodeRabbit Review

URL: https://devshot.co/reviews/coderabbit-review
Type: review
Published: 2026-09-17
Updated: 2026-09-25
Summary: CodeRabbit review for 2026: 14-day trial, free public repos, Essentials $24, Team $48, Advanced $72 per developer per month billed annually.

Review

## CodeRabbit Review

AI review on every pull request: summaries, line comments, one-click fixes. Public repos stay free. Paid seats start at $24 per developer per month billed annually.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

CodeRabbit is an AI reviewer that comments on every pull request with a plain-English summary, line notes, and one-click fixes. Public open-source repos are free forever. Private work starts with a 14-day trial, no card. Essentials is $24 per developer per month billed annually ($30 monthly), Team is $48 annual ($60 monthly), Advanced is $72 billed annually, and Enterprise is custom. Extra reviews after the included hourly cap cost $0.25 per file. The Agent add-on is $0.40 per minute. Hourly PR review caps are 5 / 8 / 10 / 12 on Essentials / Team / Advanced / Enterprise. Partner link, no extra percent off. Closest alternatives: Greptile, GitHub Copilot PR review, Graphite, Qodo, and Bito.

Key factsBest for: Teams that want an AI reviewer on every pull request, with…

- Updated: September 25, 2026
- Best for: Teams that want an AI reviewer on every pull request, with controls to cut noise.
- Price as of September 25, 2026: From $24/dev/mo (Essentials, billed annually); 14-day trial, no card; public repos free

- A drop-in PR reviewer with one-click fixes, free public repos, and seats from $24 a developer.
- Founded: 2023
- Headquarters: San Francisco, USA
- Alternatives covered: Greptile, GitHub Copilot, Graphite, Qodo, Bito

Pros

- Public open-source reviews stay free, and the 14-day trial needs no card.
- Learnings, one-click fixes, and plan-level controls (triage, custom checks) to cut noise.
- Published overages ($0.25/file, $0.40/min Agent) instead of a mystery invoice.

Cons

- Hourly caps (5 on Essentials) bite during merge rushes unless you enable usage billing.
- Can still restate the diff on large or messy PRs.
- Advanced and Enterprise are a real jump if you only wanted summaries.

CodeRabbit

Founded2023

HeadquartersSan Francisco, USA

Est. priceFree OSS to $72/dev/mo

Best forAI pull request review

CodeRabbit sits on the pull request, not in your editor. You open a PR, it posts a summary, then it leaves inline comments with suggested patches you can commit from the review. That is a different job from Copilot-in-the-IDE. The question is whether those comments catch real bugs without burying the human reviewer in nits.

This review is for engineering leads and developers deciding if a dedicated reviewer is worth another seat on top of their coding assistant. Pricing below is from coderabbit.ai/pricing as of September 2026. I have a partner link at https://coderabbit.link/louis-dupple-com.

There is no extra percent off on that link, just the referral. I am not claiming a bake-off on your repos. I am laying out the published plans, the usage meters that actually change the bill, and how Greptile, GitHub Copilot, Graphite, Qodo, and Bito differ on paper.

One billing habit to get straight before the seats look cheap. CodeRabbit sells a per-developer month, then meters bursts ($0.25 a file) and agent minutes ($0.40). A ten-person squad on Essentials annual is $240 a month before anyone hits the hourly cap.

That is the number to put next to Copilot Business, Greptile Pro, and a Graphite Team seat, not the $24 sticker alone.

## What is CodeRabbit?

CodeRabbit is an AI-first pull request reviewer built by CodeRabbit (founded 2023, San Francisco, with a large team in India). You install it on GitHub or GitLab. It reads the diff, writes a summary, and comments line by line.

Under the LLM it also runs linters and SAST tools, so a single check can cover style, known vulnerability patterns, and the obvious logic slip.

The 2026 plan names matter because older posts still say Pro and Pro Plus. Essentials is the old Pro. Team is the old Pro Plus. Existing subscriptions were renamed in place.

The product around those seats now includes agentic reviews on PRs and the CLI, one-click fixes, a learnings store so it stops repeating nits you dismissed, loops with coding agents, MCP connections, and chat. Team adds triage, custom pre-merge checks, finishing touches (unit tests, merge-conflict help, simplify), and post-merge actions.

Advanced adds blast-radius analysis, architectural impact, and a security review on every PR plus continuous security monitoring.

Public open-source repos stay free. That is a real plan, not a teaser: install on a public repo and reviews keep running. Private repos and the deeper features are what you pay for.

Git host coverage is GitHub and GitLab on the public pages. Bitbucket teams should look at Bito instead of hoping a silent adapter appears. Seats follow people who open pull requests, which is friendlier than billing every GitHub user in the org, and you can reassign a seat when someone leaves. That is the procurement detail finance will ask for after the trial.

## How CodeRabbit works

You connect the Git host, pick repos, and CodeRabbit reviews new pull requests. On each PR it posts a summary, then inline comments. Suggested fixes can be applied in one click. Learnings let you teach it "stop flagging this" so the next PR is quieter.

You can also run reviews from the CLI and keep an agent loop going while you work in Cursor or Copilot.

The meter people miss is the hourly review cap. There is no hard monthly ceiling on how many PRs you can open. Each plan allows a burst of reviews per developer per hour: 5 on Essentials, 8 on Team, 10 on Advanced, 12 on Enterprise, subject to a fair-use policy.

If a squad merges a huge stack before lunch, you hit the cap. Admins can then turn on usage-based reviews (automatic, on demand, or off) with a monthly spend cap. Continued reviews bill $0.25 per reviewed file.

The other meter is CodeRabbit Agent at $0.40 per agent minute for cloud coding tasks, Slack, and automations, after any trial or free minutes. Enabling paid Agent usage opens it for the org, again with a spend cap.

Seats are billed for developers who create pull requests, and you can reassign seats. You are not paying for every GitHub account that never opens a PR.

Noise is the other operational problem. A reviewer that restates the diff trains humans to collapse the thread. Learnings exist to stop that, but only if someone on the team actually dismisses junk and lets the store update.

Custom pre-merge checks on Team are how you encode house rules (no new raw SQL in the API package) instead of arguing with a generic model. If nobody will own that config, you will get a chatty bot and blame the vendor.

## CodeRabbit key features

Agentic PR and CLI reviewEssential
Summaries, line comments, and one-click fixes on every pull request, plus CLI reviews so the same reviewer runs outside the GitHub UI. This is the product. Everything else is a multiplier.

Learnings and coding-agent loopsEssential
Dismissed nits can become learnings so the bot stops repeating itself. Loops with coding agents keep the reviewer in the same workflow as the tool that wrote the patch.

Triage and custom pre-merge checks
Team adds triage plus custom checks you define, so the reviewer can enforce house rules instead of only generic style. This is where a noisy install usually gets usable.

Blast radius and security review
Advanced adds architectural impact, a security pass on each PR, and continuous security monitoring. Full codebase scans are a separate usage-priced Security Scan, not the seat.

Free public-repo reviews
Open-source repos get reviews forever at $0. That is the honest way to see the comment style before you put a card on a private org.

MCP, Jira, and Linear
MCP connections (5 on Essentials, more on higher plans), plus Jira and Linear, so the reviewer can see the ticket the PR claims to close. Useful, not a reason to buy by itself.

## CodeRabbit pricing

AI code review is sold as a seat per developer, then a second meter for bursts: extra files, credits, or lines of code. CodeRabbit, Greptile, GitHub Copilot, Graphite, Qodo, and Bito each publish an entry price.

Enterprise is a custom quote at CodeRabbit, Greptile, and Bito, and GitHub Copilot prices Business and Enterprise seats separately from its individual plans. The cheapest published dedicated seat is Bito Team at $12 per seat per month, billed annually.

CodeRabbit Essentials is $24 per developer per month billed annually, or $30 month to month, and usage past the hourly cap is $0.25 per file, and the Agent is $0.40 per minute.

The other jumps are Greptile at about $1 per extra credit, Bito at $5 per extra 1,000 lines after about 5,000 per seat, Graphite Team at about $40 per user per month billed annually for unlimited reviews, and Qodo Pro Team from about $30 per month for up to 30 users.

Plan | Price | Best for |

CodeRabbit public repos | $0 | Public repository reviews stay free forever |

CodeRabbit Essentials | $24/dev/mo annual ($30 monthly) | 14-day trial, no card, 5 reviews/hour, 5 MCP connections, 1-repo analysis |

CodeRabbit Team | $48/dev/mo annual ($60 monthly) | 8 reviews/hour, 10 custom checks, 10 MCP connections, 5 repos |

CodeRabbit Advanced | $72/dev/mo billed annually | No separate monthly rate, 10 reviews/hour, 15 MCP connections, 10 repos |

CodeRabbit Enterprise | Custom quote | 12 reviews per hour, billed annually |

CodeRabbit extra reviews | $0.25 per file | Past the hourly cap if usage billing is on |

CodeRabbit Agent | $0.40 per minute | Cloud coding tasks, Slack, and automations |

Greptile Free | Free | 50 credits a month, 1 developer |

Greptile Pro | $30/seat/mo | Published per-seat plan |

Greptile extra reviews | About $1 per credit | Extra reviews beyond included credits |

Greptile startup discount | Advertised discount | Advertised for companies under $2M revenue |

Greptile Enterprise | Custom quote | For teams that want self-hosted Enterprise |

GitHub Copilot Free | Free | Does not include pull request review |

GitHub Copilot Pro | $10/mo | Lowest published individual seat |

GitHub Copilot Pro+ | $39/mo | Individual seat above Pro |

GitHub Copilot Max | $100/mo | Highest published individual seat |

GitHub Copilot Business | Priced separately | Business seats, separate from individual plans |

GitHub Copilot Enterprise | Priced separately | Enterprise seats, separate from individual plans |

Graphite Hobby | Free | Free Hobby plan, with limits |

Graphite Starter | About $20/user/mo | Starter seat below the Team plan |

Graphite Team | About $40/user/mo, billed annually | Unlimited reviews |

Qodo Pro Team | From about $30/mo | Up to 30 users, plus credits, 14-day trial |

Bito Team | $12/seat/mo, billed annually | Lowest published seat in this comparison |

Bito Professional | $20/seat/mo, billed annually | Higher published seat above Team |

Bito line overage | $5 per extra 1,000 | After about 5,000 lines of code per seat |

Bito Enterprise | Custom quote | Enterprise plan at a custom price |

## CodeRabbit pros and cons

### What we like

- Public open-source reviews stay free, and the 14-day trial needs no card.
- Learnings, one-click fixes, and plan-level controls (triage, custom checks) to cut noise.
- Published overages ($0.25/file, $0.40/min Agent) instead of a mystery invoice.

### What could be better

- Hourly caps (5 on Essentials) bite during merge rushes unless you enable usage billing.
- Can still restate the diff on large or messy PRs.
- Advanced and Enterprise are a real jump if you only wanted summaries.

## Who CodeRabbit is for

CodeRabbit is a strong fit for teams that want a drop-in reviewer on every private PR and already live on GitHub or GitLab. Open-source maintainers should install it on public repos at $0 and stop there unless they also have private code.

Essentials at $24 annual is the default paid seat for a squad that wants comments and one-click fixes. Move to Team at $48 when you need triage and custom checks. Advanced at $72 is for teams that will actually use blast-radius and the security pass, not for the logo.

It is a weaker fit if you already pay for Copilot and only wanted shallow PR comments, because Copilot review is then "free" in the sense that you are not adding a second vendor.

It is also a weaker fit if your whole process is stacked diffs (Graphite owns that workflow) or if you care more about whole-repo graph context than a tunable comment stream (Greptile's bet). Watch the $0.25/file overage and the $0.40/min Agent if a few people run huge refactors or long agent sessions. Those meters are how a predictable seat becomes a surprise invoice.

Open-source maintainers should install on public repos and stop. That is a complete product at $0. Mixed orgs (public plus private) can keep OSS free and pay only for the private seats. Monorepos with 40-file refactors are exactly where the hourly cap and $0.25/file toggle matter; decide the admin policy before the first merge week, not during it.

## Best CodeRabbit alternatives

If CodeRabbit is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

CodeRabbit | Teams that want an AI reviewer on every pull request, with controls to cut noise. | From $24/dev/mo (Essentials, billed annually) | Visit → |

Greptile | Teams that want comments based on the whole codebase, not only the lines that changed. | From $30/seat/mo (Pro) | Visit → |

GitHub Copilot | Teams already paying for Copilot who want review inside GitHub. | From $10/mo (Pro) | Visit → |

Graphite | Teams that already merge stacked pull requests in Graphite. | From about $20/user/mo (Starter) | Visit → |

Qodo | Squads that want the PR checked against the Jira or Linear ticket, not just the diff. | From about $30/mo (Pro Team) | Visit → |

Bito | Companies that review GitHub, GitLab, and Bitbucket from one cheap seat. | From $12/seat/mo (Team, billed annually) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Greptile

Indexes the repo as a graph so a comment on one function knows the callers three files away.

Visit →

GitHub Copilot

PR review bundled into GitHub Copilot, shallow compared with a purpose-built reviewer but already paid for.

Visit →

Graphite

AI review for stacked pull requests, aware of the diffs underneath the one you are reading.

Visit →

Qodo

Validates a pull request against the linked ticket and flags work that skips part of the scope.

Visit →

Bito

The lowest published per-seat AI reviewer that covers all three major Git hosts.

Visit →

## The bottom line

CodeRabbit is the default dedicated reviewer if you want comments on every private PR and a price you can explain to finance. Public repos stay free. The 14-day trial needs no card. Essentials at $24 annual ($30 monthly) is the seat to model unless you already know you need Team's triage or Advanced's security pass. The partner link does not discount the list price.

Read the meters before you roll it out org-wide. Five reviews per hour on Essentials is fine until a merge week. Then $0.25 per file and $0.40 per Agent minute decide whether the bill stays boring. Already on Copilot and only want shallow review?

Stay there. Need graph context? Greptile. Stack PRs? Graphite. Ticket fidelity? Qodo. Three Git hosts on a budget? Bito. CodeRabbit wins the middle: drop-in, tunable, published prices.

If your review culture is already strong and you only wanted a summary bot, Copilot is enough. If your review culture is a rubber stamp, CodeRabbit will not fix that. It will make the mechanical pass faster. Buy the seat for that, turn learnings on, and keep a human on architecture.

## Frequently asked questions

How much does CodeRabbit cost?

Public open-source repos are free. Paid seats: Essentials $24 per developer per month billed annually ($30 monthly), Team $48 annual ($60 monthly), Advanced $72 billed annually, Enterprise custom. Extra reviews after the hourly cap are $0.25 per file. CodeRabbit Agent is $0.40 per minute. There is a 14-day trial with no credit card. Prices are from coderabbit.ai/pricing, September 2026.

Is CodeRabbit free for open source?

Yes. Sign in with GitHub or GitLab, install CodeRabbit on a public repository, and reviews stay free. Private repos and the deeper Team / Advanced features require a paid seat after the trial.

Why are there hourly review limits if reviews are unlimited?

CodeRabbit does not cap how many pull requests you can open in a month. It does cap how many reviews each developer can run per hour (5 / 8 / 10 / 12 by plan) so a burst does not stampede the queue. You can keep going with usage-based reviews at $0.25 per file, with a monthly spend cap.

Does the CodeRabbit partner link add a discount?

No extra percent off. https://coderabbit.link/louis-dupple-com is a partner referral. You still pay the published Essentials / Team / Advanced prices.

What are the best CodeRabbit alternatives?

Greptile for whole-codebase graph context, GitHub Copilot if you already pay for completions and want PR review inside GitHub, Graphite if you stack diffs, Qodo if you want the PR checked against Jira or Linear, and Bito if you need GitHub, GitLab, and Bitbucket at the lowest seat price.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [CodeRabbit pricing](https://coderabbit.ai/pricing)checked Sep 2026
- [Greptile pricing](https://www.greptile.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Graphite pricing](https://graphite.com/pricing)checked Sep 2026
- [Bito pricing](https://bito.ai/pricing)checked Sep 2026

Related guides

Ai For Code ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Cursor Review

URL: https://devshot.co/reviews/cursor-review
Type: review
Published: 2026-08-04
Updated: 2026-09-25
Summary: Honest Cursor review for engineers: a VS Code fork with best-in-class Tab autocomplete, an autonomous Agent, model choice across Claude and GPT…

Review

## Cursor Review

The AI-first code editor most engineers reach for first. A VS Code fork with a genuinely great Tab model and an Agent that can rewrite half your repo, priced with usage limits that reward you for reading the fine print.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Cursor is an AI-first code editor built as a fork of VS Code, made by Anysphere, and it is the tool most engineers name first when the topic is AI coding. Its standout is the Tab model, a next-edit predictor that finishes multi-line changes and jumps to the next spot you were going to touch, plus an Agent (Composer) mode that plans and applies edits across many files and runs terminal commands. Pricing is Hobby (free), Pro at $20 a month, Pro+ at $60, Ultra at $200, and Teams at $40 per user a month. Since 2025 the paid plans run on a usage model: Auto mode is effectively unlimited, but manually picking a frontier model like Claude Sonnet or GPT draws down an included credit pool, and heavy users hit limits sooner than expected. The strengths are speed, whole-codebase context, and model choice; the catches are surprise usage limits on Pro, extension parity that lags real VS Code, and privacy questions for proprietary code. Closest alternatives: GitHub Copilot, Devin Desktop (formerly Windsurf), Claude Code, and Zed.

Key factsBest for: Developers and teams who want AI-first editing built into…

- Updated: September 25, 2026
- Best for: Developers and teams who want AI-first editing built into VS Code, with best-in-class autocomplete and a capable multi-file agent.
- Price as of September 25, 2026: From $20/month (Pro); free Hobby plan.

- The AI-first VS Code fork most engineers reach for first, with a standout Tab model and a full-repo Agent.
- Founded: 2022 (Anysphere)
- Headquarters: San Francisco
- Alternatives covered: GitHub Copilot, Devin Desktop (formerly Windsurf), Claude Code, Zed

Pros

- Tab (next-edit prediction) is the best autocomplete in the category and hard to give up once you rely on it.
- Agent, Cmd-K, and codebase indexing make whole-repo AI editing genuinely productive on real projects.
- Model choice across Claude, GPT, and Gemini, plus MCP and rules files, gives real control over the AI.

Cons

- Usage-based limits on Pro can surprise heavy Agent users, who exhaust the included pool before month end.
- As a VS Code fork it lags upstream, so some extensions and new VS Code features arrive late.
- Privacy for proprietary code and the cost of Ultra give some teams pause.

Founded2022 (Anysphere)

HeadquartersSan Francisco

Est. price$20/mo (Pro)

Best forAI-first developers

Ask a room of software engineers which AI coding tool they actually keep open, and Cursor is the name that comes up most. It won mindshare by making the AI feel like it is sitting inside your editor rather than bolted onto the side.

The pitch is simple: take VS Code, the editor most developers already live in, fork it, and rebuild the AI layer from the keystroke up.

The harder question for a buyer in 2026 is whether that experience is worth the money, and whether the usage-based pricing model Cursor rolled out will bite you mid-sprint. This review is for working developers and engineering leads deciding whether to standardize on Cursor.

We cover what it is under the hood, how Tab, Agent, and Cmd-K work, what it really costs, where it excels and frustrates, and five competitors worth trying first.

## What is Cursor?

Cursor is a desktop code editor built as a fork of Visual Studio Code, developed by Anysphere, a San Francisco company founded in 2022. Because it is a VS Code fork, it inherits the familiar interface, keybindings, themes, and most extensions, so the switching cost for a VS Code user is close to zero. What Anysphere rebuilt is the AI stack.

Four things carry the product, and the features section unpacks each: Tab, a next-edit prediction autocomplete; Agent (built on Cursor's Composer model), an autonomous mode that plans and edits across files while running terminal commands; Cmd-K, natural-language inline editing on a selection; and codebase indexing, which embeds your repository so models answer with whole-project context, not just the open file.

On top of that, Cursor gives you model choice: route work to frontier models from Anthropic (Claude), OpenAI (GPT), and Google (Gemini), to Cursor's own fast Composer model, or let Auto pick.

It supports the Model Context Protocol (MCP) for wiring in external tools and data, and rules files (.cursor/rules, formerly .cursorrules) for steering the AI per project.

## How Cursor works

A Cursor session starts the way any VS Code session does: open a folder and Cursor indexes it in the background. With that index in place, the AI answers with context from anywhere in the repo, and you @-mention files, symbols, docs, or git history to point the model at exactly what matters.

Most of the value shows up in Tab. As you type, Cursor predicts the rest of the line, the block, and often the next edit several lines away, so renaming a variable or threading a new argument through a function becomes a chain of tab presses.

It is the feature people miss most when they leave Cursor. For scoped changes you reach for Cmd-K: select code, describe the change, and accept or reject the inline diff.

For anything larger you open the Agent. Give it a task like "add pagination to the users endpoint," and it plans, edits multiple files, runs commands, reads the output, and fixes its own mistakes, all under a diff you review before accepting.

Behind it you choose the model or let Auto decide, set guardrails with rules files, and add MCP servers so the Agent can reach your database or docs.

The rough edges are real: the Agent can wander on vague prompts, large diffs need careful review, and because Cursor tracks VS Code releases on its own cadence, some extensions and new features arrive later than upstream.

## Cursor key features

Tab (next-edit prediction)Essential
Cursor's signature feature: a model that suggests multi-line edits and predicts the next place you were about to change, so refactors flow as a chain of tab presses. More context-aware than standard single-line completion, and the main reason developers stay.

Agent and Composer modeEssential
The autonomous mode plans a task, edits across multiple files, runs terminal commands, and iterates on test output, all under a diff you approve before it lands. Composer is Cursor's own fast agent model, and you can swap in Claude or GPT for harder reasoning.

Cmd-K inline editingEssential
Select any block of code, describe the change in natural language, and get an inline diff to accept or reject. Cmd-K is the precise, low-risk counterpart to the Agent: ideal for a test, a refactor, or a bug fix without letting the AI touch the rest of the repo.

Codebase indexing and context
Cursor embeds your repository for semantic search, so answers and edits reflect the whole project, not just the open file. Steer context precisely by @-mentioning files, symbols, docs, or git history, which is what makes the AI useful on real, large codebases rather than toy examples.

Model choice and MCP
Route work to frontier models from Anthropic, OpenAI, and Google, to Cursor's own Composer, or to Auto mode, all from one dropdown. MCP support connects external tools and data (databases, issue trackers, docs) so the Agent acts on live context beyond your code.

Rules files and privacy mode
Project rules files (.cursor/rules) encode your conventions so the AI follows house style, framework choices, and do-not-touch zones. Privacy mode promises code is not stored or used for training, though security-strict teams should still read the terms first.

## Cursor pricing

Cursor, GitHub Copilot, Devin Desktop, and Zed all publish list prices, and each includes a free tier. Claude Code is included with Claude Pro or Claude Max, or billed pay-per-token on the Anthropic API. Enterprise is a custom quote at Cursor and Devin, while GitHub Copilot publishes an Enterprise price.

The cheapest credible paid entry is GitHub Copilot Pro. Costs jump on the top individual plans, especially Cursor Ultra, Devin Max, and the higher Claude Max price, all around $200 a month.

Since 2025, Cursor paid plans include a usage pool roughly equal to the plan price: Auto mode is effectively unlimited, frontier models spend the pool, and overages are optional.

Plan | Price | Best for |

Cursor Pro | $20/month | Included usage pool, optional overages |

Cursor Pro+ | $60/month | About 3x Pro Agent limits |

Cursor Ultra | $200/month | About 20x Pro limits, priority features |

Cursor Teams | $40/user/month | SSO (SAML/OIDC), admin, privacy, analytics |

Cursor Enterprise | Custom quote | SCIM, access controls, audit logs, pooled usage |

GitHub Copilot Pro | $10/month | Individual plan, AI-credit (token) billing |

GitHub Copilot Pro+ | $39/month | Higher individual plan, credit billing |

GitHub Copilot Business | $19/user/month | Per-user plan, AI-credit billing |

GitHub Copilot Enterprise | $39/user/month | Published enterprise price, credit billing |

Devin Desktop Pro | $20/month | Higher quotas that refresh daily and weekly, with frontier models |

Devin Team | $40 per full seat, $80/month minimum | Up to 200 users; two full seats cover the minimum, each with its own quota |

Devin Enterprise | Custom quote | Quote-only enterprise plan |

Claude Code on Claude Pro | $20/month | Includes Claude Code with the subscription |

Claude Code on Claude Max | $100 or $200/month | Included with Claude Max |

Claude Code API | Pay per token | Usage billing via the Anthropic API |

Zed Pro | $10/month | Unlimited edit predictions and $5 of hosted-model tokens |

Zed Business | $30/seat/month | Org-wide AI policies, data governance, unified billing |

Zed Personal | Free | 2,000 accepted edit predictions, unlimited use with your own keys |

## Cursor pros and cons

### What we like

- Tab (next-edit prediction) is the best autocomplete in the category and hard to give up once you rely on it.
- Agent, Cmd-K, and codebase indexing make whole-repo AI editing genuinely productive on real projects.
- Model choice across Claude, GPT, and Gemini, plus MCP and rules files, gives real control over the AI.

### What could be better

- Usage-based limits on Pro can surprise heavy Agent users, who exhaust the included pool before month end.
- As a VS Code fork it lags upstream, so some extensions and new VS Code features arrive late.
- Privacy for proprietary code and the cost of Ultra give some teams pause.

## Who Cursor is for

Cursor is the strongest default for developers who want AI woven into every keystroke, not parked in a sidebar. If you already use VS Code, the migration is trivial, and the Tab model alone changes how fast you move through routine edits.

Individual engineers, startups, and anyone doing greenfield or fast-iteration work get the most out of it, especially those comfortable choosing models and writing rules files to keep the AI on the rails.

It is a weaker fit in a few cases. Cost-sensitive solo developers who mostly want good autocomplete may find GitHub Copilot covers them for less, without usage anxiety. Teams in regulated or IP-strict environments should scrutinize privacy mode and the data terms of every model provider it calls.

Developers who need the latest VS Code features the moment they ship may be annoyed by the fork lagging upstream, and terminal-first engineers may get more from a command-line agent like Claude Code.

## Best Cursor alternatives

If Cursor is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Cursor | Developers and teams who want AI-first editing built into VS Code, with best-in-class autocomplete and a capable multi-file agent. | From $20/month (Pro) | Visit → |

GitHub Copilot | Developers who want affordable, reliable AI assistance tightly integrated with GitHub and standard VS Code. | From $10/month (Pro) | Visit → |

Devin Desktop (formerly Windsurf) | Developers who want a clean, agent-first AI editor with a simpler flow than Cursor's model dropdowns. | From $20/month (Pro) | Visit → |

Claude Code | Terminal-first engineers who want a powerful agentic coding assistant driven from the command line. | From $20/month (included with Claude Pro) | Visit → |

Zed | Developers who want a fast native editor and prefer to bring their own model keys or agents. | From $10/month (Pro) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

GitHub Copilot

The incumbent AI pair programmer, cheap and deeply tied to GitHub and the real VS Code.

Visit →

Devin Desktop (formerly Windsurf)

Cognition's agent-first AI editor, sold as Windsurf until its rename and as Codeium before that.

Visit →

Claude Code

Anthropic's command-line coding agent, strong on deep reasoning and large, multi-step tasks.

Visit →

Zed

A fast, open-source native editor with edit predictions, hosted models on Pro, and support for external agents.

Visit →

## The bottom line

Cursor earns its status as the default AI code editor. The Tab model is genuinely the best autocomplete in the category, the Agent and Cmd-K make whole-repo edits practical, and as a VS Code fork it costs a VS Code user almost nothing to adopt.

For an individual or a fast-moving team, Pro at $20 a month is an easy yes, and model choice, MCP, and rules files give power users real control.

The trade-offs are the usage model and the fork tax: manually driving frontier models eats your included usage faster than expected, Ultra at $200 a month is a real budget line, extensions can lag upstream, and IP-strict teams need to vet privacy mode.

Buy Cursor if you want the most complete AI-first editing experience and will spend a few minutes managing usage. If you want the cheapest reliable option, GitHub Copilot wins; for a smoother agent flow, try Devin Desktop; if you live in the terminal, Claude Code is stronger; and if you want a fast native editor that runs on your own API keys, Zed is the lighter pick.

## Frequently asked questions

How much does Cursor cost?

Cursor has a free Hobby plan and four paid options: Pro at $20 a month, Pro+ at $60, Ultra at $200, and Teams at $40 per user a month, plus custom Enterprise. Each paid plan includes a usage pool roughly equal to its price. Auto mode is effectively unlimited, but manually selecting frontier models like Claude or GPT draws down that pool, after which you wait for the monthly reset or pay usage-based overages.

Is Cursor better than GitHub Copilot?

For many developers, yes on experience: Cursor's Tab and Agent tend to feel faster and more capable than Copilot's, and the whole editor is built around the AI. Copilot wins on price ($10 versus $20) and on running inside the real VS Code and JetBrains with full extension parity, since Cursor is a fork that can lag upstream. Want the best editing experience? Pick Cursor. Want cheap assistance tied to GitHub? Pick Copilot.

Which AI models does Cursor support?

Cursor lets you choose the model per task: frontier models from Anthropic (Claude Sonnet and Opus), OpenAI (GPT), and Google (Gemini), Cursor's own fast Composer model, or Auto mode, which picks for you. Auto is effectively unlimited on paid plans, while manually chosen frontier models consume your included usage. You can also connect external tools and data through the Model Context Protocol (MCP).

Is Cursor safe to use with proprietary code?

Cursor offers a privacy mode that, when enabled, promises your code is not stored or used to train models, and Teams and Enterprise extend this with team-wide privacy and admin controls. That said, Cursor still sends context to third-party model providers to function, so regulated teams should read the data terms carefully. Teams that must keep code off third-party AI entirely can pair a plain editor such as Zed with local models, at the cost of weaker agent quality.

What are the best Cursor alternatives?

The closest alternatives are GitHub Copilot, the cheaper and more widely integrated incumbent; Devin Desktop, Cognition's agent-first editor (formerly Windsurf, and before that Codeium); Claude Code, Anthropic's command-line agent for terminal-first workflows; and Zed, a fast native editor with bring-your-own-key AI. Which one wins depends on your budget, workflow, and privacy requirements.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Devin Desktop pricing](https://cognition-labs.com)checked Sep 2026
- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Zed pricing](https://zed.dev/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsGithub Copilot ReviewCursor Student Discount 2026: Is Cursor Still Free for Students?AI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Datadog Review

URL: https://devshot.co/reviews/datadog-review
Type: review
Published: 2026-08-04
Updated: 2026-09-25
Summary: Honest Datadog review for SRE and DevOps teams: real per-host and per-GB pricing, where the single-agent platform shines, where the cost bites…

Review

## Datadog Review

The most complete observability platform for infrastructure, APM, logs, RUM, and synthetics, with one agent and more than 1,000 integrations. Genuinely powerful, genuinely priced per product, and the bill can climb fast.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Datadog is the most complete observability platform on the market, covering infrastructure, APM, logs, real user monitoring, synthetics, and security through a single agent and more than 1,000 integrations. Pricing is per product and per host or per GB: Infrastructure starts at $15 per host per month billed annually, APM at $31 per host, and Log Management at $0.10 per ingested GB plus $1.70 per million indexed events, with a free tier that covers up to 5 hosts. The biggest strength is breadth and depth in one pane of glass, with excellent dashboards, monitors, and Watchdog AI on top. The biggest catch is cost: the bill can spike unpredictably as log and custom-metric volume grows, and the per-product SKU sprawl makes forecasting hard. Cost is the number-one complaint from customers. The closest alternatives are New Relic, Grafana Cloud, Dynatrace, and Splunk.

Key factsBest for: SRE and DevOps teams that want one correlated platform…

- Updated: September 25, 2026
- Best for: SRE and DevOps teams that want one correlated platform across infrastructure, APM, logs, RUM, and synthetics.
- Price as of September 25, 2026: From $15/host/mo (Infrastructure Pro, billed annually); free for up to 5 hosts; 14-day trial

- The most complete observability platform, with a single agent and more than 1,000 integrations, priced per product.
- Founded: 2010
- Headquarters: New York, USA
- Alternatives covered: New Relic, Grafana Cloud, Dynatrace, Splunk

Pros

- Broadest and deepest feature set in the category, all correlated in one platform.
- Single agent plus 1,000+ integrations makes rollout fast across almost any stack.
- Excellent dashboards, monitors, and Watchdog AI for anomaly detection.

Cons

- The bill can spike unpredictably from log and custom-metric volume.
- Per-product SKU pricing is complex to forecast and easy to overspend on.
- Cost is consistently the number-one complaint from customers.

Founded2010

HeadquartersNew York, USA

Est. price$15 to 31/host/mo

Best forSRE and DevOps teams

Datadog is the default answer when an engineering team asks what to run for monitoring, and for good reason.

It sits at the center of the observability category, next to New Relic, Dynatrace, Grafana, and Splunk, and it has grown into a platform that touches almost every layer of a modern stack: hosts, containers, serverless functions, application traces, logs, frontend sessions, uptime checks, and security signals.

For an SRE or DevOps engineer, the real question is not whether Datadog can see your systems, because it can. It is whether the breadth justifies a bill that is famous for growing faster than anyone plans for.

This review is written for engineers and platform teams evaluating Datadog as their primary observability tool. We cover what the platform actually includes, how the single agent and integrations work day to day, what each product really costs, where Datadog is genuinely best in class, and where the pricing model and complexity work against you. We also name five alternatives worth pricing before you commit to an annual contract.

## What is Datadog?

Datadog is a cloud-based observability and security platform, founded in 2010 and headquartered in New York, that unifies monitoring across infrastructure, applications, and user experience. It describes itself as AI-powered observability and security, and in practice it is a set of tightly integrated products sold as separate SKUs.

The foundation is Infrastructure Monitoring, which collects metrics from hosts, containers, Kubernetes, and cloud services.

On top of that sit APM and distributed tracing with continuous profiling, Log Management with a split between ingest and index, Real User Monitoring and Session Replay for the frontend, and Synthetic Monitoring for API and browser uptime checks. Around those run Database Monitoring, Network Monitoring, Cloud Security, and more.

What ties it all together is the single Datadog Agent, one lightweight process you install once that reports metrics, traces, and logs, plus more than 1,000 out-of-the-box integrations (Datadog passed the 1,000 integration milestone in late 2025).

Watchdog AI, dashboards, monitors, and Incident Management sit on top of that shared data layer.

## How Datadog works

Getting started is genuinely fast. You install the Datadog Agent on a host, in a container sidecar, or as a Kubernetes DaemonSet, drop in your API key, and metrics start flowing within minutes.

Turning on an integration is usually a config file or a checkbox, and the agent auto-discovers services like Postgres, Nginx, Redis, or a cloud provider without much manual wiring.

Day to day, engineers live in dashboards, the metrics explorer, and the trace and log search. You build monitors with thresholds or anomaly detection, route alerts to Slack, PagerDuty, or Datadog On-Call, and pivot from a spiking dashboard straight into the traces and logs behind it.

Watchdog surfaces anomalies you did not think to alert on, and Bits AI can chat, investigate, and suggest remediation.

The rough edges show up in cost governance, not capability. Custom metrics, high-cardinality tags, and log volume can quietly balloon your bill, so teams end up building index filters, exclusion rules, and usage dashboards just to keep spend in check.

The learning curve to use Datadog well, especially around what to index and what to drop, is real.

## Datadog key features

Single agent and 1,000+ integrationsEssential
One lightweight Datadog Agent collects metrics, traces, and logs, and more than 1,000 out-of-the-box integrations cover cloud providers, databases, queues, and CI tools. This is why Datadog is fast to roll out: install once, enable checks, and most of your stack reports in with almost no custom code.

Infrastructure monitoring and dashboardsEssential
The core product tracks hosts, containers, Kubernetes, and serverless with tens of thousands of out-of-the-box metrics. Drag-and-drop dashboards, tag-based filtering, and flexible monitors (threshold, anomaly, forecast) make it the reference tool for seeing an entire fleet in one place.

APM and distributed tracingEssential
Application Performance Monitoring instruments services for end-to-end traces, service maps, and continuous profiling. You can follow a slow request across microservices, spot the bottleneck function, and jump from a trace to the exact log line, which is where Datadog earns its keep for SRE work.

Log Management with ingest and index split
Logs are priced in two parts: cheap ingest for everything, and paid indexing only for what you want to search and retain. Live Tail, index filters, and Flex Logs let you keep high volume affordable, but this split is also the single biggest source of surprise bills if left unmanaged.

RUM, Session Replay, and Synthetics
Digital Experience Monitoring covers the frontend: Real User Monitoring captures real sessions and errors, Session Replay records what users saw, and Synthetic Monitoring runs scheduled API and browser tests for uptime and critical flows. Together they close the loop from backend traces to actual user impact.

Watchdog AI, on-call, and incident management
Watchdog surfaces anomalies automatically, Bits AI investigates and suggests fixes, and built-in On-Call and Incident Management handle paging, escalation, and postmortems inside the same platform. It reduces the number of separate tools a team needs, though the deepest AI and automation features sit on higher tiers.

## Datadog pricing

Datadog publishes a per-product price list, billed per host or per GB. New Relic, Grafana Cloud, Dynatrace, and Splunk also have stated usage rates, so none of these five is quote-only. The cheapest credible entry is Grafana Cloud's free tier, then Pro at a $19 monthly platform fee plus $6.50 per 1,000 metric series.

Datadog's free tier covers 5 hosts, and Infrastructure Pro is $15 per host per month billed annually. On Datadog, annual rates sit below on-demand, and the bill is the sum of separate products.

Costs jump on indexed logs, APM, RUM, custom metrics, and high log volume, and again on New Relic seats, Splunk's App and Infra tier, and Dynatrace full-stack for large enterprises.

Plan | Price | Best for |

Datadog Free | $0 (up to 5 hosts) | Core metrics, 1-day retention |

Datadog Infrastructure Pro | $15/host/mo annual (about $18 on-demand) | Metrics, dashboards, 15-mo retention |

Datadog Infrastructure Enterprise | $23/host | Enterprise infrastructure plan |

Datadog APM | $31/host/mo annual, with Infrastructure | Bundled with Infrastructure Monitoring |

Datadog APM standalone | Roughly $36/host/mo | APM without the infrastructure bundle |

Datadog APM Enterprise | Up to $40/host/mo | Highest published APM tier |

Datadog APM on-demand | $36/host/mo | On-demand rate for APM with Infrastructure |

Datadog Log Management | $0.10/GB ingest or scan + $1.70/M events | 15-day retention on indexed events |

Datadog Log index on-demand | About $2.55/M indexed events | On-demand rate for indexed events |

Datadog Flex Logs | $0.05 to $0.60/M events | Cheaper storage for high log volume |

Datadog RUM | $0.15/1,000 sessions (Measure) | Full traffic; RUM Investigate is $3 per 1,000 filtered sessions |

Datadog Session Replay | $2.50/1,000 sessions | Session Replay on top of RUM |

Datadog Synthetic API | $5/10,000 API runs | Scheduled API test runs |

Datadog Synthetic browser | $12/1,000 browser runs | Scheduled browser test runs |

New Relic ingest | $0.40/GB ($0.60 on Data Plus) | After 100 GB/mo and 1 full-platform user free |

New Relic full-platform user | $99 to $349/user/mo | Standard: $10 first user, $99 each after (max 5); Pro $349 annual |

Grafana Cloud Free | Free | 10k series, 50 GB logs, 50 GB traces, 14-day retention |

Grafana Cloud Pro metrics | $6.50/1,000 series/mo | Plus a $19 monthly platform fee; 10k series included |

Grafana Cloud Pro logs and traces | $0.55/GB | $0.05 process, $0.40 write, $0.10 retain per GB; 50 GB included |

Dynatrace Infrastructure | About $0.04/host/hour | Infrastructure-only monitoring |

Dynatrace Full-stack | About $0.08/host/hour | Full-stack plus Davis Data Units for logs and events |

Dynatrace Full-stack (8 GB host) | Roughly $58/mo per 8 GB host | Monthly equivalent of the hourly full-stack rate |

Splunk Observability infrastructure | $15/host/mo annual | Infrastructure on Observability Cloud |

Splunk Observability App and Infra | $60/host/mo annual | APM plus infrastructure; End-to-End tier from $75 |

Splunk log platform | Custom quote | Ingest- or workload-based pricing through sales |

## Datadog pros and cons

### What we like

- Broadest and deepest feature set in the category, all correlated in one platform.
- Single agent plus 1,000+ integrations makes rollout fast across almost any stack.
- Excellent dashboards, monitors, and Watchdog AI for anomaly detection.

### What could be better

- The bill can spike unpredictably from log and custom-metric volume.
- Per-product SKU pricing is complex to forecast and easy to overspend on.
- Cost is consistently the number-one complaint from customers.

## Who Datadog is for

Datadog is a strong fit for teams that run real production systems at scale, especially SRE, DevOps, and platform engineers who want infrastructure, APM, logs, RUM, and synthetics in one correlated place rather than five disconnected tools.

Cloud-native shops on Kubernetes, companies with sprawling microservices, and teams that value fast time-to-value from a single agent get the most out of it.

It is a weaker fit in a few clear cases. Very small teams or side projects will find the per-host and per-product model expensive for what they need, and a generous free tier elsewhere may cover them.

Teams that are all-in on open-source telemetry (Prometheus, Loki, OpenTelemetry) and want to avoid vendor lock-in should weigh Grafana Cloud first. And cost-sensitive organizations with high log or custom-metric volume need to model the bill carefully, because that is exactly where Datadog gets expensive fastest.

## Best Datadog alternatives

If Datadog is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Datadog | SRE and DevOps teams that want one correlated platform across infrastructure, APM, logs, RUM, and synthetics. | From $15/host/mo (Infrastructure Pro, billed annually) | Visit → |

New Relic | Teams that want all-in-one observability priced on data ingest and seats rather than per host. | $0.40/GB after 100 GB free a month ($0.60 on Data Plus) | Visit → |

Grafana Cloud | Teams standardized on open source (Prometheus, Loki, Tempo, OpenTelemetry) that want managed hosting without lock-in. | Pro from $19/mo plus $6.50 per 1,000 series | Visit → |

Dynatrace | Large enterprises that want automatic, AI-driven root-cause analysis across complex environments. | From about $0.04/host/hour (infrastructure) | Visit → |

Splunk | Security and ops teams that already run Splunk for log analytics and SIEM at scale. | From $15/host/mo for Infrastructure (annual) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 5 do not publish a comparable monthly price and are left out rather than estimated.

New Relic

An all-in-one observability platform billed on data ingest plus per-user seats.

Visit →

Grafana Cloud

Managed Grafana, Prometheus, Loki, and Tempo with the most generous free tier in the category.

Visit →

Dynatrace

An enterprise observability suite built around automatic discovery and the Davis AI causal engine.

Visit →

Splunk

A log-analytics and observability heavyweight, strongest where Splunk is already the system of record.

Visit →

## The bottom line

Datadog earns its reputation as the most complete observability platform. For an SRE or DevOps team running real production systems, having infrastructure, APM, logs, RUM, and synthetics correlated in one place, fed by a single agent and more than 1,000 integrations, is a genuine advantage.

Dashboards, monitors, Watchdog AI, and built-in incident response are best in class, and rollout is fast.

The trade-off is cost and complexity. Per-product SKUs, custom metrics, and log volume can push the bill well past what you budgeted, and cost is consistently the loudest complaint.

Buy Datadog if breadth and correlation are worth paying a premium for, and if you will invest in governing usage. If you want ingest-based pricing, look at New Relic; if you want open source without lock-in, start with Grafana Cloud; if you want automatic AI root-cause at enterprise scale, weigh Dynatrace; and if Splunk already owns your logs, Splunk Observability is the natural extension.

## Frequently asked questions

How much does Datadog cost?

Datadog publishes per-product pricing. Infrastructure Monitoring is $15 per host per month on the Pro plan (billed annually) and $23 per host on Enterprise. APM starts at $31 per host per month when bundled with Infrastructure. Log Management charges $0.10 per ingested GB plus $1.70 per million events indexed with 15-day retention. RUM Measure starts at $0.15 per 1,000 sessions (RUM Investigate at $3 per 1,000 filtered sessions), and Synthetics runs $5 per 10,000 API test runs. There is a free tier for up to 5 hosts, and on-demand usage above your commitment is billed at higher rates.

Is Datadog worth it?

For teams that run production systems at scale and want one correlated view of infrastructure, applications, and user experience, yes. Datadog's breadth, single agent, integrations, and AI features are the best in the category. It is less worth it for small teams, side projects, or very cost-sensitive shops with heavy log volume, where the per-host and per-product model gets expensive quickly and a generous free tier elsewhere may cover you.

Why is my Datadog bill so high?

The most common causes are custom metrics, high-cardinality tags, and log volume. Datadog bills custom metrics per active timeseries, so a few high-cardinality tags can multiply into thousands of metrics. Logs are cheap to ingest but cost $1.70 per million events to index, so indexing everything adds up fast. APM and RUM add per-host and per-session charges on top. The fix is to use index filters, exclusion rules, and usage dashboards, and to only index and retain what you actually search.

What are the best Datadog alternatives?

New Relic is the closest all-in-one competitor, priced on data ingest plus per-user seats. Grafana Cloud is the pick for teams that want managed open source (Prometheus, Loki, Tempo, OpenTelemetry) with the most generous free tier and lowest paid pricing. Dynatrace suits large enterprises that want automatic, AI-driven root-cause analysis. Splunk Observability makes sense if your organization already runs Splunk for logs and SIEM.

Does Datadog have a free tier?

Yes. Datadog offers a free tier for Infrastructure Monitoring that covers up to 5 hosts with 1-day metric retention, plus limited free allowances on several other products. It is enough to trial the platform or monitor a small environment, but production use across APM, logs, RUM, and synthetics quickly moves you onto paid, per-product plans. There is also a 14-day trial of the full platform if you want to test the paid features before committing.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Datadog pricing](https://datadoghq.com/pricing)checked Sep 2026
- [New Relic pricing](https://newrelic.com/pricing)checked Sep 2026
- [Grafana Cloud pricing](https://grafana.com/pricing)checked Sep 2026
- [Dynatrace pricing](https://dynatrace.com/pricing)checked Sep 2026
- [Splunk pricing](https://splunk.com/pricing)checked Sep 2026

Related guides

Ai For DevopsSelf Hosting ToolsDatadog Alternatives 2026: Cheaper Monitoring Compared at 10 HostsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Dify Review

URL: https://devshot.co/reviews/dify-review
Type: review
Published: 2026-07-18
Updated: 2026-09-25
Summary: Our Dify review covers pricing, the visual workflow builder, RAG, real weaknesses, and the alternatives (Flowise, Langflow, n8n) worth comparing.

Review

## Dify Review

An open-source platform for building LLM apps, agents, and RAG workflows without wiring the plumbing yourself. Built for dev teams who want to ship fast.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Dify is one of the strongest open-source platforms for building LLM apps, and for most teams the fastest way from idea to a deployed chatbot, agent, or RAG workflow without stitching together a framework yourself. The visual Workflow Studio plus a built-in knowledge pipeline is the real draw.

Cloud pricing is honest and cheap: a free Sandbox, then Professional at $59/mo, or $590/year, and Team at $159/mo, or $1,590/year. The catch is the message-credit ceiling, which real traffic burns through fast, pushing you toward bring-your-own-key or self-hosting. If you want a lighter code-first path, Flowise is the alternative to weigh first.

Key factsBest for: Development teams shipping LLM apps, agents, and RAG…

- Updated: September 25, 2026
- Best for: Development teams shipping LLM apps, agents, and RAG workflows who can self-host later
- Price as of September 25, 2026: From $59/mo (Professional), or $590/yr billed annually; free Sandbox, free to self-host

- Open-source platform for building and operating LLM apps with a visual workflow builder and built-in RAG.
- Founded: 2023
- Free plan: Yes, Sandbox tier
- Open source: 149K+ GitHub stars
- Alternatives covered: Flowise, Langflow, n8n, Botpress, LangChain (LangSmith)

Pros

- Fastest path from idea to a deployed chatbot, agent, or RAG app without writing orchestration code
- Model-agnostic across 50-plus providers plus local Ollama, so no vendor lock-in
- Free open-source self-hosting lets you escape cloud credit caps and keep data in-house

Cons

- Message-credit model on hosted keys runs out fast; real traffic forces bring-your-own-key or self-hosting
- Visual canvas gets cluttered and hard to version-control as workflows grow complex
- The open-source license restricts reselling Dify as a multi-tenant SaaS and removing its branding

Founded2023

Free planYes, Sandbox tier

Starting price$59/mo, or $590/yr

Open source149K+ GitHub stars

Every engineering team that touches LLMs hits the same fork: build the whole app in code with a framework like LangChain, or reach for a platform that hands you the workflow engine, RAG, and observability out of the box.

Dify sits squarely in that second camp, and with 150K-plus GitHub stars it has become the default name people mention when the phrase "open-source LLM app platform" comes up.

The real question is not whether Dify can build a demo. It can, in an afternoon. The question is whether it holds up when you move past the demo: does the visual builder still make sense for complex logic, do the pricing limits bite, and can you get out cleanly if you outgrow the cloud.

This review is written from that production angle, weighing the free Sandbox, the paid cloud tiers, the self-hosted Community edition, and how Dify stacks up against Flowise, Langflow, n8n, and a raw LangChain build.

## What is Dify?

Dify is an open-source platform for developing and operating LLM applications, built by LangGenius, Inc. Think of it as the layer between raw model APIs and a shipped product: instead of writing orchestration, retrieval, and logging code, you assemble those pieces in a web console.

It covers four app types: chatbots, autonomous agents, multi-step workflows, and simple text-generation apps.

The centerpiece is the visual Workflow Studio, a node canvas where you chain LLM calls, conditionals, code blocks, HTTP requests, and tool invocations into a traceable execution path.

Around it sit the pieces you would otherwise build yourself: a Knowledge Pipeline for RAG that extracts, cleans, chunks, and indexes documents from files, websites, and drives; a Prompt IDE for iterating on prompts against multiple models side by side; and an agent engine with tool use, memory, and function calling.

Dify is model-agnostic, supporting OpenAI, Anthropic, Google, Mistral, and dozens of others plus local models through Ollama, so you are not locked to one provider. A Marketplace adds plugins, data sources, and MCP integrations.

Every app publishes as a hosted web app, an embeddable widget, a REST API, or an MCP-compatible tool, with logging, annotation, and feedback baked in.

## How Dify works

Getting started on Dify Cloud takes minutes: sign in, and you land on a workspace with app templates. Pick "Chatflow" or "Workflow" and you are dropped into the canvas.

The drag-and-connect model is genuinely fast for straightforward flows, a start node, an LLM node, a knowledge-retrieval node, an answer node, and you have a working RAG chatbot you can test in the built-in preview pane on the right.

Day to day, the loop is tight. You tweak a prompt, hit run, and watch the trace show each node's input, output, tokens, and latency. Adding a knowledge base is a matter of uploading files or pointing at a URL, choosing a chunking strategy, and picking an embedding model.

Connecting tools, a web search, a code interpreter, a custom API, happens through the same node interface or the Marketplace.

The rough edges show up as complexity grows. The visual canvas that felt clean at five nodes gets crowded at thirty, and branching logic can be fiddly to read. Version control is weaker than a code repo, diffing changes is awkward, and some teams end up managing prompts and configs outside the tool.

The API is solid, but deep customization eventually pushes you toward the self-hosted edition.

## Dify key features

Visual Workflow StudioEssential
A node-based canvas for chaining LLM calls, conditionals, code, HTTP requests, and tool calls into one traceable pipeline. You see each step's input and output, which makes debugging multi-step logic far easier than reading framework stack traces.

Knowledge Pipeline (RAG)Essential
A repeatable ingestion pipeline that extracts, cleans, chunks, and indexes content from files, websites, and connected drives, then wires it into any app. Supports multiple chunking strategies and embedding models so retrieval quality is tunable.

Agent engine with tools and MCPEssential
Build autonomous agents with tool use, memory, and function calling, extended by a Marketplace of plugins, data sources, and MCP integrations you can reuse across apps without rebuilding each one.

Prompt IDE and multi-model support
Iterate prompts against several models at once, comparing OpenAI, Anthropic, Google, Mistral, and local Ollama outputs side by side. Model-agnostic design means no lock-in and easy cost or quality swaps.

Built-in LLMOps
Every app ships with request logging, human annotation, and end-user feedback capture, plus token and latency tracking per node, giving you observability without bolting on a separate LLMOps tool.

Open-source self-hosting
The full platform is open source and runs via Docker, so you can self-host on your own infrastructure for data control or unlimited volume when the cloud message-credit caps start to get tight.

## Dify pricing

Dify, Flowise, n8n, Botpress, and LangSmith publish cloud prices, and Dify, Flowise, and n8n also offer a free self-hosted edition. Dify Professional is $59 a month, or $590 a year billed annually, with no API rate limit, and Team is $159 a month, or $1,590 a year, on top of a free Sandbox.

The cheapest published cloud entry is n8n Starter at about $24 a month, ahead of Flowise Starter at about $35 a month and LangSmith Plus at about $39 per seat per month.

n8n Pro is about $60 a month and Flowise Pro is about $65 a month, while Botpress jumps from Plus at $150 a month to Team at $750 a month, both billed annually. Enterprise is a custom quote for Dify, Flowise, n8n, Botpress, and LangSmith, and Langflow's DataStax plan stays usage-based after a free tier with no published list price.

Botpress's free plan stops at 25 conversations a month, and on Dify the plan fee is only part of the bill once message credits run out and you bring your own model key.

Plan | Price | Best for |

Dify Sandbox | Free | 1 member, 200 message credits, 5 apps, 50 documents, 30-day logs |

Dify Professional | $59/mo, or $590/yr | 3 members, 5,000 credits/mo, 50 apps, 500 documents, unlimited log history |

Dify Team | $159/mo, or $1,590/yr | 50 members, 10,000 credits/mo, 200 apps, 20GB storage |

Dify Community | Free to self-host | Open source edition, run it yourself |

Dify Enterprise | Custom quote | SSO, SAML, RBAC, and SOC 2 Type II |

Flowise self-host | Free | Open source, free to self-host |

Flowise Starter | ~$35/mo | Entry Flowise Cloud plan |

Flowise Pro | ~$65/mo | Higher Flowise Cloud plan |

Flowise Enterprise | Custom quote | Enterprise plan at a custom price |

Langflow managed (DataStax) | Usage-based | Free tier, then usage-based paid plans |

n8n Community | Free to self-host | Self-hosted community edition |

n8n Starter | ~$24/mo | Entry n8n Cloud plan |

n8n Pro | ~$60/mo | Higher n8n Cloud plan |

n8n Enterprise | Custom quote | Enterprise plan at a custom price |

Botpress Free | Free | 25 conversations, 3 seats and 3 AI agents, no overages |

Botpress Plus | $150/mo, billed annually | 250 conversations a month with $25 of AI usage included |

Botpress Team | $750/mo, billed annually | 1,500 conversations a month, unlimited seats |

Botpress Enterprise | Custom quote | Enterprise plan at a custom price |

LangChain open source | Free | Free open-source framework |

LangSmith Developer | Free | Free LangSmith developer tier |

LangSmith Plus | ~$39/seat/mo | Paid per seat each month |

LangSmith Enterprise | Custom quote | Enterprise plan at a custom price |

## Dify pros and cons

### What we like

- Fastest path from idea to a deployed chatbot, agent, or RAG app without writing orchestration code
- Model-agnostic across 50-plus providers plus local Ollama, so no vendor lock-in
- Free open-source self-hosting lets you escape cloud credit caps and keep data in-house

### What could be better

- Message-credit model on hosted keys runs out fast; real traffic forces bring-your-own-key or self-hosting
- Visual canvas gets cluttered and hard to version-control as workflows grow complex
- The open-source license restricts reselling Dify as a multi-tenant SaaS and removing its branding

## Who Dify is for

Dify is a strong fit for engineering teams that want to ship LLM features fast without owning the orchestration stack. Startups building an internal RAG chatbot, a customer-support agent, or a document-Q&A tool will get a working prototype in a day and a production deploy in a week.

It also suits teams that want an open-source escape hatch: build on the cloud, self-host later, keep your data in your own VPC.

It is a weaker fit in two cases. If your product is a deeply custom agent with unusual control flow, you will eventually fight the visual builder and be better served by a code-first stack like LangChain with LangSmith, where everything lives in your repo.

And if you are a non-technical team that just wants a support bot with a polished chat widget and CRM integrations, Botpress or a purpose-built chatbot tool will feel more finished than Dify's developer-oriented console.

Solo builders on a tight budget can lean on the free Sandbox or self-host indefinitely, but should expect to manage infrastructure themselves.

## Best Dify alternatives

If Dify is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Dify | Development teams shipping LLM apps, agents, and RAG workflows who can self-host later | From $59/mo (Professional), or $590/yr billed annually | Visit → |

Flowise | Developers who want a lightweight, LangChain-based visual builder they fully control | From ~$35/mo (Starter) | Visit → |

Langflow | Teams prototyping RAG and agent flows who want a Python open-source visual builder | Free to self-host | Visit → |

n8n | Teams wiring AI steps into broader automation across hundreds of apps | From ~$24/mo (Starter) | Visit → |

Botpress | Teams building polished customer-facing support and chat agents | From $150/mo (Plus, billed annually) | Visit → |

LangChain (LangSmith) | Engineering teams that want full control in code, with everything in their repository | From ~$39/seat/mo (LangSmith Plus) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Flowise

Open-source drag-and-drop tool for building LLM flows and agents on top of LangChain.

Visit →

Langflow

Open-source visual builder for RAG and agent pipelines, backed by DataStax and IBM.

Visit →

n8n

Source-available workflow automation platform with AI agent nodes and 400-plus integrations.

Visit →

Botpress

Chatbot and AI agent platform focused on conversational customer experiences and channel deployment.

Visit →

LangChain (LangSmith)

Code-first framework plus LangSmith observability for building custom LLM apps and agents.

Visit →

## The bottom line

Dify earns its reputation. For a dev team that wants to ship an LLM chatbot, agent, or RAG app without spending weeks on orchestration and observability code, it is one of the best starting points available, and the free Sandbox plus $590/year Professional tier makes it cheap to prove out.

The open-source Community edition is the safety net that makes committing low-risk: outgrow the cloud limits and you self-host.

Pick something else when your needs pull to an extreme. Choose LangChain with LangSmith if your agent logic is genuinely custom and you want everything in code with real version control.

Choose Botpress if you are building a polished customer-support bot more than a developer tool, and choose n8n if the AI is one step inside a much larger automation. Flowise and Langflow are the closest like-for-like open-source rivals, worth a weekend of comparison if the visual-builder approach appeals. For most teams, though, Dify is the one to try first.

## Frequently asked questions

How much does Dify cost?

Dify Cloud is free on the Sandbox tier (200 message credits, 5 apps). Professional is $59 a month, or $590 a year billed annually, for 5,000 monthly credits and 50 apps, and Team is $159 a month, or $1,590 a year, for 10,000 credits and 50 members. The Community edition is free to self-host, and Enterprise is custom-priced.

Is Dify worth it?

For teams shipping LLM apps, yes. It replaces weeks of orchestration, RAG, and logging code with a visual console, and the open-source edition removes lock-in risk. It is less worth it if your agent needs highly custom control flow, where a code-first LangChain build pays off, or if you just want a finished support bot.

Does Dify have a free plan or trial?

Yes. The Sandbox plan is free forever with 200 message credits, 5 apps, 1 member, 50 knowledge documents, and 30-day log history. It is meant for building and testing, not production traffic. Beyond that, the self-hosted Community edition is completely free if you run it on your own infrastructure.

What are the best Dify alternatives?

Flowise and Langflow are the closest open-source visual builders. n8n fits if AI is one step in a larger automation, Botpress if you want a polished customer-support bot, and LangChain with LangSmith if you prefer full code-level control. Flowise is usually the first alternative most Dify shoppers compare.

Can I self-host Dify and use my own model API keys?

Yes to both. The Community edition runs via Docker on your own servers for free, with no message-credit caps. On any tier you can plug in your own OpenAI, Anthropic, Google, Mistral, or local Ollama keys, which bypasses Dify's hosted credit limits and routes model cost to your own provider bill.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Dify pricing](https://dify.ai/pricing)checked Sep 2026
- [Flowise pricing](https://flowiseai.com/pricing)checked Sep 2026
- [Langflow pricing](https://langflow.org/#pricing)checked Sep 2026
- [n8n pricing](https://n8n.io/pricing)checked Sep 2026
- [Botpress pricing](https://botpress.com/pricing)checked Sep 2026
- [LangChain pricing](https://langchain.com/pricing)checked Sep 2026

Related guides

Ai App Development PlatformsBright Data ReviewN8n ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# DigitalOcean Review

URL: https://devshot.co/reviews/digitalocean-review
Type: review
Published: 2026-08-04
Updated: 2026-09-25
Summary: Honest DigitalOcean review: flat-rate Droplets from $4/mo, an App Platform PaaS, managed Postgres and Kubernetes, plus GPU Droplets for AI.

Review

## DigitalOcean Review

A developer-first cloud with flat, predictable pricing and a simple UX. Genuinely easy to ship on, with a smaller catalog of managed services and regions than AWS or GCP.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

DigitalOcean is a developer-first cloud built on flat, predictable pricing, with Droplets (Linux VMs) starting at $4 per month and most services published openly instead of hidden behind a calculator. Its biggest strength is simplicity: a clean UX, fast provisioning, and a stack that covers Droplets, the App Platform PaaS, managed Postgres, MySQL, Valkey and Redis, DOKS Kubernetes, Spaces object storage, and now GPU Droplets for AI. The catch is breadth: fewer managed services and regions than AWS or GCP, paid support tiers, and bandwidth overage once you pass the included transfer. The closest alternatives are Linode (now Akamai), Vultr, Hetzner, and AWS Lightsail. For indie hackers, startups, and small teams, DigitalOcean hits a rare sweet spot of price, clarity, and speed.

Key factsBest for: Indie hackers, startups, and small teams that want simple…

- Updated: September 25, 2026
- Best for: Indie hackers, startups, and small teams that want simple cloud hosting and a bill they can predict.
- Price as of September 25, 2026: From $4/mo (Basic Droplet, billed per second); App Platform free for three static sites.

- A developer-first cloud with flat, published pricing and a genuinely simple UX, now extended to GPUs and AI.
- Founded: 2011
- Headquarters: New York, USA
- Alternatives covered: Linode (Akamai), Vultr, Hetzner, AWS Lightsail

Pros

- Flat, transparent pricing you can forecast before the month ends.
- Clean UX plus world-class docs and community tutorials.
- Covers Droplets, PaaS, managed databases, Kubernetes, and GPUs in one place.

Cons

- Fewer managed services and regions than AWS or GCP.
- Not ideal for complex enterprise or multi-account architectures.
- Bandwidth overage and paid support tiers add up at scale.

DigitalOcean

Founded2011

HeadquartersNew York, USA

Est. priceFrom $4/mo

Best forStartups & developers

DigitalOcean is one of the first names developers reach for when AWS feels like overkill. It sits in the same conversation as Linode, Vultr, and Hetzner, and it has built its reputation on one simple promise: cloud infrastructure you can actually understand.

Instead of a pricing calculator with hundreds of line items, you get a Droplet, a flat monthly price, and a dashboard you can navigate in about a minute.

This review is written for software engineers, indie hackers, and startups deciding whether DigitalOcean is the right home for their apps.

We look at what you actually get, what it costs in real numbers, how it feels to deploy and run day to day, where it genuinely shines, and where it falls short against the hyperscalers and its flat-rate rivals. We also cover who should look elsewhere, and five alternatives worth pricing before you commit.

## What is DigitalOcean?

DigitalOcean is a cloud infrastructure provider founded in 2011 and headquartered in New York, built specifically for developers and small-to-mid-size businesses rather than large enterprises. It went public in 2021 and runs data centers across North America, Europe, and Asia.

At the center are Droplets, DigitalOcean's Linux virtual machines. Around them sits a full platform: the App Platform, a Heroku-style PaaS that deploys straight from a Git repo; Managed Databases for PostgreSQL, MySQL, Valkey, Redis, MongoDB, Kafka, and OpenSearch; DOKS, managed Kubernetes; Spaces, S3-compatible object storage; plus Block Storage, Load Balancers, Functions (serverless), and a container registry.

The newer story is AI. DigitalOcean now offers GPU Droplets (NVIDIA H100 and H200, plus AMD Instinct hardware), an inference engine, and a GenAI platform with knowledge bases and agents.

The pitch across all of it is the same: the predictable, published pricing and clean UX that made the core product popular, extended to modern AI workloads.

## How DigitalOcean works

Getting started is genuinely fast. You create a Droplet in under a minute, pick a region and a base image (Ubuntu, Debian, or a one-click app like Docker or a LAMP stack), add your SSH key, and you have a working server. Billing is per second with a small minimum charge, so a throwaway test box costs pennies.

For teams that would rather not manage servers, the App Platform reads a Git repo, detects the language, builds it, and deploys, handling TLS, scaling, and health checks for you. Managed Databases work the same way: pick an engine and a size, and DigitalOcean handles backups, patching, failover, and connection pooling.

The developer experience is the real product. The docs and community tutorials are among the best on the internet, the API and Terraform provider are clean, and doctl (the CLI) covers most tasks.

The rough edges show at scale: observability is thinner than CloudWatch, IAM and fine-grained permissions are simpler than AWS, and a few managed services lag the feature depth of their hyperscaler equivalents.

## DigitalOcean key features

Droplets (Linux VMs)Essential
Droplets are DigitalOcean's core: Linux virtual machines that spin up in under a minute. Basic shared-CPU plans run from $4 to $96 per month, with CPU-Optimized, General Purpose, and Memory-Optimized ranges above that. Per-second billing, snapshots, backups, and a large library of one-click images make them the workhorse most projects start on.

App Platform (PaaS)Essential
The App Platform is a Heroku-style PaaS that deploys directly from GitHub, GitLab, or a container image. It builds, runs, and scales your app while handling TLS, health checks, and rollbacks. A free tier hosts up to three static sites, and container apps start at $5 per month, making it a fast path to production without touching a server.

Managed Databases
Managed Databases cover PostgreSQL, MySQL, Valkey, Redis, MongoDB, Kafka, and OpenSearch, starting around $15 per month. DigitalOcean handles backups, patching, failover, standby nodes, and connection pooling, so a small team gets a production-grade database without a dedicated DBA. Read replicas and larger clusters are available as you grow.

DOKS managed Kubernetes
DOKS, DigitalOcean Kubernetes, gives you a managed control plane (free on the standard tier) and you pay only for the worker-node Droplets, from about $12 per month. It integrates with load balancers, block storage, and the container registry, giving startups real Kubernetes without the operational weight of running the control plane themselves.

Spaces and Block Storage
Spaces is S3-compatible object storage from $5 per month, bundling 250 GB of storage and 1 TB of outbound transfer, with a built-in CDN. Block Storage volumes ($0.10 per GB per month) and Network File Storage add persistent, resizable disks. Together they cover static assets, backups, and shared file needs with the S3 tooling you already know.

GPU Droplets and GenAI platform
GPU Droplets bring NVIDIA H100 and H200 and AMD Instinct accelerators on demand, priced from about $0.76 to $11.19 per GPU per hour, with cheaper multi-month reservations. Paired with the inference engine (from $0.05 per million tokens) and the GenAI platform for agents and knowledge bases, small teams can train and serve models without a hyperscaler contract.

## DigitalOcean pricing

DigitalOcean, Linode (Akamai), Vultr, Hetzner, and AWS Lightsail all publish prices, and none require a custom quote. The lowest entry is Vultr at $2.50 per month (IPv6-only), followed by Lightsail at $3.50 (also IPv6-only) and DigitalOcean at $4 per month; Hetzner lists its EU cloud plans in euros and bills the primary IPv4 separately.

DigitalOcean bills Droplets per second, the App Platform is free for up to three static sites, and Lightsail falls under the AWS Free Tier for new accounts. Costs jump on specialized Droplets: CPU-Optimized from $42 per month, General Purpose from $63, and Memory-Optimized from $84.

GPU Droplets run about $0.76 to $11.19 per GPU per hour on demand, or from $1.91 when reserved. Transfer overage is $0.01 per GB after the included transfer, and business and premium support adds a monthly percentage of spend.

Plan | Price | Best for |

DigitalOcean Basic Droplet | From $4/mo, billed per second | 512 MB RAM, 1 vCPU, 10 GB SSD, 500 GB transfer |

DigitalOcean Basic Droplet sizes | $6, $12, $24, $48, $96/mo, per second | Larger shared-CPU sizes, up to 16 GB RAM |

DigitalOcean CPU-Optimized | From $42/mo, per second | Specialized Droplets for CPU-heavy work |

DigitalOcean General Purpose | From $63/mo, per second | Balanced Droplets above the Basic range |

DigitalOcean Memory-Optimized | From $84/mo, per second | Droplets for memory-heavy workloads |

DigitalOcean App Platform | From $5/mo | Container apps; free for up to three static sites |

DigitalOcean Managed Database | From $15.15/mo | Single node; Postgres, MySQL, Valkey, Redis, and more |

DigitalOcean DOKS | From $12/mo | Worker node; control plane free on the standard tier |

DigitalOcean Spaces | $5/mo | 250 GB storage and 1 TB transfer, with a CDN |

DigitalOcean Block Storage | $0.10 per GB per month | Persistent, resizable disks billed per GB |

DigitalOcean GPU Droplet | About $0.76 to $11.19/GPU/hr | On-demand NVIDIA H100, H200, and AMD Instinct |

DigitalOcean GPU reserved | From $1.91/GPU/hr | Lower rate for sustained GPU workloads |

DigitalOcean transfer overage | $0.01 per GB | After pooled transfer, from 500 GB per month |

Linode Shared Nanode | From $5/mo | 1 GB RAM, 1 vCPU, 25 GB SSD, 1 TB transfer |

Linode Dedicated CPU | From $36/mo | Dedicated CPU; managed databases and Kubernetes |

Vultr Cloud Compute IPv6 | From $2.50/mo | IPv6-only entry, 32-plus global locations |

Vultr Cloud Compute | From $5/mo | 1 GB RAM, 1 vCPU. Also A100, H100, GH200, bare metal |

Hetzner Cloud VM | Check current pricing | 20 TB traffic; server auction and storage boxes |

AWS Lightsail Linux | From $5/mo with IPv4 | 512 MB RAM, 2 vCPU, 20 GB SSD, 1 TB transfer |

## DigitalOcean pros and cons

### What we like

- Flat, transparent pricing you can forecast before the month ends.
- Clean UX plus world-class docs and community tutorials.
- Covers Droplets, PaaS, managed databases, Kubernetes, and GPUs in one place.

### What could be better

- Fewer managed services and regions than AWS or GCP.
- Not ideal for complex enterprise or multi-account architectures.
- Bandwidth overage and paid support tiers add up at scale.

## Who DigitalOcean is for

DigitalOcean is a strong fit for indie hackers, startups, and small-to-mid-size engineering teams who want to ship without wrestling a hyperscaler. If you value predictable bills, a clean dashboard, and world-class docs over an exhaustive service catalog, it is one of the best homes for your app.

Side projects, SaaS MVPs, agency client sites, and early-stage products all land squarely in its sweet spot, and the App Platform makes it a credible Heroku replacement.

It is a weaker fit in a few clear cases. Large enterprises with complex, multi-account architectures, strict compliance regimes, or deep AWS and GCP dependencies will hit the ceiling on managed services, regions, and fine-grained IAM.

Teams that need a specific managed service DigitalOcean does not offer (exotic databases, dozens of global edge regions, or niche compliance tooling) should look at the hyperscalers. And cost-obsessed teams running heavy, steady compute in Europe can often get more raw power per dollar from Hetzner.

For most small teams, though, DigitalOcean trades breadth for clarity, and that is usually the right trade.

## Best DigitalOcean alternatives

If DigitalOcean is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

DigitalOcean | Indie hackers, startups, and small teams that want simple cloud hosting and a bill they can predict. | From $4/mo (Basic Droplet, billed per second) | Visit → |

Linode (Akamai) | Developers who want predictable Linux VMs now backed by Akamai's global edge network. | From $5/mo (Shared Nanode) | Visit → |

Vultr | Developers who want the most global regions and low-cost access to part of a GPU. | From $2.50/mo (IPv6-only) or $5/mo (1 GB RAM, 1 vCPU) | Visit → |

Hetzner | Cost-conscious developers and startups, especially in Europe, who want the most compute per dollar. | EU cloud plans priced in euros, primary IPv4 billed separately | Visit → |

AWS Lightsail | Developers who want DigitalOcean-style simplicity but a bridge into the full AWS ecosystem. | From $3.50/mo (IPv6-only) or $5/mo with a public IPv4 | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

Linode (Akamai)

A veteran developer cloud, now part of Akamai, with flat-rate Linux VMs and a huge documentation library.

Visit →

Vultr

A DigitalOcean-style flat-rate cloud with the widest region coverage and cheap fractional GPUs.

Visit →

Hetzner

A German host famous for the cheapest high-spec VMs and dedicated servers, with generous included traffic.

Visit →

AWS Lightsail

Amazon's simplified, fixed-price VPS that bundles compute, storage, and transfer with an on-ramp to full AWS.

Visit →

## The bottom line

DigitalOcean earns its following. For indie hackers, startups, and small teams, it delivers the rare combination of flat, published pricing, a genuinely simple UX, and documentation that has taught a generation of developers.

Droplets, the App Platform, managed databases, and DOKS cover the vast majority of what a growing product needs, and the new GPU and GenAI stack keeps it relevant as workloads shift toward AI.

The trade-off is breadth. You give up the sprawling service catalog, regional reach, and enterprise tooling of AWS and GCP, and you pay for support beyond the basics.

Choose DigitalOcean if simplicity and predictable cost matter more than having every service under one roof. If you want raw price-to-performance, compare Hetzner; if you want the widest regions or cheap GPUs, look at Vultr; and if you may grow into the full AWS stack later, AWS Lightsail is the softer on-ramp.

## Frequently asked questions

How much does DigitalOcean cost?

DigitalOcean publishes flat pricing. Droplets (Linux VMs) start at $4 per month for a Basic plan (512 MB RAM, 1 vCPU, 10 GB SSD) and scale through $6, $12, $24, $48, and $96 tiers, billed per second. The App Platform has a free static-site tier with container apps from $5 per month, Managed Databases start around $15 per month, DOKS Kubernetes runs from about $12 per month, and Spaces object storage is $5 per month for 250 GB. GPU Droplets cost from roughly $0.76 to $11.19 per GPU per hour.

Is DigitalOcean cheaper than AWS?

For simple, predictable workloads, usually yes, and it is far easier to forecast. DigitalOcean's flat monthly Droplets and bundled transfer avoid the metered surprises that make AWS bills hard to predict. That said, Hetzner often beats DigitalOcean on raw price-to-performance, and at very large or complex scale AWS reserved instances and breadth can win. The real DigitalOcean advantage is a bill you can understand before the month ends.

What is a Droplet?

A Droplet is DigitalOcean's name for a Linux virtual machine, its core product. You choose a region, a size (from 512 MB of RAM upward), and a base image such as Ubuntu, Debian, or a one-click app, then get a running server in under a minute with a public IP and SSH access. Droplets are billed per second and can be resized, snapshotted, and backed up as your needs change.

Does DigitalOcean support Kubernetes and AI or GPU workloads?

Yes. DOKS (DigitalOcean Kubernetes) offers a managed control plane, free on the standard tier, with worker nodes from about $12 per month. For AI, GPU Droplets provide NVIDIA H100 and H200 and AMD Instinct hardware on demand, alongside an inference engine from $0.05 per million tokens and a GenAI platform for agents and knowledge bases. It is a real option for training and serving models without a hyperscaler contract.

What are the best DigitalOcean alternatives?

The closest flat-rate alternatives are Linode (now part of Akamai) and Vultr, which offer similar Droplet-style VMs, with Vultr adding more regions and cheap fractional GPUs. Hetzner is the pick for the most compute per dollar, especially in Europe, though its footprint is smaller. AWS Lightsail gives you DigitalOcean-style simplicity with a built-in path into the full AWS ecosystem if you expect to outgrow a simple VPS.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [DigitalOcean pricing](https://digitalocean.com/pricing)checked Sep 2026
- [Linode pricing](https://linode.com)checked Sep 2026
- [Vultr pricing](https://vultr.com/pricing)checked Sep 2026
- [Hetzner pricing](https://hetzner.com)checked Sep 2026

Related guides

Self Hosting ToolsDocker Desktop AlternativesCheap VpsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Firebase Review

URL: https://devshot.co/reviews/firebase-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Firebase review 2026: what Spark actually includes, what Blaze bills past the daily quotas, and how Supabase, Appwrite, Amplify, Convex, and Nhost compare.

Review

## Firebase Review

Worth it when offline sync and the Google client SDKs are the product. A weak default when you need a SQL bill you can forecast.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Firebase is worth it in 2026 when a mobile or web app needs Google's client SDKs and offline sync, and you can live inside the free quotas. Spark costs nothing, while Blaze is pay as you go: in us-central1, Firestore document reads are $0.03 per 100,000 after 50,000 free reads a day, verified on Google's pricing pages in September 2026.

One project covers Firestore, Auth, Storage, Hosting, Crashlytics, and Cloud Messaging, so a mobile team ships before anyone writes a server. The catch is read amplification: a listener, an offset, or a security rule bills extra reads, and a Google Cloud budget alerts you without stopping spend. For a schema you can leave, start with the Supabase review.

Key factsBest for: Mobile and web apps that need Google's client SDKs and…

- Updated: September 25, 2026
- Best for: Mobile and web apps that need Google's client SDKs and offline sync.
- Price as of September 25, 2026: Spark no-cost; Blaze pay as you go after daily quotas.

- Google's app backend: a real free tier, then a usage bill listeners can push past any forecast.
- Parent: Google
- Free plan: Spark, no card
- Paid plan: Blaze, pay as you go
- Alternatives covered: Supabase, Appwrite, AWS Amplify, Convex, Nhost

Pros

- Offline persistence and client SDKs for iOS, Android, web, and Flutter sit in the same project as Auth and Storage.
- Spark includes 50,000 Firestore reads a day and 1 GiB stored, and Crashlytics plus Cloud Messaging stay no-cost.
- SQL Connect adds PostgreSQL on Cloud SQL without leaving the Firebase project, though existing documents stay documents.

Cons

- Cloud Functions require Blaze, and a budget alert does not stop spend past the free quotas.
- Listeners, query offsets, and security-rule reads each add to the document-read bill, which is how a small screen gets expensive.
- Phone SMS is priced by destination country, so a global login is not a flat auth fee.

ParentGoogle

Free planSpark, no card

Paid planBlaze, pay as you go

Best forMobile apps with offline sync

Firebase is the backend mobile teams still reach for first, because the iOS, Android, and web SDKs get a signed-in user and a synced document store running before anyone has written a server.

The buying question in 2026 is whether the bill stays predictable once listeners and Cloud Functions are in production, since an alert does not stop the spend. This review prices Spark and Blaze, then sets them next to Supabase, Appwrite, AWS Amplify, Convex, and Nhost.

Toolradar data: the [September 2026 low-code backend ranking](https://toolradar.com/best/low-code-backend) evaluated 109 tools, and it lists Supabase as an open-source Firebase alternative.

[Firebase](https://toolradar.com/tools/firebase) and [Supabase](https://toolradar.com/tools/supabase) both have directory pages there. A separate [serverless ranking](https://toolradar.com/best/serverless) evaluated 53 tools the same month, with Convex outside the top 10.

How we compared: vendor pricing and product pages read on September 23, 2026. Firestore unit prices are Google Cloud's list rates for us-central1 (Iowa), the region selected on that pricing page.

## What is Firebase?

Firebase is Google's mobile and web app development platform, a set of managed products on one Google Cloud project. Authentication, Cloud Firestore, Realtime Database, Cloud Storage, Cloud Functions, Hosting, App Hosting, and SQL Connect share that project, so the mobile app calls one stack and one bill.

Analytics, Crashlytics, Cloud Messaging, Performance Monitoring, Remote Config, and App Distribution stay no-cost on both plans. Firestore is a schemaless document database, so you ship without migrations and pay in reads instead of seats.

Realtime Database is the older JSON tree, and it is still the simpler offline sync when the data is one nested object. SQL Connect, marked new on the product list, attaches Cloud SQL for PostgreSQL when the data is relational, and it does not turn your documents into tables.

There is no seat fee, and the meter is reads, writes, storage, egress, function time, and monthly active users only if you turn on Identity Platform. App Hosting is the newer deploy path, GitHub in and Cloud Run underneath, while classic Hosting remains the static CDN.

If that git-push loop is the whole requirement, the Vercel review is the closer comparison than another database.

## How Firebase works

You create a project, register an iOS, Android, or web app, and drop in the client SDK. Security Rules, not a server you wrote, decide which documents a signed-in user can touch. That loop is why a prototype lands before a backend repo exists.

The rough edge is that the client talks to the database directly. A rule that calls get() or exists() bills those extra document reads. A query offset bills every skipped document, which makes deep pagination expensive.

A listener bills a read when a result is added, updated, or dropped, and a reconnect after 30 minutes offline is billed like a fresh query.

Cloud Functions hold the logic that must not live on the device, and they are absent on Spark, so a webhook needs a billing account. Linking that account moves the project to Blaze, and unlinking it moves the project back. Phone sign-in has required that billing link since September 2024, so a texted code cannot stay on Spark.

On Spark, crossing a plan limit in a calendar month shuts the app off for the rest of that month. On Blaze the same traffic is billed, and a budget alert does not reject the request. For other ways to run that code, see serverless platforms.

## Firebase key features

Firestore listeners and offline persistenceEssential
A document database with realtime listeners and offline persistence in the mobile and web SDKs, so a phone app can open with no network. Listeners bill a read on each added, updated, or removed result, and a reconnect after 30 minutes offline bills like a new query.

Auth, with Identity Platform as the paid upgradeEssential
Email, social, and anonymous sign-in are included on Spark and Blaze, and phone auth needs Blaze and bills per SMS, which makes a global login a country-priced line. Identity Platform, used for SAML and OIDC, includes 50,000 monthly active users before Google Cloud's per-user rate.

Cloud Functions, Blaze onlyEssential
Server logic without a VM you patch, and Spark cannot run it, so a webhook forces a billing account. On Blaze the free invocation pool is counted on the billing account, not the project, so a second Firebase project does not get a second free pool.

SQL Connect for PostgreSQL
Attaches Cloud SQL for PostgreSQL and syncs relational reads and writes to the client, which suits joins that documents handle badly. The first instance includes a three-month trial, and staying on Spark afterward gets it archived, then deleted after another 90 days.

Hosting and App Hosting
Hosting is the static CDN, with a custom domain and SSL on both plans. App Hosting deploys full-stack apps from GitHub onto Cloud Run. Uncached bandwidth has its own monthly allowance, separate from Hosting's daily transfer cap, so a cache miss does not eat the static allowance.

Crashlytics, Analytics, and Cloud Messaging
Crashlytics, Analytics, Cloud Messaging, Performance Monitoring, and App Distribution stay no-cost on Spark and Blaze, so crash reports survive once you pay for reads. They are not a trace backend, and production latency still wants a tool built for that job.

## Firebase pricing

Firebase publishes two plans, not a seat grid, so you forecast usage rather than headcount. Spark needs no card, and Blaze keeps those allowances, then bills what you use past them. If the billing account is eligible, Blaze includes $300 in credit.

Google retired the old flat Flame plan in February 2022, so there is no fixed monthly ceiling left.

In us-central1, document reads are $0.03 per 100,000 and writes are $0.09 per 100,000, after the daily free reads and free writes. The caps reset around midnight Pacific time and do not roll forward, so a quiet week does not pay for a busy Tuesday.

A day with 1,000,000 reads past the free reads costs $0.30, and a day with 1,000,000 writes past the free writes costs $0.90. Only one database per project gets the free 1 GiB, so a second database pays from the first byte.

A team of five and a team of twenty pay the same Firebase bill, because there is no seat line. Convex Professional is the contrast: five developers are $125 a month and twenty are $500 before usage.

Supabase Pro, Appwrite Pro, and Nhost Pro are organization fees, so headcount does not move them, and Supabase Team and Nhost Team are each $599 a month when you need the compliance tier.

Prices below were verified on each vendor's pricing page alongside those Firestore rates. EU Convex deployments are 1.3 times the US list, and the included US usage does not apply there.

Plan | Price | Best for |

Firebase Spark | No-cost | No card. No Cloud Functions. Monthly cap shuts the app off. |

Firebase Blaze | Pay as you go | Spark quotas included. $300 credit if the account is eligible. |

Firestore reads, us-central1 | $0.03 per 100,000 | After 50,000 free reads a day. Quota does not roll. |

Firestore writes, us-central1 | $0.09 per 100,000 | After 20,000 free writes a day. |

Cloud Functions invocations | $0.40 per million | After 2 million a month, counted on the billing account. |

Functions outbound networking | $0.12 per GB | After 5 GB a month on Blaze. |

Hosting data transfer | $0.15 per GB | After 360 MB a day. Storage is $0.026 per GB after 10 GB. |

App Hosting uncached bandwidth | $0.20 per GiB | After 10 GiB a month, in effect since August 1, 2025. |

Realtime Database storage | $5 per GB | After 1 GB. Downloads are $1 per GB after the daily free transfer. |

Identity Platform users | $0.0055 per MAU | After 50,000 MAU, up to 100,000. Email auth without it stays included. |

Identity Platform SAML or OIDC | $0.015 per MAU | After 50 MAU a month. |

SQL Connect, Cloud SQL | From $9.37 a month | After a 3-month trial. Spark instances are archived, then deleted. |

Remote Config fetches | $0.06 per 10,000 | Since September 1, 2026, after 100,000 fetches a day. |

Phone SMS, Indonesia | $0.35 per SMS | Identity Platform country table. First 10 SMS each day are not billed. |

Supabase Free | No-cost | 50,000 MAU, 500 MB database, pauses after a week, 2 projects. |

Supabase Pro | $25/mo | Includes a Micro compute credit. Extra MAU, disk, and egress are usage. |

Supabase Team | $599/mo | SOC 2 and ISO 27001. HIPAA is an add-on. |

Appwrite Free | No-cost | 75,000 MAU, 2 GB storage, pauses after a week, 2 projects. |

Appwrite Pro | From $25/mo | 200,000 MAU and 2 TB bandwidth per project. Extra project is $15. |

AWS Amplify build minutes | $0.01 per minute | After 1,000 free minutes a month. No per-seat fee. |

Amplify example, their page | $8.08 a month | AWS's own 5-developer, 300-daily-user hosting example, before backend services. |

Convex Professional | $125/mo for 5 developers | List rate is per developer. Twenty developers are $500 before usage. |

Convex Business | $2,500 monthly minimum | SSO, SLAs, and dedicated deployments above Professional. |

Nhost Pro | Same org rate as Supabase Pro | 10 GB database and 50 GB storage included, then usage. |

Nhost Team | Compliance tier, quote above it | SOC 2 Type II. Enterprise is a separate quote. |

## Firebase pros and cons

### What we like

- Offline persistence and client SDKs for iOS, Android, web, and Flutter sit in the same project as Auth and Storage.
- Spark includes 50,000 Firestore reads a day and 1 GiB stored, and Crashlytics plus Cloud Messaging stay no-cost.
- SQL Connect adds PostgreSQL on Cloud SQL without leaving the Firebase project, though existing documents stay documents.

### What could be better

- Cloud Functions require Blaze, and a budget alert does not stop spend past the free quotas.
- Listeners, query offsets, and security-rule reads each add to the document-read bill, which is how a small screen gets expensive.
- Phone SMS is priced by destination country, so a global login is not a flat auth fee.

## Who Firebase is for

Firebase fits a mobile team that wants offline sync and Google's client SDKs, and a web team that will live with documents, security rules, and a usage bill. Spark covers Auth, Firestore inside the daily caps, Hosting, Crashlytics, and messaging with no card, which holds until server logic or a texted login forces a billing account.

It is the wrong default when the data is relational and you want SQL you can dump, because leaving means rewriting the client rather than exporting tables. SQL Connect is Cloud SQL with a trial, not a conversion of your documents, so stored documents stay on the read meter.

For a Postgres host you pick on its own, read Postgres hosting and keep a database client in the workflow.

Skip Firebase when you need a hard cost ceiling or a server you run yourself. Blaze will not stop at a budget, so the alert pages a human and the requests keep succeeding.

Supabase and Appwrite both document a Docker install, covered with the self-hosting tools roundup, and Appwrite's guide includes a DigitalOcean image that is a VM rather than another hosted backend (DigitalOcean review).

Crashlytics will not replace traces either, so cross-service latency starts with the Datadog review and the monitoring tools list. Devshot's daily brief is where a meter change like Remote Config's September price belongs. Subscribe free if you want that note in the inbox.

## Best Firebase alternatives

If Firebase is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Firebase | Mobile and web apps that need Google's client SDKs and offline sync. | Spark no-cost | Visit → |

Supabase | Teams that want Postgres, row-level security, and a Docker exit. | From $25/mo (Pro) | Visit → |

Appwrite | Developers who want one open-source backend with a Docker install. | Pro is a monthly org fee | Visit → |

AWS Amplify | Teams already on AWS who will accept Cognito, Lambda, and S3 bills. | No seat fee | Visit → |

Convex | TypeScript apps that want reactive queries and a per-developer price. | Professional is per developer | Visit → |

Nhost | GraphQL apps that want Hasura on Postgres with a published Pro plan. | Pro at the same monthly org rate | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 6 does not publish a comparable monthly price and is left out rather than estimated.

Supabase

Auth, storage, and edge functions on a real Postgres database you can dump and self-host.

Visit →

Appwrite

Auth, database, storage, and functions in one subscription, with a documented self-hosted Docker path.

Visit →

AWS Amplify

AWS's front-end host, with auth and data billed as separate services behind it.

Visit →

Convex

A reactive database and function runtime with a free cap, a usage Starter plan, and a per-developer Professional plan.

Visit →

Nhost

Postgres, Hasura GraphQL, auth, storage, and functions, with a free project that pauses after a week.

Visit →

## The bottom line

Firebase is worth paying for when offline sync and the Google SDKs are why you are here, and when someone will watch the read chart. The product will not refuse traffic when that chart spikes. Start on Spark, move to Blaze before Cloud Functions or phone auth, and treat the budget alert as a page rather than a brake.

Choose Supabase or [Appwrite](https://toolradar.com/tools/appwrite) when you want a Docker exit, Amplify when the rest of the company is already on AWS, and [Convex](https://toolradar.com/tools/convex) when a per-developer price is easier to defend than a read bill.

Choose [Nhost](https://toolradar.com/tools/nhost) when the API should be Hasura GraphQL, and use the GraphQL tools list if that layer is the real decision.

The Devshot brief is the short version of this stack call. Subscribe free if you want the next pricing note in the inbox.

Cite this: Devshot, "Firebase Review 2026", September 2026.

## Frequently asked questions

Is Firebase worth it in 2026?

Yes, when a mobile or web app needs Google's client SDKs and offline document sync, and someone will watch the usage bill, because a surprise invoice is the real failure. Spark costs nothing and covers Auth, Firestore within the daily caps, Hosting, Crashlytics, and Cloud Messaging. It is the wrong default when you need SQL you can dump, a hard spending stop, or Cloud Functions without Blaze. For relational data, compare Supabase before the schema is committed, because documents will not become tables.

How much does Firebase cost?

Spark is no-cost, and Blaze is pay as you go on top of the same allowances. In us-central1, Firestore charges $0.03 per 100,000 document reads after 50,000 free reads a day, and $0.09 per 100,000 writes after 20,000 free writes a day, which stays small until a listener multiplies the reads. Cloud Functions require Blaze and cost $0.40 per million invocations after 2 million a month, counted on the billing account rather than the project. There is no per-seat fee, so hiring does not raise the line. Prices were verified on Google's pricing pages in September 2026.

Does Firebase have a free plan?

Yes: Spark needs no payment method and includes the no-cost products plus daily Firestore and Hosting allowances. It does not include Cloud Functions or phone authentication, so a login code or a webhook is when you leave it. Cross a Spark limit during a calendar month and Firebase shuts the app off for the rest of that month instead of billing you, which is safe for a side project and unacceptable for customers. Production traffic belongs on Blaze, where those allowances still apply and the overage is what you pay.

How does Firebase compare with Supabase?

Firebase wins on mobile offline sync and on client SDKs that already cover iOS, Android, and web. Supabase wins when you want Postgres, row-level security, and a documented Docker install, because you can dump the database. Supabase Pro is $25 a month and Team is $599 a month for SOC 2, while the free plan pauses after a week idle, so a quiet demo will sleep. Firebase has no flat Pro price: you pay for reads, writes, and egress, or you stay on Spark and accept the monthly shutoff.

Can Firebase use SQL?

Yes, through SQL Connect, which attaches Cloud SQL for PostgreSQL so relational data can sit beside Firestore. The first instance includes a three-month no-cost trial, and after that Cloud SQL starts as low as $9.37 a month, varying by region and machine size, a standing fee even in a quiet month. Leave the project on Spark when the trial ends and Google archives the instance, then deletes it after another 90 days. Firestore stays a document database, and SQL Connect does not turn existing documents into tables.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Firebase pricing](https://firebase.google.com/pricing)checked Sep 2026
- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026
- [Appwrite pricing](https://appwrite.io/pricing)checked Sep 2026

Related guides

Supabase ReviewVercel ReviewServerless PlatformsPostgres HostingSelf Hosting ToolsDatabase ClientsGraphql ToolsDatadog ReviewDigitalocean ReviewMonitoring ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Firecrawl Review

URL: https://devshot.co/reviews/firecrawl-review
Type: review
Published: 2026-09-17
Updated: 2026-09-25
Summary: Firecrawl review for 2026: Free 1,000 credits, Hobby $16, Standard $83, Growth $333, Scale $599. Credit math, rollover rules, 10% partner discount…

Review

## Firecrawl Review

The scrape/crawl/search API that returns markdown and JSON for agents and RAG. Free 1,000 credits a month, then $16 to $599. Credits do not roll over until Scale.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Firecrawl is a web-data API for agents: scrape, crawl, map, search, and interact, with clean markdown or JSON out. Free is 1,000 credits a month, no card, 2 concurrent requests. Paid (effective September 4, 2026): Hobby $16/mo billed annually ($19 monthly) for 5,000 credits, Standard $83/mo for 100,000, Growth $333/mo for 500,000, Scale $599/mo for 1 million plus one-month rollover, Enterprise custom with ZDR and SSO. Scraping, crawling or mapping one page costs 1 credit; search is 2 credits per 10 results; interact is 2 credits per browser minute. Partner 10% off via the referral link. Closest alternatives: Bright Data, Apify, Browserbase, and Context.dev.

Key factsBest for: Agent and RAG teams that want markdown or JSON from the…

- Updated: September 25, 2026
- Best for: Agent and RAG teams that want markdown or JSON from the live web.
- Price as of September 25, 2026: From $16/mo (Hobby, billed yearly); free 1,000 credits/mo, 2 concurrent, no card

- A scrape/crawl/search API billed in credits, built to feed LLMs, from $0 to $599.
- Founded: 2024
- Headquarters: San Francisco, USA
- Alternatives covered: Bright Data, Apify, Browserbase, Context.dev

Pros

- Markdown/JSON output and a free 1,000-credit month with no card.
- One credit model across scrape, crawl, map, search, and interact.
- Published concurrency, rollover rules, and a 10% partner discount.

Cons

- Hobby/Standard/Growth credits expire at month end; only Scale rolls over.
- Search, interact, and JSON extras burn the pool faster than "1 page = 1 credit."
- Not a residential-proxy or actor-marketplace platform.

Firecrawl

Founded2024

HeadquartersSan Francisco, USA

Est. priceFree to $599/mo

Best forLLM-ready web data

Firecrawl is what you call when an agent needs the live web as clean markdown, not a pile of HTML and a headless-browser babysitter. You pass a URL (or a search, or a crawl seed). It returns text an LLM can actually use. That is a narrower job than Bright Data's proxy-and-dataset empire, and a higher-level job than renting raw browsers from Browserbase.

This review is for developers wiring RAG, research agents, or enrichment pipelines. Prices are from firecrawl.dev/pricing as of September 2026, with the current rate card effective September 4, 2026. I disclose a partner discount: 10% off at https://firecrawl.link/louis-dupple-com.

I am not claiming a scrape bake-off on your target sites. I am walking through the credit math, which plans roll over, and when you should pick Bright Data, Apify, Browserbase, or Context.dev instead.

Credit math is the whole pricing page. A hobby project that scrapes 200 docs pages a month fits in Free. An agent that searches, then scrapes, then interacts with a login wall, does not. I would rather you model 30 days of real calls on Free or Hobby than jump to Standard because a landing page said 100,000 credits sounds like a lot.

## What is Firecrawl?

Firecrawl is a hosted web-data API (open-source core, paid "fire-engine" for proxies and hard pages) from the Mendable team in San Francisco. The product is built for AI systems: scrape one URL, crawl a site, map the URL tree, search the web, or interact with a page (click, type, wait) when the data sits behind a UI.

Default output is markdown. You can also ask for HTML, screenshots, metadata, or JSON that matches a schema you send in.

The hosted service handles JavaScript render, smart wait, and anti-bot plumbing so you are not maintaining Playwright plus a proxy pool for every new domain. The open-source repo is there if you want to self-host the simpler path.

Interact, the dashboard, and the reliability layer are why most agent teams pay for hosted.

Firecrawl's own January 13, 2026 scrape-content benchmark (1,000 URLs) quotes 96% coverage and a P95 around 3.4 seconds. Treat that as vendor-published, not an independent lab. The useful product claim is simpler: one API for search, scrape, and interact, billed in a shared credit pool.

Firecrawl is not a proxy reseller and it is not a no-code robot recorder. If you need to click through a site as a human would, for hours, Browserbase is the closer tool. If you need a maintained Amazon actor, Apify's store is the closer tool.

Firecrawl's bet is that most agent work wants clean text from URLs, and that a shared credit is easier than a GB-of-residential-proxy invoice. Agree with that bet before you migrate a Bright Data job.

## How Firecrawl works

You create an API key and call scrape, crawl, map, search, or interact. Scrape takes a URL and comes back with markdown or structured JSON. Crawl follows links from a seed with depth and path filters. Map returns the URL tree without fetching every body.

Search spends credits on results, then you scrape the ones you care about. Interact is the browser minute: the agent clicks and types when a static scrape is not enough.

Credits are shared across endpoints. 1 credit is 1 page on a basic scrape, crawl, or map. Search is 2 credits per 10 results. Interact is 2 credits per browser minute. Advanced scrape formats (JSON, Question, Highlight) add 4 credits per page on top.

Firecrawl's monitor/extract engine bills 7 credits per page per check when you use that path. A scrape that returns no result is not charged. A 403 or 404 that is returned to you still costs 1 credit.

Concurrency and rate limits climb with the plan (2 concurrent on Free, 5 on Hobby, 25 on Standard, 50 on Growth, 100 on Scale). Pay-as-you-go, on paid plans only, tops up in $5 increments when you run out: 1,000 extra credits on Hobby, 2,000 on Standard, 2,500 on Growth, 5,000 on Scale.

You set a monthly cap or turn top-ups off. Unused plan credits roll over only on Scale (one month) and Enterprise (custom). Hobby, Standard, and Growth reset.

Concurrency is the other silent limit. Free's 2 concurrent requests will serialize a naive batch over 50 URLs. Standard's 25 is the first plan that feels like a pipeline. Rate limits on scrape/map/search versus crawl/agent are published on the same comparison table; read them if you batch.

And robots.txt for the FirecrawlAgent directive is respected on crawl. Do not treat that as optional just because you have credits left.

## Firecrawl key features

LLM-ready scrape and crawlEssential
One call returns markdown or schema JSON, with JavaScript rendering and smart wait included. That is the reason agent and RAG teams pick Firecrawl over a raw HTML scraper.

Shared credit poolEssential
Scrape, crawl, map, search, and interact draw from one balance. Easy to reason about if you stay on basic page credits; easy to blow if you lean on search, interact, or JSON extras.

Search plus interact
Search finds pages. Interact drives the ones that need clicks. Together they cover the "agent that uses the web" loop without you standing up a separate browser vendor on day one.

Open-source core, hosted fire-engine
You can read the repo and self-host the simple path. Hosted adds proxies, interact, and the reliability layer most production agents actually need.

Scale rollover and Enterprise ZDR
Scale is the first self-serve plan that keeps unused credits for a month. Enterprise adds zero data retention, SSO, SCIM, static egress IPs, and key restrictions.

Agent / MCP onboarding
CLI, MCP, and an agent-onboarding skill so coding agents can sign themselves up. Convenient if you live in that workflow, irrelevant if you just want a curl key.

## Firecrawl pricing

Firecrawl, Apify, and Browserbase publish list prices and a free plan. Bright Data publishes unit rates, $1.50 per 1K successful records on its Scraper APIs pay as you go and from $5/GB for residential proxies, plus a free tier of 5K records a month.

Firecrawl Enterprise, Bright Data enterprise, and Browserbase Scale are custom quotes, and Context.dev gives 1,000 free credits a month, then paid credit plans. The cheapest named subscription is Firecrawl Hobby at $16/mo billed yearly (about 16.7% less than paying month to month), with Apify Starter and Browserbase Developer ($20/mo) a few dollars above.

The bill jumps at Firecrawl Standard ($83/mo), Growth ($333/mo), and Scale ($599/mo yearly, 20% under monthly, with one-month rollover). Apify Business, the top published plan, is $999/mo or $899/mo yearly.

Plan | Price | Best for |

Firecrawl Hobby | $16/mo billed yearly ($19 monthly) | 5,000 credits; 10% partner discount |

Firecrawl Standard | $83/mo | 100,000 credits, 25 concurrent |

Firecrawl Growth | $333/mo | 500,000 credits, 50 concurrent, Slack support |

Firecrawl Scale | $599/mo billed yearly (20% off) | 1 million credits, one-month rollover, invoice billing |

Firecrawl Enterprise | Custom quote | ZDR, SSO, and dedicated support |

Bright Data Web Scraper API | $1.50 per 1K records | Per 1K successful records |

Bright Data residential proxies | From $5/GB | Proxy traffic, lower per-GB rate on monthly commitments |

Bright Data Enterprise | Custom quote | Enterprise plan, custom pricing |

Apify Starter | $19/mo | Fee returns as platform usage credit; 10% off billed annually |

Apify Scale | $199/mo | $199 of usage included, lower compute and proxy rates |

Apify Business | $999/mo | $999 of usage included, lowest published unit rates |

Browserbase Developer | $20/mo, then $0.12/hr | 100 hours included, then hourly overage |

Browserbase Startup | $99/mo, then $0.10/hr | 500 hours included, then hourly overage |

Browserbase Scale | Custom quote | Custom plan above Startup |

## Firecrawl pros and cons

### What we like

- Markdown/JSON output and a free 1,000-credit month with no card.
- One credit model across scrape, crawl, map, search, and interact.
- Published concurrency, rollover rules, and a 10% partner discount.

### What could be better

- Hobby/Standard/Growth credits expire at month end; only Scale rolls over.
- Search, interact, and JSON extras burn the pool faster than "1 page = 1 credit."
- Not a residential-proxy or actor-marketplace platform.

## Who Firecrawl is for

Firecrawl is a strong fit for developers who want LLM-ready pages without building a scraping company. RAG over docs sites, research agents, enrichment, price monitors: if the output you want is markdown or JSON, start here.

Free covers the spike of curiosity. Hobby at $16 annual is the side-project plan. Standard at $83 is where I would put an internal agent that runs every day. Scale at $599 is the first plan that keeps leftover credits for a month, which matters if your volume is lumpy.

It is a weaker fit if you need a huge residential proxy network and 1,000 pre-built site scrapers (Bright Data), a marketplace of ready-made actors (Apify), or raw headless browsers you drive with Playwright (Browserbase).

Firecrawl interact exists, but Browserbase is still the browser platform. Context.dev is the closer cousin if you want scrape plus brand/company data in one API. And if your monthly volume is "millions of pages of Amazon," you are shopping infrastructure, not a 5,000-credit hobby key.

Legal still applies. Public pages, site terms, personal data, and logged-in content are your problem, not a credit setting. Firecrawl can redact PII and sells ZDR on Enterprise. That does not bless a scrape your counsel would reject.

If the job is regulated or the target is angry about bots, talk to a lawyer and look at vendors that run KYC, not just at $16 Hobby.

## Best Firecrawl alternatives

If Firecrawl is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Firecrawl | Agent and RAG teams that want markdown or JSON from the live web. | From $16/mo (Hobby, billed yearly) | Visit → |

Bright Data | Engineering teams scraping hard anti-bot targets at scale. | $1.50 per 1K records pay as you go (Scraper APIs) | Visit → |

Apify | Teams that want a marketplace of ready-made scrapers plus compute. | From $19/mo (Starter, plus pay as you go) | Visit → |

Browserbase | Developers who need fleets of headless browsers, not a scrape-to-markdown API. | From $20/mo (Developer) | Visit → |

Context.dev | Agents that want scrape/crawl plus company or brand context in one API. | Free 1,000 credits/mo | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 5 do not publish a comparable monthly price and are left out rather than estimated.

Bright Data

Proxies, scraper APIs, a headless browser, and ready-made datasets under one roof.

Visit →

Apify

A scraper platform and actor store: rent someone else's crawler, or run your own, billed on usage.

Visit →

Browserbase

Cloud browsers for agents: Playwright/Puppeteer/Stagehand, billed on browser hours.

Visit →

Context.dev

A web-data API aimed at agents: scrape, crawl, extract JSON, and pull brand/company data.

Visit →

## The bottom line

Firecrawl is the right default when the consumer is an LLM. Free 1,000 credits, no card. Hobby at $16 annual is enough to prove a pipeline. Standard at $83 is the first grown-up plan. Scale at $599 is the first one that keeps unused credits for a month. Take the 10% partner discount if you are paying list.

Do not force it into a job it is not pricing for. Hard targets and datasets: Bright Data. Actor marketplace: Apify. Raw browsers: Browserbase. Scrape plus brand data: Context.dev. Watch search, interact, and JSON extras, because those are not 1 credit per page. And if your volume is lumpy, skip Growth's use-it-or-lose-it pool and look at Scale.

Start on Free. If you finish the month with 200 credits left, stay. If you finish it on hour two, buy Hobby or Standard and take the 10% partner cut. Move to Scale when rollover would have saved you money last quarter. Switch vendors when the job is proxies, actors, or raw browsers, not when a tweet said Firecrawl is the only scraper.

## Frequently asked questions

How much does Firecrawl cost?

Free is 1,000 credits a month with no card. Hobby is $16 per month billed annually ($19 monthly) for 5,000 credits. Standard is $83 per month for 100,000 credits. Growth is $333 per month for 500,000. Scale is $599 per month for 1 million credits with one-month rollover. Enterprise is custom. Prices are from firecrawl.dev/pricing, effective September 4, 2026. A partner link offers 10% off.

How do Firecrawl credits work?

1 credit equals 1 page on a basic scrape, crawl, or map. Search costs 2 credits per 10 results. Interact costs 2 credits per browser minute. JSON, Question, and Highlight formats add 4 credits per page. A scrape that returns nothing is free; a returned 403 or 404 still costs 1 credit.

Do Firecrawl credits roll over?

Not on Free, Hobby, Standard, or Growth. Scale carries unused plan credits for one month. Enterprise rollover is custom. Credits you buy as pay-as-you-go stay until you use them, and they expire if you cancel.

Is Firecrawl free?

Yes, with a permanent free tier: 1,000 credits every month, 2 concurrent requests, no credit card. It is enough to try scrape and search. Production volume needs Hobby or above.

What are the best Firecrawl alternatives?

Bright Data for proxies, hard anti-bot sites, and datasets. Apify for a marketplace of ready-made scrapers plus compute. Browserbase if you need raw headless browsers and Playwright control. Context.dev if you want scrape plus brand or company data in one API.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Bright Data pricing](https://brightdata.com/pricing)checked Sep 2026
- [Apify pricing](https://apify.com/pricing)checked Sep 2026

Related guides

Web Scraping ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# GitHub Copilot Review

URL: https://devshot.co/reviews/github-copilot-review
Type: review
Published: 2026-08-04
Updated: 2026-09-25
Summary: Honest GitHub Copilot review for engineers: real 2026 pricing across Free, Pro, Pro+, Max, Business, and Enterprise, plus inline completions…

Review

## GitHub Copilot Review

The most widely deployed AI coding assistant, wired straight into GitHub, VS Code, and the tools engineers already live in. Excellent value at $10 a month, with a usage-based billing model that rewards reading the fine print.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

GitHub Copilot is the default AI pair programmer for most engineering teams, and in 2026 it is a genuinely strong product rather than just the incumbent. Pricing is clear and cheap to start: a real Free tier at $0, Pro at $10 a month, Pro+ at $39 a month, Max at $100 a month, Business at $19 per user per month, and Enterprise at $39 per user per month. Inline completions and next-edit suggestions are unlimited on every paid plan and cost nothing extra, while chat, agent mode, code review, and the CLI draw from a monthly allowance. The big 2026 change is billing: on June 1 GitHub swapped fixed premium-request quotas for token-metered AI Credits, so heavy agent users can now run over their included pool. The strongest reasons to buy are ecosystem depth and multi-model access: Claude, GPT, and Gemini in one subscription, first-class VS Code, JetBrains, Neovim, and Visual Studio support, plus Copilot on github.com and in the CLI. The catch is that it is best inside the GitHub and Microsoft world, chat quality swings with the model you pick, and it has historically felt less agentic than Cursor. The closest alternatives are Cursor, Devin Desktop, Amazon Q Developer, and Zed.

Key factsBest for: Professional engineers and teams already on GitHub who want…

- Updated: September 25, 2026
- Best for: Professional engineers and teams already on GitHub who want multi-model AI in their existing IDE at low cost.
- Price as of September 25, 2026: From $10/mo (Pro); free plan at $0.

- The default AI coding assistant, deeply integrated with GitHub and priced to win, with multi-model access across Claude, GPT, and Gemini.
- Launched: 2021
- Vendor: GitHub (Microsoft)
- Alternatives covered: Cursor, Devin Desktop (formerly Windsurf), Amazon Q Developer, Zed

Pros

- Unbeatable value: a real Free tier and a $10 Pro plan with unlimited completions.
- Multi-model access to Claude, GPT, and Gemini in one governed subscription.
- Works everywhere engineers already do: VS Code, JetBrains, Neovim, the CLI, and github.com.

Cons

- Premium chat and agent usage are metered, and heavy users can hit overage costs.
- Historically less agentic and autonomous than Cursor for big multi-file tasks.
- At its best inside the GitHub and Microsoft ecosystem; chat quality varies by chosen model.

Launched2021

VendorGitHub (Microsoft)

Est. priceFree to $39/user/mo

Best forEngineers on GitHub

GitHub Copilot is the tool that made AI-assisted coding normal. It shipped as a technical preview in 2021, went general availability in 2022, and by 2026 it is used by millions of developers and a large share of the Fortune 500.

For most engineers the real question is no longer whether an AI assistant helps, because it does, but whether Copilot is still the right one now that Cursor, Devin Desktop, and Amazon Q have raised the bar on agentic coding.

This review is written for working software engineers and the leads who buy their tools.

We cover what Copilot actually is in 2026, how it fits into a real editing workflow, the model lineup, the premium-request and AI-credit billing that trips people up, what each plan costs, where Copilot is genuinely excellent, and where it falls short of the newer agent-first editors.

We also name four alternatives worth trying before you standardize a team on any one of them.

## What is GitHub Copilot?

GitHub Copilot is an AI coding assistant built by GitHub, owned by Microsoft, that lives inside your editor and across the GitHub platform.

At its core it does three things: inline code completion that suggests the next line or whole function as you type, Copilot Chat for asking questions about code, generating snippets, and debugging in natural language, and agent mode, where Copilot plans a change, edits multiple files, runs commands, and iterates toward a goal you describe.

What separates it from a single-model tool is model choice. One subscription gives you access to several frontier models, including Anthropic Claude (Sonnet and, on higher tiers, Opus), OpenAI GPT, and Google Gemini, so you can pick a fast, cheap model for boilerplate and a stronger reasoning model for hard refactors.

Copilot runs in VS Code, Visual Studio, the JetBrains IDEs, Neovim, and Xcode, plus Copilot in the CLI for shell commands and Copilot on github.com for asking questions about a repo, reviewing pull requests, and assigning issues to the Copilot coding agent.

On the top plans it adds enterprise codebase indexing, so answers and completions are grounded in your organization's own code rather than generic training data.

## How GitHub Copilot works

Day to day, most of the value is invisible. You type, and Copilot proposes greyed-out completions you accept with Tab. Next-edit suggestions predict the follow-on change across the file, which is where a lot of the real time savings hide.

None of this consumes your paid allowance, so the autocomplete you use hundreds of times a day is effectively free once you are on a paid plan.

The heavier features work differently. Copilot Chat opens a side panel where you reference files, symbols, or the whole workspace and ask for an explanation, a test, or a fix.

Agent mode takes a prompt like "add pagination to this endpoint and update the tests," then reads context, edits several files, runs the test suite, and shows you a diff to approve.

On github.com you can hand an issue to the Copilot coding agent, which spins up its own environment, pushes a branch, and opens a pull request for human review. Copilot code review can leave inline comments on PRs automatically.

The part engineers need to understand is metering. Completions are unlimited, but chat, agent runs, code review, and CLI requests draw from a monthly pool.

Historically that pool was counted in premium requests, and in June 2026 GitHub moved paid plans to token-based AI Credits, so a single long agent session on a large model can burn far more of your allowance than a quick chat.

Run out and you either stop, wait for the monthly reset, or pay overage. In practice this means picking the right model for the job is not just about quality, it is about cost.

## GitHub Copilot key features

Inline completions and next-edit suggestionsEssential
The original Copilot feature and still the best reason to keep it on. As you type it predicts the next line, block, or entire function, and next-edit suggestions extend that to the follow-on changes elsewhere in the file. Crucially, completions are unlimited on every paid plan and never touch your credit allowance, so the highest-frequency feature carries no per-use cost.

Copilot Chat with multi-model accessEssential
An in-editor chat that understands your open files and workspace, explains unfamiliar code, writes tests, and fixes errors in place. The differentiator is model choice: Claude, GPT, and Gemini variants are all selectable from one subscription, letting you trade speed for reasoning depth. Chat quality does vary noticeably by model, so the picker matters.

Agent mode and the Copilot coding agent
Agent mode plans a change, edits multiple files, runs commands and tests, and iterates until the task is done, all from a natural-language prompt. The coding agent on github.com goes further, taking an assigned issue in its own environment and opening a pull request. Powerful, though historically less autonomous and polished than Cursor's agent for large multi-step tasks.

Broad IDE and CLI coverage
First-class extensions for VS Code and Visual Studio, solid support in the JetBrains suite, Neovim, and Xcode, plus Copilot in the CLI for generating and explaining shell commands. Wherever your team already works, Copilot is there, which is a real advantage over editor-locked tools like Cursor and Devin Desktop that require adopting a new IDE.

Copilot on github.com and code review
Beyond the editor, Copilot answers questions about a repository, summarizes and reviews pull requests, and can leave automated inline review comments. For teams that live in GitHub pull requests, this pulls AI assistance into the exact place code already gets shipped, with no context switch.

Enterprise codebase indexing and knowledge
On Enterprise, Copilot indexes your organization's repositories so chat and completions are grounded in your actual codebase, conventions, and internal libraries rather than generic patterns. Combined with knowledge bases and policy controls, this is what makes answers useful in a large private monorepo instead of merely plausible.

## GitHub Copilot pricing

GitHub Copilot, Cursor, Devin Desktop, Amazon Q Developer, and Zed all publish list prices and a free plan, so none of this set is quote-only. The cheapest paid entry is a tie at $10 a month between Copilot Pro, the plan most individual engineers want, and Zed Pro, and Copilot pairs it with a real free plan at $0, not a trial.

Costs jump at Copilot Pro+ ($39 a month) and Max ($100 a month), Cursor Ultra and Devin Max ($200 a month), and team seats such as Copilot Business at $19 per user per month and Cursor Teams from $40 per user per month.

Copilot's included AI Credits total 1,500 on Pro, 7,000 on Pro+, and 20,000 on Max, at $0.01 each, and premium overage moved from $0.04 per request to about $0.01 per credit in June 2026. Students, teachers, and maintainers of popular open-source projects can get Copilot Pro free, and inline completions on paid plans stay outside that credit pool.

Plan | Price | Best for |

GitHub Copilot Free | $0 | 2,000 completions, 50 chats, limited agent, Haiku-class and GPT mini |

GitHub Copilot Pro | $10/mo | Unlimited completions and chat, about $15 credits, was 300 requests |

GitHub Copilot Pro+ | $39/mo | Claude Opus for heavy use, about $70 credits, was 1,500 requests |

GitHub Copilot Max | $100/mo | 20,000 AI credits a month and priority access to premium models |

GitHub Copilot Business | $19/user/mo | Org policy, license control, IP indemnity, data protections, about $19 credits |

GitHub Copilot Enterprise | $39/user/mo | github.com, codebase indexing, knowledge bases, about $39 credits |

GitHub Copilot legacy overage | $0.04 per request | Extra premium requests under the old quota |

GitHub Copilot credit overage | Near $0.01 per credit | 2026 token rate after included credits |

Cursor Hobby | Free | Free Hobby tier |

Cursor Pro | $20/mo | Individual paid plan |

Cursor Ultra | $200/mo | For heavy usage |

Cursor Teams | From $40/user/mo | Team seats |

Devin Desktop Free | Free | Free tier with a daily quota |

Devin Desktop Pro | $20/mo | Higher daily and weekly quotas with frontier models |

Devin Desktop Max | $200/mo | Top plan for heavy use |

Devin Team | $40 per full seat, $80/mo minimum | Up to 200 users; two full seats cover the minimum, each with its own quota |

Amazon Q Developer Free | Free | Free tier available |

Amazon Q Developer Pro | $19/user/mo | Monthly allowance of agentic requests and code transformations |

Zed Personal | Free | 2,000 accepted edit predictions, unlimited use with your own keys |

Zed Pro | $10/mo | Unlimited edit predictions and $5 of hosted-model tokens |

Zed Business | $30/seat/mo | Org-wide AI policies, data governance, unified billing |

## GitHub Copilot pros and cons

### What we like

- Unbeatable value: a real Free tier and a $10 Pro plan with unlimited completions.
- Multi-model access to Claude, GPT, and Gemini in one governed subscription.
- Works everywhere engineers already do: VS Code, JetBrains, Neovim, the CLI, and github.com.

### What could be better

- Premium chat and agent usage are metered, and heavy users can hit overage costs.
- Historically less agentic and autonomous than Cursor for big multi-file tasks.
- At its best inside the GitHub and Microsoft ecosystem; chat quality varies by chosen model.

## Who GitHub Copilot is for

Copilot is the safe, high-value default for most professional engineers, and an especially easy call for anyone already on GitHub.

If your team reviews code in GitHub pull requests, uses VS Code, Visual Studio, or JetBrains, and wants Claude, GPT, and Gemini in one governed subscription without adopting a new editor, Copilot is hard to beat on value and fit.

The $10 Pro plan is one of the best deals in developer tooling, and the real Free tier means there is no reason not to try it.

It is a weaker choice in a few cases. Teams that want the most autonomous, agent-first editing experience for large multi-file changes may find Cursor or Devin Desktop ahead of Copilot's agent. Shops deep in the AWS ecosystem who want infrastructure-aware assistance and security scanning will lean toward Amazon Q Developer.

And developers who want a fast native editor running on their own API keys should look at Zed. Copilot is at its best inside the GitHub and Microsoft world, and slightly less compelling the further you get from it.

## Best GitHub Copilot alternatives

If GitHub Copilot is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

GitHub Copilot | Professional engineers and teams already on GitHub who want multi-model AI in their existing IDE at low cost. | From $10/mo (Pro) | Visit → |

Cursor | Engineers who want the most polished agent-first editing experience and will adopt a dedicated IDE. | From $20/mo (Pro) | Visit → |

Devin Desktop (formerly Windsurf) | Developers who want an agentic AI IDE with a clean flow-based UX and a generous free tier. | From $20/mo (Pro) | Visit → |

Amazon Q Developer | Teams building on AWS who want cloud-aware assistance, security scanning, and code transformation. | From $19/user/mo (Pro) | Visit → |

Zed | Developers who want a fast native editor and prefer their own model keys or external agents. | From $10/mo (Pro) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Cursor

A VS Code fork rebuilt around AI, widely seen as the leader for agentic, multi-file coding.

Visit →

Devin Desktop (formerly Windsurf)

Cognition's agent-first AI editor, formerly Windsurf, and Cursor's closest rival on approachable UX.

Visit →

Amazon Q Developer

Amazon's AI coding assistant, strongest for AWS-heavy stacks and large-scale code modernization.

Visit →

Zed

A fast, open-source native editor with edit predictions, hosted models on Pro, and bring-your-own-key AI.

Visit →

## The bottom line

GitHub Copilot remains the smart default, and in 2026 that is a compliment rather than a hedge. It offers the best value in the category, a real Free tier, unlimited completions that cost nothing extra, multi-model access to Claude, GPT, and Gemini, and coverage across every major IDE plus the CLI and github.com.

For any team already shipping code through GitHub, it is the path of least resistance and, usually, the right one.

The honest caveats are worth pricing in. The move to token-based AI Credits means heavy agent users need to watch overage, chat quality depends on the model you select, and for the most autonomous multi-file editing Copilot's agent has historically trailed Cursor.

Buy Copilot if value, governance, and fit with your existing tools matter most. Try Cursor or Devin Desktop if you want the sharpest agent and will switch editors, reach for Amazon Q Developer if you live on AWS, and pick Zed if you want a fast native editor on your own keys. For most engineers, though, the $10 Pro plan is the easiest yes in developer tooling.

## Frequently asked questions

How much does GitHub Copilot cost?

GitHub Copilot has a free plan at $0 and five paid tiers: Pro at $10 a month, Pro+ at $39 a month, Max at $100 a month, Business at $19 per user per month, and Enterprise at $39 per user per month. Inline code completions are unlimited on every paid plan and never consume your allowance. Chat, agent mode, code review, and CLI usage draw from a monthly AI-credit pool that scales with the plan, and going over it is billed as overage.

Is the GitHub Copilot free plan any good?

Yes, for light use and evaluation. The Free tier includes roughly 2,000 code completions and about 50 chat messages a month, limited agent usage, and access to a smaller set of models. It is genuinely useful for personal projects and the best way to test Copilot before paying, but heavy daily users will hit the limits and want Pro, where completions and chat are unlimited.

What AI models does GitHub Copilot use?

Copilot is multi-model. From one subscription you can pick between Anthropic Claude (Sonnet, and Opus on higher tiers), OpenAI GPT models, and Google Gemini, choosing a faster or cheaper model for routine work and a stronger reasoning model for complex refactors. The Free tier exposes a smaller, lighter model set. Because chat and agent quality vary by model, the picker is worth learning rather than leaving on the default.

GitHub Copilot vs Cursor: which is better?

They serve overlapping but different needs. Copilot wins on value, ecosystem reach, multi-model choice, and working inside the IDEs and GitHub workflow you already use. Cursor, a dedicated AI editor, is widely considered ahead for autonomous, agent-driven multi-file changes but costs $20 a month and requires adopting its own editor. Many engineers run Copilot as their everyday assistant and reach for Cursor when they want the strongest agent.

What changed with GitHub Copilot billing in 2026?

On June 1, 2026, GitHub moved paid plans from fixed premium-request quotas to token-based AI Credits. Instead of counting a set number of premium requests per month, Copilot now meters premium features like chat, agent mode, and code review by token usage against an included credit allowance. Inline completions stay unlimited and free of charge. The practical effect is that long agent sessions on the largest models can consume credits quickly, so heavy users should monitor usage and expect possible overage.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [Devin Desktop pricing](https://cognition-labs.com)checked Sep 2026
- [Amazon Q Developer pricing](https://aws.amazon.com/q/developer/pricing/)checked Sep 2026
- [Zed pricing](https://zed.dev/pricing)checked Sep 2026

Related guides

Ai Coding AssistantsCursor ReviewFree Ai Coding ToolsHow to Get GitHub Copilot for Free in 2026: Free Plan, Students, Teachers and Open SourceAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Heroku Alternatives in 2026

URL: https://devshot.co/reviews/heroku-alternatives
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Heroku alternatives in 2026 for a web process and a worker, with USD prices from each vendor.

Expert Guide

## The Best Heroku Alternatives in 2026

A Standard-1X dyno is still 512 MB, so pick the host that keeps a web process, a worker, and Postgres.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 8 tools compared

TL;DR

[Render](https://toolradar.com/tools/render) is the best Heroku alternative in 2026 when the app still needs a web process, a worker, and Postgres. A Starter instance is $7/mo for 512 MB and 0.5 CPU. Use [Railway](https://toolradar.com/tools/railway) when the bill should follow CPU and RAM, [Fly.io](https://toolradar.com/tools/flyio) when the process has to sit in Ashburn, or [DigitalOcean App Platform](https://toolradar.com/tools/digitalocean-app-platform) when the account already pays for Droplets.

Stay on [Heroku](https://toolradar.com/tools/heroku) when the workload needs a Shield dyno or Heroku Connect into Salesforce, because a new host will not recreate either. Eco is a personal-account pool that sleeps after 30 minutes and allows only two process types, which makes it a sandbox.

Key facts8 tools compared: Render, Railway, Fly.io, DigitalOcean App Platform, Northflank, Koyeb…

- Updated: September 25, 2026
- Top pick: Render (best for: Teams moving a web process and a worker off dynos)
- Top pick price as of September 25, 2026: Render: Starter web service $7/mo (512 MB, 0.5 CPU); Hobby workspace has no fee

- 8 tools compared: Render, Railway, Fly.io, DigitalOcean App Platform, Northflank, Koyeb, Coolify, Dokku
- Railway (best for: Apps whose CPU and RAM should set the bill): Hobby $5/mo with a matching usage credit; Pro $20/mo per workspace
- Fly.io (best for: Processes that must run in a chosen city, not the cheapest database): shared-cpu-1x, 256 MB, $1.94/mo in Ashburn; no workspace fee
- DigitalOcean App Platform (best for: Accounts that already run Droplets and want a container): Smallest container: 1 vCPU, 512 MiB, 50 GiB transfer; see the container row

A Heroku alternative is a host for a web process, a background worker, and a database, deployed from git or a container, without a dyno formation. For a Rails or Django app with a Procfile, Render is the closest shape in 2026, because it still sells a sized instance.

A usage meter suits idle apps, and a VPS you patch suits a team that already maintains a box.

Toolradar data: On 24 September 2026 the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) we publish scored 297 products.

Heroku is named in the group ranked 11 through 30, outside the ten at the top, alongside DigitalOcean App Platform.

Every dollar below was read on the vendor's own pricing page on 24 September 2026, in USD.

A Next.js app that never ran a worker belongs on the serverless platforms guide and the Vercel review, because it has no worker to replace.

## Top Picks

Based on features, real-world fit, and value for money.

Best Heroku Alternatives in 2026: 8 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Render](https://toolradar.com/tools/render) | Starter web service $7/mo (512 MB, 0.5 CPU); Hobby workspace has no fee | Teams moving a web process and a worker off dynos |

[Railway](https://toolradar.com/tools/railway) | Hobby $5/mo with a matching usage credit; Pro $20/mo per workspace | Apps whose CPU and RAM should set the bill |

Fly.io | shared-cpu-1x, 256 MB, $1.94/mo in Ashburn; no workspace fee | Processes that must run in a chosen city, not the cheapest database |

DigitalOcean App Platform | Smallest container: 1 vCPU, 512 MiB, 50 GiB transfer; see the container row | Accounts that already run Droplets and want a container |

Northflank | nf-compute-20 at $5.40/mo (0.2 shared vCPU, 512 MB); sandbox does not sleep | Buildpack apps that may later run in your cloud account |

Koyeb | eco-nano $1.61/mo in Washington; Pro is $29/mo with compute included | A tiny always-on instance, or one free web service |

[Coolify](https://toolradar.com/tools/coolify) | Cloud base fee for 2 servers; self-hosted has no license fee | Teams who want a PaaS panel on servers they already operate |

[Dokku](https://toolradar.com/tools/dokku) | No hosted price; open source, git push onto a VPS you operate | A git-push workflow on one server you already patch |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 8 do not publish a comparable monthly price and are left out rather than estimated.

1

### Render

Top Pick

Best for: Teams moving a web process and a worker off dynos

PricingStarter web service $7/mo (512 MB, 0.5 CPU); Hobby workspace has no fee

+Starter matches a Standard-1X dyno's 512 MB, adds 0.5 CPU, and a worker uses the same menu

+Pro workspaces drop the per-seat fee, include SOC 2 and ISO 27001 reports, and include 15 custom domains

+Paid Postgres starts at 0.1c-256mb, with 1 GB of storage included and expandable disk at $0.30/GB

−A free web service spins down after 15 minutes idle, wakes in about a minute, and cannot run a worker

−Hobby is a single seat and 5 GB of bandwidth, and a free Postgres database expires 30 days after you create it

Visit Render →

2

### Railway

Best for: Apps whose CPU and RAM should set the bill

PricingHobby $5/mo with a matching usage credit; Pro $20/mo per workspace

+Stopped services cost nothing, and service builds are not billed for CPU or memory

+Pro is a per-workspace fee with unlimited seats, 30-day logs, and up to 42 replicas on a service

+Egress is $0.05/GB and volume storage is $0.15/GB for a full month, both metered in smaller units

−A full month of 1 GB of RAM is $10 before CPU, so a hot Hobby service pays the usage total, not a capped dyno

−The free plan is one project, three services, one replica, and no custom domains, with a $1 monthly credit

Visit Railway →

3

### Fly.io

Best for: Processes that must run in a chosen city, not the cheapest database

Pricingshared-cpu-1x, 256 MB, $1.94/mo in Ashburn; no workspace fee

+Ashburn and Secaucus share the 256 MB preset price on the card, and extra RAM is $5.00 per GB

+Egress is $0.02/GB in North America and Europe, and a dedicated IPv4 is $2/mo

+A stopped Machine bills $0.15 per GB of root filesystem per month, not the running preset

−Managed Postgres Basic is $38/mo for a shared-2x with 1 GB of RAM, and disk is $0.28/GB on top

−There is no free machine tier, and paid support starts at the Standard row in the table

Visit Fly.io →

4

### DigitalOcean App Platform

Best for: Accounts that already run Droplets and want a container

PricingSmallest container: 1 vCPU, 512 MiB, 50 GiB transfer; see the container row

+The smallest paid container is 1 vCPU and 512 MiB with 50 GiB of transfer included

+Three static-only apps are free, each with 1 GiB of transfer, and extra transfer is $0.02 per GiB

+A development database is 512 MiB of Postgres attached to that one app, at the row in the table

−Fixed shared sizes allow a single container instance, and autoscaling is limited to dedicated instances

−That development database is not backed up by default and is destroyed when the app is destroyed

Visit DigitalOcean App Platform →

5

### Northflank

Best for: Buildpack apps that may later run in your cloud account

Pricingnf-compute-20 at $5.40/mo (0.2 shared vCPU, 512 MB); sandbox does not sleep

+The getting-started container is 0.2 shared vCPU and 512 MB, prorated to the second

+The sandbox includes two services, one database, and two cron jobs, and that compute does not sleep

+The pricing page lists no seat fee and no added platform charge for a workload in your own VPC

−Network egress is $0.06/GB, and disk is $0.15/GB per month

−Bring-your-own-cloud still leaves the AWS, GCP, or Azure bill in that account

Visit Northflank →

6

### Koyeb

Best for: A tiny always-on instance, or one free web service

Pricingeco-nano $1.61/mo in Washington; Pro is $29/mo with compute included

+eco-nano is 0.1 vCPU and 256 MB at $1.61/mo in Washington, Frankfurt, or Singapore

+The pricing FAQ states charges are in USD, and a paused service is not billed

+Each organization gets one free web instance: 512 MB, 0.1 vCPU, and 2 GB of SSD

−That free instance cannot be a worker, cannot attach a volume, and scales to zero after 1 hour without traffic

−The pricing FAQ says scale-to-zero for paid services is still on the roadmap, and a card check places a $29 hold

Visit Koyeb →

7

### Coolify

Best for: Teams who want a PaaS panel on servers they already operate

PricingCloud base fee for 2 servers; self-hosted has no license fee

+Self-hosted is free and the pricing page says it includes current and upcoming features

+Cloud's base fee covers two connected servers, with unlimited deployments on each

+Annual Cloud billing is listed at a 20% saving against the monthly base fee

−The Cloud subscription does not include the machines, so the VPS invoice is still yours

−An extra team on Cloud requires another subscription, and a third server is an add-on slot

Visit Coolify →

8

### Dokku

Best for: A git-push workflow on one server you already patch

PricingNo hosted price; open source, git push onto a VPS you operate

+The default build path is Herokuish buildpacks, and a Dockerfile deploy is the other documented path

+The official Postgres plugin links a service so the app receives DATABASE_URL

+There is no license fee and no per-seat charge from the Dokku project

−A new app does not include a database, and installing the Postgres plugin requires root on the host

−The git remote user must be dokku or the push fails, and there is no vendor uptime agreement

Visit Dokku →

## What it is

You are replacing a formation, not a logo: a web process, a worker, and a Postgres add-on, built with a buildpack or a Dockerfile. A host that cannot price all three is not an exit.

Eco allows two process types and sleeps after 30 minutes of inactivity, so a team app that must stay awake should skip it.

Basic allows ten process types and stays up, which makes it the first always-on shape. Standard-1X is the first tier with unlimited process types, at 512 MB.

Cedar is the classic runtime a Procfile app is already on.

Fir is the Kubernetes runtime, and every Fir size on the pricing page requires a Private Space, so a Fir quote is isolation rather than a cheaper dyno. Postgres Essential plans use shared RAM: Essential-0 is 1 GB of disk, 20 connections, and a listed maximum downtime of 4 hours a month, the outage you accept on that plan.

Standard-0 is the first dedicated tier, the step when shared RAM is the complaint.

How we compared: on 24 September 2026 we read those eight vendors' USD pricing pages plus Heroku's dyno and Postgres lists. The 297-product ranking is the wider set, and this is a price comparison rather than a timed cutover.

## Why it matters

The bill that pushes a team off Heroku is two always-on dynos plus a database, not the Eco sticker. Two Standard-1X dynos and Essential-0 Postgres add up to $55/mo, the number to beat. Performance-M, the first common-runtime dyno with dedicated compute, is $250 for 2.5 GB.

The smallest Private dyno, Private-S, is $125 and it requires a Private Space.

Render prices the instance and then a workspace, not a seat per engineer. Two Starter instances plus 0.1c-256mb Postgres are the cheap always-on stack in the table.

A free web service spins down after 15 idle minutes, and the workspace only includes 750 free instance hours. A Pro workspace does not lift either limit, so an upgrade will not wake that free app.

Railway's Hobby fee includes a matching usage credit, and the bill becomes the usage total once you pass it.

Memory is $10 per GB for a full month, and CPU is the per-vCPU row in the table, metered per second. A service left scaled up is not covered by the credit, while a stopped service costs nothing, so idle time is the saving. Fly.io has no workspace fee, and a shared-cpu-1x with 256 MB is $1.94/mo in Ashburn, but Managed Postgres Basic is $38/mo before disk.

That database line is why a Fly Machine is not an automatic saving when the data is the product, with the longer comparison in the Postgres hosting guide.

A VM exit starts with the VPS hosting guide, the DigitalOcean review, and the self-hosting guide.

A cluster is the Kubernetes tools guide, pipelines are the CI/CD tools guide, and a phone backend is the Firebase alternatives list. Later price changes show up in the Devshot brief.

## Key features to look for

Whether the worker is a first-class process
Heroku bills another dyno for the worker, so a queue is a second invoice line. Render, Railway, Fly.io, and Northflank bill another service, and Koyeb's free instance cannot be a worker.

What happens when nobody sends a request
Eco sleeps after 30 minutes, which suits a personal app and fails a late webhook. Render's free web tier sleeps after 15, Railway bills nothing while stopped, and Fly bills root filesystem only, so idle cost is the real difference.

Where Postgres sits on the invoice
Heroku and Render sell a database plan you can show finance. Railway meters a service plus a volume, Fly starts at the Basic cluster, and DigitalOcean's dev database is not a backup.

The unit you cannot forget
Render Pro adds a workspace fee on top of the instances, the line teams forget at renewal. Railway Pro is per workspace with seats included, Fly.io and Northflank have no seat line, and Coolify bills connected servers.

Who owns the cloud account
Render, Railway, Fly.io, and Koyeb run on their cloud if you do not want an account of your own. Northflank can target your VPC, and Coolify and Dokku run on a machine you patch.

The compliance tier that is not a checkbox
Heroku Shield is its own price list, so compliance is a separate buy. Render Scale adds a HIPAA premium, Fly sells a HIPAA package, and Koyeb Scale is the plan with an SLA.

## Pricing

Read the formation before the sticker, because two Standard-1X dynos plus a database plan are the usual production shape. Essential-0 is shared RAM with a 4-hour downtime ceiling. Standard-0, at $50/mo, is the first dedicated Postgres plan: 4 GB of RAM, 64 GB of disk, and 120 connections.

Essential-1, at $9/mo, only adds disk on that shared tier, to 10 GB.

Render bills a workspace and then compute, so the instance is not the whole invoice. Hobby has no workspace fee, one seat, 5 GB of bandwidth, and two custom domains, then $0.25 per extra domain.

Scale is $499/mo, a Pro-size service is $85/mo for 4 GB and 2 CPU, and a persistent disk is $0.25/GB per month. Past the bandwidth pool, use the rate in the table once the app serves files.

Railway's invoice becomes the usage total once you pass the included credit, and Postgres is a service plus a volume rather than a named plan.

Fly's Ashburn Machine is the preset on the card, managed disk is $0.28/GB, and the HIPAA package is $99/mo, a compliance add-on rather than a bigger Machine. Coolify adds $3/mo per server after the first two, on top of the VPS.

Northflank's nf-compute-100-2 is $24/mo for 1 vCPU and 2 GB, Koyeb nano is $2.68/mo, and a small paid Postgres instance is $29.76/mo before storage at $0.50/GB. Checked on each vendor's pricing page in September 2026.

Plan | Price | Best for |

Heroku Eco dyno pool | $5/mo | 1,000 shared hours, 512 MB, sleeps after 30 minutes, two process types |

Heroku Basic dyno | $7/mo max | 512 MB, always on, ten process types, about $0.01 per hour |

Heroku Standard-1X dyno | $25/mo max | 512 MB, unlimited process types, preboot, one dyno unit per full month |

Heroku Postgres Essential-0 | $5/mo | Shared RAM, 1 GB disk, 20 connections, 4-hour monthly downtime ceiling |

Heroku Postgres Standard-0 | $50/mo | Dedicated 4 GB RAM, 64 GB disk, 120 connections, 1-hour downtime ceiling |

Render Starter instance | $7/mo | 512 MB and 0.5 CPU, web service or worker |

Render Standard instance | $25/mo | 2 GB and 1 CPU, four times Standard-1X RAM |

Render Pro workspace | $25/mo | Unlimited seats, 25 GB bandwidth, 15 domains, 1,000 pipeline minutes |

Render Scale workspace | $499/mo | 1 TB bandwidth, SSO, HIPAA workspace at a 20% compute premium |

Render Postgres 0.1c-256mb | $6/mo | 0.1 CPU, 256 MB RAM, 100 connections, 1 GB storage included |

Render Postgres 0.5c-1g | $19/mo | 0.5 CPU and 1 GB RAM |

Render bandwidth past the included pool | $0.15/GB | Hobby includes 5 GB, Pro includes 25 GB, then this rate |

Railway Hobby | Credit equals the fee | Matches the Hobby line on the card; overage bills as the usage total |

Railway Pro | $20/mo | Per workspace, unlimited seats, usage credit equals the fee, 30-day logs |

Railway CPU | $20/vCPU | Full-month equivalent for containers, not the sandbox VM rate |

Fly.io shared-cpu-1x, Ashburn | $1.94/mo | 256 MB, $0.0027 per hour, same sticker in Secaucus |

Fly.io performance-1x, Ashburn | $31/mo | 2 GB on the performance preset |

Fly.io Managed Postgres Basic | $38/mo | shared-2x, 1 GB RAM, high availability and backups included |

Fly.io Standard support | $29/mo | 36-hour first response; community support stays free |

DigitalOcean App Platform container | $5/mo | 1 vCPU, 512 MiB, 50 GiB transfer, one instance only |

DigitalOcean shared container, 1 GiB | $12/mo | 1 vCPU, 1 GiB, 150 GiB transfer, still no autoscaling |

DigitalOcean development database | $7/mo | 512 MiB Postgres, no default backups, deleted with the app |

DigitalOcean dedicated egress IP | $25/mo | Per app, a pair of static addresses |

Northflank nf-compute-100-2 | $24/mo | 1 dedicated vCPU and 2 GB, marked most popular on the list |

Koyeb Pro | $29/mo | Includes $10 of compute, 10 users, 100 services |

Koyeb Scale | $299/mo | Includes $100 of compute and a 99.9% uptime SLA |

Coolify Cloud | $5/mo | Two connected servers included; you still pay for those servers |

Dokku hosted | No hosted price | Open source on a VPS; Postgres is a plugin you install |

Mistakes to avoid

×Treating Eco as production fails, because the pool is 1,000 hours, it sleeps after 30 minutes, and a second always-on process does not fit.

×Assuming a Render Pro workspace wakes a free web service is wrong, because an upgrade does not remove the 15-minute spin-down.

×Pricing Fly.io from the Machine and skipping Managed Postgres Basic hides the bill, because the Essential tier is the cheap Heroku database.

×Putting a worker on Koyeb's free instance fails, because it cannot be a worker and paid scale-to-zero is still on the roadmap.

Expert tips

→Name the web process, the worker, and the database before you compare stickers, because a host that cannot price the worker is not a Heroku exit.

→If Postgres must stay a named plan, start with Render, because Railway and Dokku make the database a service you run.

→Put Coolify or Dokku on a machine someone already patches, because the license is the small number and the server is the cost.

## The bottom line

Move a dyno app to Render when you want a web service, a worker, and a Postgres plan with prices you can point at. Keep Heroku when Shield, a Private Space, or Heroku Connect is the feature you would have to rebuild.

Railway fits apps that sit idle, because a stopped service should leave the bill.

Fly.io fits a chosen city, and its database plan sits far above the Essential tier, so the Machine is not the invoice. DigitalOcean App Platform fits an account that already runs Droplets, and Northflank fits when the workload has to stay in your cloud.

Koyeb is a small instance rather than a free worker, and Coolify and Dokku belong on a server you already patch.

The prices were read for a dyno exit, separate from that 297-product ranking. Cite this: Devshot, "Best Heroku Alternatives in 2026", September 2026.

## Frequently asked questions

What is the best Heroku alternative in 2026?

Render, when you want a web process, a worker, and Postgres as its own plan. Starter is 512 MB and 0.5 CPU at the table price, and the Hobby workspace has no fee. Railway fits a usage meter, Fly.io fits a chosen city, and DigitalOcean App Platform fits an account that already pays DigitalOcean. Stay on Heroku for Shield dynos or Heroku Connect.

How much does it cost to leave Heroku?

Two Standard-1X dynos plus Essential-0 Postgres are $55/mo on Heroku's list, and each dyno is 512 MB. Render's matching RAM is the Starter row, and 0.1c-256mb Postgres is $6/mo. Fly.io's Ashburn preset is $1.94/mo, while managed Postgres is the Basic cluster row, so the preset alone is not the saving. Railway bills the usage total once you pass the included credit. Rates were read on vendor pricing pages in September 2026.

Is there a free Heroku alternative?

Render's free web service sleeps after 15 minutes, and its free database expires after 30 days, so leave production off both. Railway's free plan is a $1 monthly credit, three services, and no custom domains. Koyeb's one free web instance has 512 MB and cannot be a worker. Coolify self-hosted has no license fee, and Dokku publishes no hosted price, so both still need a server you pay for. Heroku Eco is a paid hour pool that sleeps after 30 minutes.

Render vs Railway: which replaces a dyno?

Render replaces the dyno with a sized instance, Starter on the card and a Standard instance at 2 GB and 1 CPU in the table. Railway replaces it with memory and CPU rates, and Hobby's credit only equals the subscription. Pick Render when you want a ceiling you can read before deploy, and pick Railway when stopped time should cost nothing.

Can I self-host a Heroku alternative?

Dokku is the git-push version: buildpacks by default, a Dockerfile path if you want it, and Postgres only after you install the plugin as root. Coolify is the panel, self-hosted with no license fee, or Cloud for two servers you still rent. Northflank can target your VPC with no added platform charge, and you still pay that cloud. Heroku publishes no self-hosted edition, so a box you run is a different product.

Does Fly.io replace Heroku Postgres?

The Machine can replace the dyno, but not the database beside it. Managed Postgres Basic is $38/mo for shared-2x and 1 GB of RAM, plus disk at $0.28/GB, so it does not replace Essential-0 or Essential-1. Standard-0 is $50/mo for 4 GB of dedicated RAM. Use Fly when the city matters more than a cheap Essential database, and see the Postgres hosting guide for the longer comparison.

What breaks when you migrate off Heroku?

The Procfile, add-on variables, and any code that assumed a dyno sleep schedule, so price the worker as its own service before you cut over. Eco sleeps after 30 minutes, Render's free tier sleeps after 15, and a paid Render instance does not, which changes what your code can assume. Essential-0 caps connections at 20, so a worker pool above that cap will be refused. On Railway the database is a volume plus a container, not a clicked add-on.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Render pricing](https://render.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026
- [Coolify pricing](https://coolify.io/pricing)checked Sep 2026
- [Dokku pricing](https://dokku.com)checked Sep 2026

Related guides

Digitalocean ReviewVercel ReviewServerless PlatformsVps Hosting For DevelopersPostgres HostingSelf Hosting ToolsKubernetes ToolsCi Cd ToolsFirebase AlternativesDevshot briefAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Heroku Review

URL: https://devshot.co/reviews/heroku-review
Type: review
Published: 2026-09-24
Updated: 2026-09-25
Summary: Heroku review 2026: what Eco hours actually buy, why the metrics dyno stays on the smallest RAM, and how Render, Railway, Fly.io, App Platform…

Review

## Heroku Review

Worth it for a git-push app with a Postgres add-on. A weak default once the web process outgrows the smallest always-on dyno.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Heroku is worth it in 2026 when a small team wants git-push deploys and a managed Postgres add-on, and will pay for an always-on dyno. Eco is $5 for 1,000 shared dyno hours a month and sleeps after 30 minutes idle, verified on Heroku's pricing page in September 2026. Basic, the always-on size, maxes at $7 per dyno.

A production web dyno with metrics is Standard-1X and still only 0.5 GB of RAM. Team apps cannot use Eco at all, so a company account starts on the always-on size. The closest replacements are [Render](https://toolradar.com/tools/render), [Railway](https://toolradar.com/tools/railway), Fly.io, DigitalOcean App Platform, and Northflank.

Key factsBest for: Small teams that will pay for an always-on dyno and a…

- Updated: September 25, 2026
- Best for: Small teams that will pay for an always-on dyno and a Postgres add-on.
- Price as of September 25, 2026: Eco $5 for the hour pool; Basic $7/dyno/mo max, always on.

- Salesforce's PaaS: a Procfile and git push, and the dyno bill starts before the app has traffic.
- Founded: 2007
- Parent: Salesforce
- Cheapest dyno: Eco, 1,000 hours
- Alternatives covered: Render, Railway, Fly.io, DigitalOcean App Platform, Northflank

Pros

- A sleeping web dyno on Eco does not spend the hour pool, so a quiet personal app can last the month.
- Standard-1X adds preboot, metrics, and horizontal scale without leaving the common runtime.
- Essential-0 Postgres, Key-Value, and Kafka share the dyno invoice.

Cons

- Team and Enterprise apps cannot use Eco, so the cheapest always-on dyno is Basic.
- Standard-1X is still 0.5 GB, so the 1 GB common-runtime size is Standard-2X at $50.
- An empty Eco pool sleeps every Eco dyno for the rest of the month, and extra hours are not for sale.

Founded2007

ParentSalesforce

Cheapest dynoEco, 1,000 hours

Best forGit-push apps with Postgres

Heroku is still the platform a Rails or Node team means when they say the app should go live from a git push, and in 2026 that habit has a floor price. Eco is a paid hour pool, not a free dyno, and the always-on size bills even when nobody is hitting the app. A personal project can sleep and a Team app cannot, which is the split in this purchase.

This review prices Cedar dynos, Postgres, and the Private Space that Fir requires, then sets them next to Render, Railway, Fly.io, DigitalOcean App Platform, and Northflank. Stay if the Procfile and the add-on invoice are the workflow you want. Leave if you need more memory than the smallest metrics dyno.

Toolradar data: the [September 2026 hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) evaluated 297 tools, and 230 of them offer a free or freemium plan. [Heroku](https://toolradar.com/tools/heroku) is in that catalog and sits outside the top 10. A dyno that stays awake is a paid line on day one.

How we compared: vendor pricing and product pages read on September 24, 2026. Dyno and data prices are Heroku's published monthly maximums for a resource that runs the whole month. Eco is the exception, a flat subscription whether you use the hours or not.

## What is Heroku?

Heroku is Salesforce's platform as a service, acquired in 2011, which is why Enterprise features and a Salesforce sync sit behind a contract rather than a self-serve switch. You push code, Heroku builds it, and runs each process type in a dyno, so the bill follows process count, not a single app price.

The common runtime is Cedar, where Eco, Basic, Standard, and Performance dynos live, along with Postgres, Key-Value Store, and Kafka. Fir is the Kubernetes generation, and every Fir dyno requires a Private Space, so those shapes are not a cheaper Basic.

Heroku Connect syncs Postgres with Salesforce, and the Demo edition starts free, enough to see the mapping and not enough for production data. Enterprise and Shield editions are limited to Heroku Enterprise customers, so a production sync is a contract.

Managed Inference sits on the same pricing page, with token rates rather than a seat fee, so you pay only when a model is called.

## How Heroku works

The everyday loop is a git push and a Procfile, which is the habit you are paying for. The web process and a worker are separate formations, so both running is two dynos, not a bundled app fee.

One-off commands, including Heroku Scheduler, provision a dyno for the length of the command and that time is billed, so a short nightly task is not free.

Eco is only for personal apps, meaning apps that are not in a Heroku Team or an Enterprise Team. The plan is a fixed hour pool for the month, shared by every Eco dyno, review app, and Heroku CI run on that account, so a pull request and the live app spend the same hours.

A web dyno sleeps after 30 minutes without requests, and sleep does not consume hours, so idle time stays inside the pool. A worker-only Eco app does not sleep, because it never sees web traffic, and it can burn the pool around the clock.

When the account hits the full pool, every Eco dyno is forced to sleep until the next month, and Heroku does not sell extra Eco hours. The documented move for an app that must stay up is Basic, roughly $0.01 an hour.

Standard-1X adds metrics, threshold alerts, preboot, and horizontal scale, and the RAM does not increase with those features, so you pay for a safer deploy rather than a larger process. Performance dynos are limited to accounts with an established payment history, so a new team cannot use them as the first production size.

## Heroku key features

Cedar dynos, from Eco to PerformanceEssential
Eco is a flat monthly pool for personal apps only, and an idle web dyno sleeps so it stops drawing on that pool. Basic stays up and allows 10 process types, which a Team app has to buy. Standard adds metrics, preboot, and horizontal scale, but the first Standard size does not add RAM. Performance dynos are dedicated compute for larger workloads.

Postgres, Key-Value, and Kafka add-onsEssential
Essential-0 Postgres is 1 GB of disk, 20 connections, and a 4-hour monthly downtime ceiling on shared RAM, so treat it as a prototype. Standard-0 is the first dedicated database, with 4 GB of RAM, 64 GB of disk, and rollback of up to 4 days. Premium-0 adds high availability on that same footprint, a standby rather than more memory. Key-Value Mini is 25 MB, and it runs out once the key set grows.

Fir dynos inside a Private Space
Fir sizes such as dyno-1c-4gb run only in a Fir-generation Private Space, so a Basic app never sees them. The space itself is a separate Elements charge, starting near $1.389 an hour, and the dynos inside are billed on top, so the space fee starts before any Fir dyno does.

Review apps and pipelines
Review apps and Heroku CI draw from the app owner's Eco hour pool when the app is personal, so a busy pull-request week can empty the hours production also needs. Collaborators do not have their own pool: usage lands on the owner, and a teammate's CI run can sleep those dynos for the rest of the month.

Managed Inference
Nova Lite and Claude Opus 4.8 are listed per million tokens, in US and EU, rather than as a seat you pay whether you call them or not, so an unused model adds nothing. MCP tool calls can run as one-off dynos, which means an agent loop pays dyno time on top of the tokens.

Heroku Connect
Connect syncs rows between Heroku Postgres and Salesforce, and Demo is the free start, enough to try the mapping and not enough for production data. Enterprise and Shield Connect are sold only to Heroku Enterprise customers, so a startup on Basic cannot turn the production sync on from the dashboard.

## Heroku pricing

Heroku bills dynos on wall-clock time, not CPU time. Anything scaled above zero accrues usage whether or not requests arrive, prorated to the second, except Eco, which is a flat fee for the month you are subscribed. A dyno left scaled up still costs the full month, traffic or not.

A Basic dyno that runs the whole month is the always-on common-runtime size. A Standard-1X web dyno plus a Standard-1X worker is two of those $25 lines, and the database is a third line.

Essential-0 Postgres matches the Eco monthly fee, with a 4-hour downtime ceiling and shared RAM, which fits a side project and fails if that outage would wake a customer, and the dedicated step up is Standard-0.

High availability is Premium-0 at $200, still on the same RAM and disk, with rollback out to 7 days and a 15-minute downtime ceiling.

Kafka's cheapest plan, Basic 0, is $100 a month, often a larger line than the web dyno.

Render's always-on Starter matches that Basic monthly maximum, and its free web instance spins down after 15 minutes inside a 750-hour pool. Railway Pro is $20 per workspace with a matching usage credit and unlimited seats, plus the per-vCPU and per-GB rates in the table.

Fly.io's shared-cpu-1x is $2.02 a month in Ashburn and Amsterdam, and $2.43 in San Jose, so the city is part of the price. App Platform's smallest shared container matches Eco's monthly rate, and a development database is another line at Basic's monthly maximum. Northflank's nf-compute-10 is $2.70 a month for a container you can leave running.

Prices below were verified on each vendor's pricing page in September 2026. Heroku Enterprise, Render Enterprise, Railway Enterprise, and Northflank Enterprise are quotes. Fly.io support is a separate purchase from the machine price.

Plan | Price | Best for |

Heroku Eco | $5 for 1,000 hours | Personal apps only. Sleeps after 30 idle minutes. Pool does not expand. |

Heroku Basic | $7/dyno/mo max | Always on, 10 process types. About $0.01/hour. |

Heroku Standard-1X | $25/dyno/mo max | Metrics, preboot, horizontal scale. Still 0.5 GB, so extra RAM is a larger dyno. |

Heroku Standard-2X | $50/dyno/mo max | First common-runtime dyno with 1 GB of RAM. |

Heroku Performance-M | $250/dyno/mo max | 2.5 GB dedicated. Requires an established payment history. |

Heroku Postgres Essential-0 | Same monthly fee as Eco | 1 GB disk, 20 connections, 4-hour downtime ceiling, shared RAM. |

Heroku Postgres Standard-0 | $50/mo max | Dedicated 4 GB RAM, 64 GB disk, rollback up to 4 days. |

Heroku Postgres Premium-0 | $200/mo max | High availability on the same 4 GB. Rollback up to 7 days. |

Heroku Key-Value Mini | $3/mo max | 25 MB, 20 connections. |

Heroku Kafka Basic 0 | $100/mo max | Smallest Kafka plan. 3.73 GB, one week retention. |

Heroku Private Space | From ~$1.389/hour | Space fee on Elements. Dynos inside are extra. Fir requires a space. |

Heroku Fir dyno-1c-4gb | $80/dyno/mo max | 1 vCPU, 4 GB. Private Space only. |

Heroku Nova Lite | $0.06 / $0.24 per 1M tokens | Input / output. US and EU on the pricing page. |

Heroku Claude Opus 4.8 | $5.50 / $27.50 per 1M tokens | Input / output. US and EU. |

Render Free web | $0 | Spins down after 15 idle minutes. 750 free instance hours a month. |

Render Starter web | Matches Basic | 512 MB, 0.5 CPU. Billed while the instance is on. |

Render Pro workspace | $25/mo | Unlimited seats. 25 GB bandwidth, then $0.15/GB. |

Render Postgres 0.1c-256mb | $6/mo | Free Postgres expires 30 days after creation. |

Railway Hobby | Fee equals its usage credit | Same monthly number as Eco. Stopped services cost nothing. |

Railway Pro | $20/workspace/mo | Includes a matching usage credit. Seats are unlimited. |

Railway memory | $10 per GB-month | CPU is $20 per vCPU-month past the credit. Egress is $0.05/GB. |

Fly.io shared-cpu-1x, Ashburn | $2.02/mo | 256 MB, markup 1. San Jose is $2.43. Card required. |

Fly.io volume | $0.15/GB/mo | Billed even if the Machine is stopped. Rootfs on a stopped Machine is the same rate. |

DigitalOcean App Platform | Smallest shared container matches Eco | 512 MiB, 50 GiB transfer. Scalable 1 GiB size is $12/mo. |

DigitalOcean dev database | Matches Basic, per 512 MB | Extra outbound transfer is $0.02/GiB. Free tier is static sites only. |

Northflank nf-compute-10 | $2.70/mo | 0.1 shared vCPU, 256 MB. Egress is $0.06/GB. |

Northflank Sandbox | No-cost services | 2 services, 1 database, 2 cron jobs. They do not sleep. |

## Heroku pros and cons

### What we like

- A sleeping web dyno on Eco does not spend the hour pool, so a quiet personal app can last the month.
- Standard-1X adds preboot, metrics, and horizontal scale without leaving the common runtime.
- Essential-0 Postgres, Key-Value, and Kafka share the dyno invoice.

### What could be better

- Team and Enterprise apps cannot use Eco, so the cheapest always-on dyno is Basic.
- Standard-1X is still 0.5 GB, so the 1 GB common-runtime size is Standard-2X at $50.
- An empty Eco pool sleeps every Eco dyno for the rest of the month, and extra hours are not for sale.

## Who Heroku is for

Heroku fits a small team that wants a Procfile, a Postgres add-on, and a bill made of named dynos rather than a cloud calculator. Personal projects can live on Eco if idle sleep is acceptable. A customer-facing app on a Team starts at Basic, because Eco is refused there, and it should move to Standard once you want preboot and metrics.

It is the wrong default when the process needs more memory than the smallest metrics dyno and you are not ready to step up, or when high availability on Postgres means the Premium plan. Fir and network isolation add a Private Space before the first Kubernetes-shaped dyno, so isolation is a platform fee.

If the workload is a VM you SSH into, read VPS hosting for developers and the DigitalOcean review.

If it is a container you run yourself, the self-hosting tools list is the other path, and it pays off only when someone will patch the host.

Skip Heroku when the front end is the product. A Next.js app is usually a closer call on the Vercel review and the serverless platforms roundup, because an always-on dyno is the wrong meter for a page.

Managed Postgres you want to shop on its own is the Postgres hosting page, with Supabase if the database should be the platform. Pipelines that outgrow review apps belong with the CI/CD tools list.

A monitor that pings an Eco app every minute will keep it awake and spend the hour pool, which is the same failure mode Heroku documents for uptime checks. Production alerts belong on the monitoring tools list, not on a cron that defeats sleep.

Dyno price changes show up in the Devshot brief. Subscribe free if you want that note in the inbox.

## Best Heroku alternatives

If Heroku is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Heroku | Small teams that will pay for an always-on dyno and a Postgres add-on. | Eco $5 for the hour pool | Visit → |

Render | Teams that want web, worker, and Postgres services with a free instance for trials. | Starter web matches Basic | Visit → |

Railway | Backends where a stopped service should drop to zero compute. | Pro $20/workspace/mo with a matching usage credit | Visit → |

Fly.io | Long-running Machines placed in a specific region, billed while started. | shared-cpu-1x 256MB is $2.02/mo in Ashburn | Visit → |

DigitalOcean App Platform | Git deploys when the rest of the stack is already on DigitalOcean. | Smallest shared container matches Eco | Visit → |

Northflank | Preview environments and containers, including a sandbox that stays up. | From $2.70/mo (nf-compute-10) | Visit → |

Render

A service catalog close to Heroku's, with a free web instance that sleeps and a paid Starter that stays up.

Visit →

Railway

Per-second CPU and memory, with a workspace credit so a stopped service drops the compute line.

Visit →

Fly.io

Firecracker Machines priced by region markup, for a team that cares which city the process runs in.

Visit →

DigitalOcean App Platform

DigitalOcean's PaaS: per-container sizes, a static-only free tier, and a dev database at a flat monthly rate.

Visit →

Northflank

Git-push containers and previews, with a sandbox that stays awake and per-second compute once you outgrow it.

Visit →

## The bottom line

Heroku is worth paying for when the workflow you want is a Procfile, review apps, and Postgres on one invoice, and someone will watch dyno count. Eco is a reasonable personal sandbox if sleep is fine.

A Team app should budget Basic at minimum and Standard-1X once metrics matter, and it should not expect the smallest RAM to carry a real worker, because that worker is a second dyno.

Choose [Render](https://toolradar.com/tools/render) when you want that service list with a free instance that sleeps on a timer. Choose Railway when a stopped process should cost nothing and seats should not multiply the fee.

Choose [Fly.io](https://toolradar.com/tools/flyio) when the region price is the point.

Choose App Platform when the account is already DigitalOcean, and [Northflank](https://toolradar.com/tools/northflank) when previews and a non-sleeping sandbox matter more than the Heroku add-on catalog.

A cluster, rather than a dyno, is the Kubernetes tools decision.

The Devshot brief is the short version of this bill. Subscribe free if you want the next dyno change in the inbox.

Cite this: Devshot, "Heroku Review 2026", September 2026.

## Frequently asked questions

Is Heroku worth it in 2026?

Yes, when the team wants git-push deploys, review apps, and a Postgres add-on on one bill, and will pay for an always-on dyno. Eco is a paid hour pool on personal apps and sleeps when the web process is idle, which suits a portfolio and fails a customer URL. It is a weak default for a Team account, which cannot use Eco, and for any process that needs more than 0.5 GB. The metrics tier keeps that same RAM, and high-availability Postgres jumps to Premium-0 at $200.

How much does Heroku cost?

Eco is a $5 flat subscription for 1,000 dyno hours on personal apps only, so a Team account cannot start there. Basic maxes at $7 per dyno for a full month, Standard-1X at $25, and Standard-2X at $50, all prorated to the second if you scale down. Essential-0 Postgres uses the same monthly fee as Eco, Standard-0 matches that Standard-2X monthly maximum, and Premium-0 is $200. A Private Space starts near $1.389 an hour before the dynos inside it. Prices were verified on Heroku's pricing page in September 2026.

Does Heroku have a free plan?

There is no free dyno, and Eco bills for the month even if you use none of the hour pool, so an abandoned personal app still costs money. Heroku Connect's Demo edition is free, and Managed Inference has no seat fee, but both sit beside dyno charges once an app is running. When the hour pool runs out, dynos sleep until the next month rather than switching you to a paid overage. Render, by contrast, still offers a $0 web instance.

How does Heroku compare with Render?

Both bill a small always-on instance at the Basic monthly maximum, about 512 MB, and Heroku calls that size Basic with no $0 dyno. Render's free web instance spins down after 15 minutes and draws on 750 free instance hours, while Heroku Eco sleeps after 30 minutes inside a paid hour pool. Render's free Postgres expires 30 days after creation, so a trial database is not a long-term store. Heroku Essential-0 does not expire on a timer, and it states a 4-hour monthly downtime ceiling.

What is the difference between Cedar and Fir?

Cedar is the common runtime, home to Eco, Basic, Standard, and Performance, and it is the menu a normal app should stay on. Fir is the Kubernetes generation, with sizes such as dyno-1c-4gb at $80, and it only runs inside a Private Space. The space is its own hourly charge on the Elements listing, billed on top of the dyno, so Fir is a network-isolation purchase rather than a cheaper big dyno. Performance dynos on Cedar also require an established payment history, which blocks a new account from jumping to dedicated compute.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Render pricing](https://render.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026

Related guides

Vercel ReviewDigitalocean ReviewServerless PlatformsPostgres HostingVps Hosting For DevelopersSelf Hosting ToolsKubernetes ToolsCi Cd ToolsMonitoring ToolsSupabase ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Hetzner Cloud Review

URL: https://devshot.co/reviews/hetzner-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Hetzner Cloud review 2026: the Europe row carries the large outbound allowance. A new US server is a higher cap before IPv4…

Review

## Hetzner Cloud Review

Worth it for a Linux VM in Germany or Finland, where the outbound allowance is what keeps the month predictable. A new US order is a higher cap before the address, and the cheap CX row is labeled not available.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Hetzner Cloud is worth it in 2026 when the VM sits in Germany or Finland. The EU CPX12 is $13.49/mo before IPv4 and includes 20 TB of outbound traffic.

A new server in Ashburn or Hillsboro is the CPX11 at $20.49/mo before a $0.60 IPv4, with 1 TB included. The $6.49 CX23 is labeled not available, so skip that row.

Prices below are USD and exclude VAT, verified on Hetzner's price list on September 24, 2026. If the box has to live in the US with the address in the sticker, start with the DigitalOcean review.

Key factsBest for: Developers who can place a Linux VM in Germany or Finland…

- Updated: September 25, 2026
- Best for: Developers who can place a Linux VM in Germany or Finland and care about egress
- Price as of September 25, 2026: EU CPX12 $13.49/mo excl. IPv4; US CPX11 $20.49/mo excl. IPv4; IPv4 is $0.60/mo.

- A German cloud whose Europe traffic allowance is the product, and whose US cap is a separate price.
- US entry: CPX11, before IPv4
- EU entry: CPX12, 20 TB out
- IPv4: Billed apart from the VM
- Alternatives covered: DigitalOcean, Vultr, Akamai Cloud (Linode), OVHcloud, UpCloud

Pros

- The Europe CPX12 includes 20 TB outbound before IPv4, which is the row that still beats a metered US pool
- Hourly billing stops at a monthly cap, and incoming traffic plus same-zone traffic are free
- Firewalls and private networks have no extra charge, and primary IPv6 is free

Cons

- The US CPX11 includes 1 TB, so the Europe bargain is not the Ashburn price
- CX23 at $6.49/mo and the rest of the cost-optimized page were labeled not available on September 24, 2026
- A powered-off server and an unused primary IPv4 both keep billing until you delete them

Hetzner

US entryCPX11, before IPv4

EU entryCPX12, 20 TB out

IPv4Billed apart from the VM

Best forEU apps with heavy egress

Buy Hetzner when the workload can live in Falkenstein, Nuremberg, or Helsinki and the month is mostly egress. The Europe shared-AMD row is still the reason developers open this vendor.

The United States row is its own higher price list, and a currency conversion of the Europe sticker will not match what Ashburn bills.

Toolradar data: the [VPS and cloud infrastructure statistics](https://toolradar.com/statistics/vps-servers) track 5 published tools as of September 24, 2026, out of 10,680 tools in the directory.

[Hetzner](https://toolradar.com/tools/hetzner) has its own tool page, and so do [DigitalOcean](https://toolradar.com/tools/digitalocean) and [Vultr](https://toolradar.com/tools/vultr).

How we compared: Hetzner, DigitalOcean, Vultr, Akamai, OVHcloud, and UpCloud, priced in USD on September 24, 2026. We did not provision servers. The wider sheet is the VPS hosting guide for these same hosts.

## What is Hetzner?

Hetzner Cloud is a Linux virtual machine you SSH into, sold by a German host that also rents dedicated servers. Locations are Falkenstein, Nuremberg, Helsinki, Ashburn, Hillsboro, and Singapore, so the city is a price choice as well as a latency choice.

Hetzner owns the three European parks, while the US and Singapore sites are colocation, and support for every zone is the team in Germany.

Shared AMD plans (CPX) and dedicated AMD plans (CCX) are listed in every zone. Shared Intel/AMD and Ampere plans are listed only for Germany and Finland, so a US order cannot fall back to those cheaper shapes.

Object storage is listed for those European locations and not for Ashburn, Hillsboro, or Singapore, so a US VM cannot keep its files on this bill.

Images are Ubuntu, Debian, Fedora, CentOS, Rocky Linux, and AlmaLinux. Hetzner will not support Windows, even if you install it yourself, so keep Windows off this bill. There is no app platform on this invoice, so a git push with a sleeping dyno belongs on the Heroku review.

## How Hetzner works

You create a project, pick a location and a plan, and attach a primary IP if the server needs a public interface. There is no public interface until at least one primary IP is assigned. Primary IPv6 is free, and a primary IPv4 that sits on no server still bills until you delete the address.

The server has an hourly rate and a monthly cap, so an early delete costs the hourly rate rounded up to a full hour. Leave it powered off and you still pay, because the resources stay reserved.

A snapshot, a delete, and a later rebuild is the cheaper pause, since snapshots bill for compressed gigabytes and skip attached volumes.

Currency is chosen when the account is created, USD or EUR, and it cannot be changed later. Moving a server into a project with the other currency, or rescaling it, switches that server onto the current price list.

A rebuild, or a move between projects that share a currency, leaves the older cap in place.

Firewalls and private networks have no extra charge, and official libraries exist for Go and Python. Outbound speed is not a guarantee on the shared 10 Gbit/s host link. Hetzner says to expect about 300 to 500 Mbit/s, which is the rate a transfer job should use.

## Hetzner key features

Two price lists, not one conversionEssential
Germany and Finland share one USD column, while Ashburn and Hillsboro share a higher one, and Singapore is a third. A Europe screenshot will understate what a new US order costs. The June 15, 2026 adjustment hits new orders and rescales. An older server keeps its cap until you rescale it, move it across currencies, or restore it after a delete.

Primary IPv4 as its own lineEssential
Primary IPv4 has an hourly rate of $0.0010, and primary IPv6 is free, so a v6-only box can skip the address charge. One of each type can sit on a server before you need another address product. A floating IPv4, the address you move between servers, is a different product at $3.50/mo.

Outbound traffic, by zoneEssential
EU cloud servers include at least 20 TB of outbound traffic, and the US CPX11 includes 1 TB. A file host that stays inside the bundle in Germany runs into overage in Ashburn. Incoming traffic and traffic inside the same network zone are free. Traffic to another Hetzner zone counts as ordinary outbound, billed in 100 MB blocks.

Cost-optimized plans labeled not available
On September 24, 2026 the cost-optimized page labeled every CX and CAX plan not available, including CX23, so that sticker may not be a server you can create. Regular Performance CPX plans did not carry that label, which leaves the CPX row as the orderable path.

Backups are 20% and seven slots
Daily backups cost 20% of that server's monthly price and keep seven slots. The oldest drops off, and the copies die with the server, so deleting the VM deletes the copies too. In Hillsboro and Singapore the backup stays in the only data center, and volumes are not inside the backup, so an attached disk needs its own copy.

Dedicated vCPU when the share is not enough
CCX13 is 2 dedicated AMD vCPUs, 8 GB of RAM, and 80 GB of NVMe, for when a shared core is too noisy. Europe is $50.49/mo before IPv4, and Ashburn and Hillsboro are $50.99/mo before IPv4, with 1 TB outbound instead of the Europe allowance.

## Hetzner pricing

Every Hetzner dollar below is the USD column, excluding VAT, as of September 24, 2026, after the June 15, 2026 price adjustment. Treat the euro sticker as the wrong column for a US budget, and VAT still sits on top of either cap.

The orderable Europe entry on Regular Performance is CPX12, with 1 shared AMD vCPU, 2 GB of RAM, 40 GB NVMe, and 20 TB outbound. The monthly cap is in the table, or $0.0216 per hour if you destroy the box early.

Add the address and the marketing "starting from" figure on that family is $14.09/mo, the cap plus IPv4.

The US entry is CPX11: 2 shared AMD vCPUs, 2 GB, 40 GB, and 1 TB outbound. It uses the higher cap in the table, or $0.0328 per hour. With the primary IPv4 the monthly cap is $21.09, so price the address before comparing hosts that include it.

CPX21 in those cities is 3 vCPUs, 4 GB, 80 GB, and 2 TB at $37.49/mo before IPv4, the next step when 2 GB is tight. Europe CPX22 is 2 vCPUs, 4 GB, 80 GB, and 20 TB at $22.99/mo before IPv4.

Singapore's CPX12 is $17.99/mo before IPv4, with a smaller traffic bundle than Europe, so Asia is its own column.

Extra outbound is $1.20 per TB on the US and Europe rate those plan pages use, billed in 100 MB blocks once the included bundle is gone. Singapore's extra terabyte is higher on the same product.

A CX23 at $6.49/mo before IPv4, and a CAX11 at $6.99/mo before IPv4, were labeled not available, so do not treat them as orderable. The cost-optimized starting-from card is $7.09/mo, the CX23 cap plus IPv4, on a plan you may not be able to create.

DigitalOcean's 2 GiB, 2 vCPU plan is $18/mo with IPv4 included, and outbound past the pool is $0.01 per GiB. Passing the pool by 1,024 GiB costs $10.24, which is the overage a heavy month adds. Vultr's 1 GB plan is $5.00/mo, and the $2.50 row is IPv6 only, so an A record needs the higher row.

Akamai shared CPU starts at $5/mo for buyers whose risk is extra egress. Lightsail's Linux bundle with a public IPv4 is $5/mo, and outbound overage starts at $0.09 per GB, the steepest rate in this set.

OVHcloud's US VPS-1 starts at $4.54/mo with unlimited traffic, a bandwidth ceiling rather than a terabyte pool. UpCloud's Starter 1 GB plan is $3.5/mo with IPv4 included, a small disk with the address already in the price.

Plan | Price | Best for |

Hetzner EU CPX12 | $13.49/mo | Excl. IPv4 and VAT. 1 AMD vCPU, 2 GB, 40 GB, 20 TB out |

Hetzner US CPX11 | $20.49/mo | Ashburn and Hillsboro, excl. IPv4. 2 vCPU, 2 GB, 1 TB, $0.0328/hr |

Hetzner primary IPv4 | $0.60/mo | Excl. VAT. $0.0010/hr. Primary IPv6 is free |

Hetzner US CPX11 with IPv4 | $21.09/mo | Server cap plus one primary IPv4, excl. VAT |

Hetzner EU CPX22 | $22.99/mo | Excl. IPv4. 2 vCPU, 4 GB, 80 GB, 20 TB out |

Hetzner US CPX21 | $37.49/mo | Excl. IPv4. 3 AMD vCPUs, 4 GB, 80 GB, 2 TB out |

Hetzner CX23, listed | $6.49/mo | Excl. IPv4. Europe only. Page label: not available |

Hetzner CAX11, listed | $6.99/mo | Excl. IPv4. 2 Ampere cores, 4 GB. Labeled not available |

Hetzner extra outbound | $1.20/TB | US and EU plan-page rate. Billed in 100 MB blocks |

Hetzner EU CCX13 | $50.49/mo | Excl. IPv4. 2 dedicated vCPU, 8 GB, 80 GB |

Hetzner US CCX13 | $50.99/mo | Excl. IPv4. Same shape, 1 TB outbound in the US |

Hetzner Singapore CPX12 | $17.99/mo | Excl. IPv4. Smaller traffic bundle than the EU row |

Hetzner floating IPv4 | $3.50/mo | Address you can move. Not the primary IPv4 line |

DigitalOcean Basic 512 MiB | $4/mo | 1 vCPU, 10 GiB SSD, 500 GiB transfer, IPv4 included |

DigitalOcean Basic 1 GiB | $6.00/mo | 1 vCPU, 25 GiB, 1,000 GiB transfer |

DigitalOcean Basic 2 GiB, 2 vCPU | $18/mo | 60 GiB SSD, 3,000 GiB transfer, IPv4 included |

DigitalOcean outbound overage | $0.01/GiB | Inbound is free. Snapshots are $0.06/GB per month |

Vultr regular, IPv6 only | $2.50/mo | 1 vCPU, 0.5 GB, 10 GB disk, 0.50 TB |

Vultr regular 1 GB | $5.00/mo | 1 vCPU, 25 GB, 1 TB, $0.007/hr |

Vultr regular 2 GB, 2 vCPU | $15/mo | 65 GB disk, 3 TB bandwidth |

Akamai shared CPU | $5/mo | 1 GB size includes 1 TB. Extra egress $0.005/GB in most regions |

Akamai dedicated CPU | $36/mo | Starting price on the shared-versus-dedicated table |

OVHcloud US VPS-1 | $4.54/mo | Starts at. 2 vCores, 4 GB, 40 GB, unlimited traffic, 500 Mbps |

OVHcloud US VPS-2 | $8.50/mo | Starts at. 4 vCores, 8 GB, 75 GB, 1 Gbps |

UpCloud Starter 1 GB | $3.5/mo | IPv4 included, 10 GB disk, egress at zero under Fair Transfer |

Lightsail Linux with IPv4 | $5.00/mo | 0.5 GB, 2 vCPUs, 20 GB, 1 TB. Overage from $0.09/GB outbound |

## Hetzner pros and cons

### What we like

- The Europe CPX12 includes 20 TB outbound before IPv4, which is the row that still beats a metered US pool
- Hourly billing stops at a monthly cap, and incoming traffic plus same-zone traffic are free
- Firewalls and private networks have no extra charge, and primary IPv6 is free

### What could be better

- The US CPX11 includes 1 TB, so the Europe bargain is not the Ashburn price
- CX23 at $6.49/mo and the rest of the cost-optimized page were labeled not available on September 24, 2026
- A powered-off server and an unused primary IPv4 both keep billing until you delete them

## Who Hetzner is for

Hetzner fits a developer who can put the VM in Germany or Finland and who will delete boxes that are finished. Staging clusters and file hosts that would burn a metered transfer pool are where the Europe allowance still changes the month.

It is a weak default when users are in the United States and latency matters, because the cheap rows are not in Ashburn. It is also a weak default for object storage beside a US server, a supported Windows image, or a managed Postgres.

For the database, use the Postgres hosting comparison, and for a request-billed app, use the serverless platforms guide.

Devshot data: 27,887 developers are on this list, and the average open rate is 30.3%, from the audience file updated September 20, 2026. The next price-list change belongs in the daily Devshot brief.

## Best Hetzner alternatives

If Hetzner is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Hetzner | Developers who can place a Linux VM in Germany or Finland and care about egress | EU CPX12 $13.49/mo excl | Visit → |

DigitalOcean | Developers who want a US region, a public IPv4 in the sticker, and a monthly cap | From $4/mo (512 MiB) or $6.00/mo (1 GiB) | Visit → |

Vultr | Developers who need a specific city and will skip the IPv6-only row | From $2.50/mo (IPv6 only) or $5.00/mo (1 GB, 1 TB, $0.007/hr) | Visit → |

Akamai Cloud (Linode) | Teams whose risk is egress, and who want shared CPU with a published transfer bundle | Shared CPU from $5/mo | Visit → |

OVHcloud | Buyers who want 4 GB of RAM and unlimited traffic more than hourly cloud billing | US VPS-1 starts at $4.54/mo (2 vCores, 4 GB, unlimited traffic) | Visit → |

UpCloud | Quiet projects that want IPv4 in the price and egress listed at zero | Starter from $3.5/mo (1 GB, 10 GB disk, IPv4 included) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

DigitalOcean

The predictable Droplet bill, with transfer pooled across the team and inbound left off the meter.

Visit →

Vultr

A flat-rate cloud with a wide region list, a 672-hour monthly cap, and a per-gigabyte egress charge past the bundle.

Visit →

Akamai Cloud (Linode)

Linode's shared and dedicated plans, now on Akamai, with extra egress priced on the card.

Visit →

OVHcloud

A VPS card with a stated bandwidth cap and a one-day backup, priced from a US starting rate.

Visit →

UpCloud

A small-disk Starter plan with IPv4 included and no egress fee under a fair-use policy.

Visit →

## The bottom line

Order Hetzner in Germany or Finland when egress is the line you are trying to avoid. Price the CPX12 row plus the separate IPv4, and treat CX23 as listed until the not-available label comes off.

Do not approve an Ashburn budget from a post that still quotes the old Europe CX price. The US CPX11 cap plus IPv4 is the sheet number, and 1 TB is included before the $1.20 per TB rate.

Choose [DigitalOcean](https://toolradar.com/tools/digitalocean) when the region and the included IPv4 matter more than the Europe allowance, and read the DigitalOcean review first.

Choose [Vultr](https://toolradar.com/tools/vultr) when the city matters and you will skip the IPv6-only row.

Choose [Akamai](https://toolradar.com/tools/linode) when extra egress at $0.005 per GB is the risk. Choose [OVHcloud](https://toolradar.com/tools/ovhcloud) when 4 GB and unlimited traffic are worth a VPS card.

Choose [UpCloud](https://toolradar.com/tools/upcloud) when a tiny disk and zero egress fit.

A cluster belongs on the Kubernetes tools guide, and a frontend host on the Vercel review. Apps that assume a cheap European box are in the self-hosting tools guide.

Cite this: Devshot, "Hetzner Cloud Review 2026", September 2026.

## Frequently asked questions

Is Hetzner Cloud worth it in 2026?

Yes, if the server can live in Germany or Finland and outbound traffic is a real part of the bill. The EU CPX12 includes 20 TB and is priced before a separate IPv4, so add the address before you compare. It is a weaker buy for a new Ashburn or Hillsboro server, because that CPX11 cap is much higher and includes 1 TB. The cheaper CX plans were labeled not available on the cost-optimized page on September 24, 2026.

How much does Hetzner Cloud cost?

In USD, excluding VAT, a new Europe CPX12 is $13.49/mo before IPv4, and a new US CPX11 is $20.49/mo before IPv4, or $21.09/mo with one primary IPv4. Those caps were verified on Hetzner's price list in September 2026. Hourly use is rounded up to one hour and cannot exceed the monthly cap, so a short-lived box costs less than a full month. Servers ordered before June 15, 2026 keep the older price until you rescale them or move them into a project with a different currency.

Does Hetzner charge for IPv4?

Yes, primary IPv4 is $0.60/mo excluding VAT, billed even when the address is not attached to a server. You delete the address to stop that line, and primary IPv6 is free. A server needs at least one primary IP before it has a public interface. A floating IPv4, which you can move between servers in the same network zone, is $3.50/mo and sits on top of the primary.

Is Hetzner cheaper than DigitalOcean?

For a Europe server with heavy outbound traffic, usually yes. The EU rows include 20 TB, and extra traffic on those plan pages is $1.20 per TB, so a large transfer month stays predictable. For a small US server, often no, because DigitalOcean's 2 GiB, 2 vCPU Basic plan is $18/mo with IPv4 included and 3,000 GiB of transfer. The Hetzner US CPX11 cap plus IPv4 is $21.09/mo with 1 TB, so compare those shapes rather than the oldest Europe sticker.

What are the best Hetzner alternatives?

DigitalOcean, when you want a published US Droplet with IPv4 inside the price. Vultr, when the city matters and you avoid the $2.50 IPv6-only plan, because that row cannot serve an A record. Akamai Cloud, when egress overage is the line that would dominate the bill. OVHcloud, when the US VPS-1 start of $4.54/mo and unlimited traffic fit. UpCloud, when the Starter plan at $3.5/mo and zero egress fit a small disk. The full sheet is the VPS hosting guide on Devshot.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Hetzner pricing](https://hetzner.com)checked Sep 2026
- [DigitalOcean pricing](https://digitalocean.com/pricing)checked Sep 2026
- [Vultr pricing](https://vultr.com/pricing)checked Sep 2026

Related guides

Vps Hosting For DevelopersDigitalocean ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Lovable Review

URL: https://devshot.co/reviews/lovable-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Honest Lovable review for developers and technical founders: what the AI app builder actually generates, its Supabase-backed stack, real credit pricing…

Review

## Lovable Review

An AI app builder that turns a chat prompt into a deployed full-stack product, with a real database, auth and export-ready code, priced by a credit meter rather than a flat monthly fee.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Lovable is worth it in 2026 if you want a real, deployable full-stack app from a chat prompt, not just a UI mockup. The free plan caps out at 30 build credits a month, and Pro starts at $25/month for 100 monthly credits, with Business at $50/month for teams that need SSO.

Its edge over rivals is the output: a working React frontend, Supabase-backed backend, auth and a database, as real editable code you can export to GitHub. The catch is the credit meter: a single feature request can burn more than a credit, and a busy building week means a top-up or an upgrade. If you already write code and mainly want fast UI generation, Bolt or v0 start at a lower monthly price with more direct control.

Key factsBest for: Non-technical founders and developers who want a real…

- Updated: September 25, 2026
- Best for: Non-technical founders and developers who want a real full-stack app, not a mockup, from a chat prompt.
- Price as of September 25, 2026: From $25/mo (Pro, 100 monthly credits; $21/mo billed annually); free plan capped at 30 build credits a month.

- An AI app builder that outputs a working product with a real database, not just a UI.
- Headquarters: Stockholm, Sweden
- Free plan: Yes, capped at 30 build credits a month
- Backend: Supabase (via Lovable Cloud)
- Alternatives covered: Bolt, v0, Replit, Cursor

Pros

- Generates a working frontend, Supabase-backed backend, auth and database together, as real exportable code.
- GitHub sync keeps a real repository in step, so a developer can take over at any point.
- Unlimited workspace members on every paid plan; the bill scales with credits, not seats.

Cons

- Credit pricing is a floor, not a ceiling: a landing page built with images can run 2 credits per message, and a busy week needs a top-up.
- No free plan for a serious project, just a capped 30 credits a month that covers a small prototype.
- The AI makes its own architecture calls, so course-correcting a large generated app takes real review time.

HeadquartersStockholm, Sweden

Free planYes, capped at 30 build credits a month

BackendSupabase (via Lovable Cloud)

Pricing modelCredit-based, not per seat

Lovable pitches itself as an AI software engineer: describe an app in plain English in the chat box, and it writes a working product, not a slide of components.

Built by Lovable Labs, headquartered in Stockholm, Sweden, it has become one of the names developers hear most in the vibe coding conversation, alongside Bolt, v0 and Replit.

The real test for a developer, not just a non-technical founder, is what happens to the code after generation: can you own it, extend it, and hand it to a team?

This review covers what Lovable builds and how the credit system prices that work. It also looks at where the Supabase-backed backend helps and where it gets in the way, and how Lovable stacks up against Bolt, v0, Replit and Cursor for a developer deciding where to prototype next.

Non-technical founders and product people are the audience Lovable is built for. Engineers get more mileage treating it as a fast first draft.

## What is Lovable?

Lovable is a web-based AI app builder made by Lovable Labs, a company headquartered in Stockholm, Sweden.

You describe an app or website in a chat interface, and it generates a working product: a React frontend, a Supabase-backed backend with a Postgres database, authentication and file storage, deployed to a lovable.app subdomain or a custom domain.

Unlike a drag-and-drop no-code tool, the output is real, exportable code you can push to GitHub and keep editing outside Lovable.

Beyond the chat-to-app flow, Lovable adds design system support so generated interfaces stay visually consistent, plus connectors for services like Stripe and other third-party APIs. It also runs an MCP server for wiring in external tools, and Lovable Cloud, its managed hosting and backend layer that sits on Supabase as the infrastructure provider.

Business and Enterprise plans add single sign-on, role-based access, a security center and audit logs for teams. It sits closer to a junior full-stack engineer than a website template tool.

## How Lovable works

A Lovable project starts with a prompt: describe the app, and Lovable scaffolds pages and wires routing. If the app needs one, it also provisions a Supabase database and auth behind the scenes. Every follow-up message, from "make the button gray" to "add authentication with sign up and login," draws credits from your plan's monthly balance.

Cost is priced by how much work the change actually requires. A cosmetic tweak runs about half a credit; a full authentication flow runs just over one.

The visual editor lets you click an element and describe a change instead of writing a prompt from scratch. Plan Mode switches Lovable into a step-by-step planning flow for larger features, at a flat rate per message.

GitHub sync keeps a real repository in step with every change, so a developer can open the code in a normal editor at any point and pick up where the chat left off.

Once code leaves Lovable, hosting choices like Vercel, Netlify or Firebase take over.

The rough edge is the credit meter itself: free accounts get a daily grant of five build credits, capped at 30 a month. That covers a small prototype and not much else before you hit a wall.

## Lovable key features

Chat-to-app generationEssential
Describe an app in plain English and Lovable scaffolds a working product: pages, routing, a database schema and authentication, not just static screens. It is the core loop the whole product is built around.

Supabase-backed backend (Lovable Cloud)Essential
Generated apps get a real Postgres database, authentication and file storage through Lovable Cloud, which runs on Supabase as the infrastructure provider. Most prototypes need no separate backend setup.

Credit-based usage pricingEssential
Every message costs credits, with the rate tied to how much work the change requires: a small style tweak costs a fraction of a credit, a new feature with auth can cost more than one. The monthly plan sets the credit balance, not a fixed number of messages.

Visual edit mode
Click an element in the preview and describe a change instead of writing a prompt from scratch, useful for small styling fixes without burning a full chat turn.

GitHub sync and exportable code
Every change syncs to a real GitHub repository, so a developer can open the generated code in a normal editor, review diffs and keep working outside Lovable at any point.

Design systems and templates
Pro and Business plans add design system support so generated screens stay visually consistent across a larger app; Business and Enterprise add a template library for common app types to start from.

## Lovable pricing

Lovable, Bolt, v0, Replit and Cursor all publish list prices, checked directly on each vendor's pricing page, September 2026. Enterprise is a custom quote at all five. Lovable and v0 price against a shared credit balance rather than per seat, so inviting more workspace members does not raise the bill on its own.

The cheapest credible paid entry is Replit Core or Cursor Pro, both $20 a month. Lovable and Bolt Pro start at the same price as each other, one step up. Team tiers cost the most at v0 and Replit. Cursor is the only one of the five that does not generate a deployed app end to end, since it edits an existing codebase instead.

Devshot data: [Toolradar's AI app builders guide](https://toolradar.com/guides/best-ai-app-builders), last refreshed in May 2026 and still live today, tracks 153 tools in the category. It lists Lovable, Bolt, v0 and Replit among six top picks.

See each tool's full profile on Toolradar: [Lovable](https://toolradar.com/tools/lovable), [Bolt](https://toolradar.com/tools/bolt), [v0](https://toolradar.com/tools/v0), [Replit](https://toolradar.com/tools/replit) and [Cursor](https://toolradar.com/tools/cursor).

Plan | Price | Best for |

Lovable Pro | $25/mo | 100 monthly credits, unlimited users, custom domains |

Lovable Pro (annual) | $21/mo billed annually | Same as Pro, billed yearly |

Lovable Business | $50/mo | 100 monthly credits plus SSO, role-based access, priority support |

Lovable Business (annual) | $42/mo billed annually | Same as Business, billed yearly |

Lovable Enterprise | Custom platform fee | Volume-based pricing, SCIM, audit logs, custom SLA |

Bolt Pro | $25/mo | 10M tokens to start, no daily cap, custom domains |

Bolt Teams | $30/mo per member | Centralized billing, team access management |

Bolt Enterprise | Custom quote | SSO, audit logs, dedicated account manager |

v0 Plus | $30/mo per user (promo; list $90/mo) | Monthly credit pool per user, all models, team billing |

v0 Business | $100/user/mo | Training opt-out by default, priority support |

v0 Enterprise | Custom quote | SAML SSO, role-based access, guaranteed SLA |

Replit Core | $20/mo | AI Agent, up to 60 Free Mode projects, unlimited workspaces |

Replit Core (annual) | $18/mo billed annually | Same as Core, billed yearly |

Replit Pro | $100/mo | 10 parallel agents, up to 15 collaborators, 28-day database rollback |

Replit Pro (annual) | $90/mo billed annually | Same as Pro, billed yearly |

Replit Enterprise | Custom quote | SSO/SAML, single-tenant environments, custom seat limits |

Cursor Pro | $20/mo | Extended Agent limits, frontier model access, MCP support |

Cursor Pro Plus | $60/mo | Higher usage pool for daily agent users |

Cursor Ultra | $200/mo | Largest usage pool, for agent power users |

Cursor Teams (Standard) | $40/user/mo | Centralized billing, SAML/OIDC SSO, Bugbot reviews |

Cursor Teams (Premium) | $120/user/mo | 5x the Standard seat's agent usage limits |

Cursor Enterprise | Custom quote | Pooled usage, SCIM, audit logs, priority support |

## Lovable pros and cons

### What we like

- Generates a working frontend, Supabase-backed backend, auth and database together, as real exportable code.
- GitHub sync keeps a real repository in step, so a developer can take over at any point.
- Unlimited workspace members on every paid plan; the bill scales with credits, not seats.

### What could be better

- Credit pricing is a floor, not a ceiling: a landing page built with images can run 2 credits per message, and a busy week needs a top-up.
- No free plan for a serious project, just a capped 30 credits a month that covers a small prototype.
- The AI makes its own architecture calls, so course-correcting a large generated app takes real review time.

## Who Lovable is for

Lovable is a strong fit for non-technical founders, product managers and solo builders who need a real, working product fast. They would rather pay a subscription than hire a developer for the first version.

It also suits developers who want to skip repetitive full-stack setup: scaffolding auth, a database and CRUD screens in minutes, then taking the exported code the rest of the way. Teams standardizing on Supabase get the most out of Lovable Cloud, since the backend maps directly onto infrastructure they likely already use.

Who should skip it: engineers who want fine-grained control over architecture from the first message, since the AI still makes its own structural choices you may need to unwind later.

Budget-conscious solo developers comfortable writing their own UI get more raw capability per dollar from Bolt or v0. Teams that want AI woven into an existing codebase, rather than a new app generated from scratch, are better served by an editor like Cursor or GitHub Copilot.

Heavy builders should budget beyond the sticker price: the credit meter, not the plan name, decides what a busy month actually costs.

## Best Lovable alternatives

If Lovable is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Lovable | Non-technical founders and developers who want a real full-stack app, not a mockup, from a chat prompt. | From $25/mo (Pro, 100 monthly credits | Visit → |

Bolt | Developers who want a fast browser prototype with full control of the running code and no daily cap once paid. | From $25/mo (Pro, 10M tokens to start) | Visit → |

v0 | Teams already on React, Next.js and Vercel who want generated UI and apps that drop straight into their stack. | From $30/mo (Plus, per user | Visit → |

Replit | Builders who want one tool for building, the database and hosting, without stitching services together. | From $20/mo (Core, $18/mo billed annually) | Visit → |

Cursor | Developers with an existing codebase who want AI woven into the project, not a new app generated from scratch. | From $20/mo (Pro) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Bolt

StackBlitz's AI builder, fast in the browser with token-metered pricing and full code control.

Visit →

v0

Vercel's AI builder, tuned for teams that will deploy the output straight to Vercel.

Visit →

Replit

A full browser IDE with an AI Agent, its own database and hosting, in one workspace.

Visit →

Cursor

An AI-first code editor for engineers extending real projects, not generating new ones from a prompt.

Visit →

## The bottom line

Lovable earns its place for anyone who wants a real, working product out of a chat prompt rather than a mockup. The Supabase-backed backend, exportable code and GitHub sync mean the output is a genuine starting point, not a dead end.

Pro is priced the same as Bolt's entry plan, a reasonable ask for a solo founder or a developer skipping repetitive setup work.

The trade-off is the credit meter: the sticker price is a floor. Heavy building weeks force a top-up or an upgrade to Business at $50/month, and the free plan's 30 credits do not stretch far.

Choose Bolt if you want more headroom per dollar and full code control, v0 if your stack is already Next.js and Vercel, and Replit if you want the database and hosting bundled into one IDE. Pick Cursor or Windsurf instead if you are extending an existing codebase rather than generating a new one. Buy Lovable when the credit cost buys a working app faster than building the scaffolding yourself.

Methodology: prices, plans and features were checked directly on each vendor's pricing and product pages, September 2026. The four alternatives are the tools developers most often compare against Lovable in the AI coding and app-builder categories, not a paid list. None of the five carries a sponsored placement on this page.

Cite this: Devshot, "Lovable Review 2026: Pricing, Pros, Cons and Alternatives," September 2026.

## Frequently asked questions

How much does Lovable cost?

Lovable's free plan gives 30 build credits a month, enough for a small prototype. Pro is the cheapest paid plan, matching Bolt's entry price ($21/month if billed annually), and covers 100 monthly credits for unlimited workspace members. Business costs $50/month ($42/month billed annually) for the same 100 credits plus SSO, role-based access and priority support. Enterprise is a custom, volume-based platform fee. Credits are not sold per seat, so adding people to a workspace does not raise the price on its own.

Is Lovable worth it in 2026?

For a non-technical founder or a developer who wants a working full-stack app fast, yes: the output is real, exportable code with a Supabase-backed database and auth already wired up, not just a UI mockup. For an engineer who wants tight control over architecture from the first message, or a budget-conscious builder who writes their own UI, Bolt or v0 give more capability per dollar.

Does Lovable have a free plan?

Yes, but it is capped tightly: 5 build credits a day up to 30 a month, plus a monthly grant of Cloud credits for hosting a small app. That covers a first prototype, not ongoing development. Bolt's free plan is more generous on raw usage, with a monthly token allowance and no daily cap, while Lovable's free tier runs out faster on complex builds.

What are the best Lovable alternatives?

Bolt is the closest match on price and output, with no daily token cap once you are on Pro. v0 fits teams already building on Next.js and Vercel. Replit bundles the database and hosting into a full browser IDE with database rollback. Cursor is a different category entirely, an AI-first editor for extending an existing codebase rather than generating a new app.

What tech stack does Lovable generate?

Lovable outputs a React frontend with a Supabase-backed backend by default, including a Postgres database, authentication and file storage through Lovable Cloud. The code is real and exportable to GitHub, so you can keep editing it in a normal editor or hand it to a development team once the prototype outgrows the chat interface.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Lovable pricing](https://lovable.dev/pricing)checked Sep 2026
- [Bolt pricing](https://bolt.new/pricing)checked Sep 2026
- [v0 pricing](https://v0.dev/pricing)checked Sep 2026
- [Replit pricing](https://replit.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026

Related guides

Vibe Coding ToolsCursor ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Neon Review

URL: https://devshot.co/reviews/neon-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Neon is worth it in 2026 when Postgres should sleep between deploys, and a weak fit when the bill has to stop itself.

Review

## Neon Review

Worth it when Postgres should sleep between deploys. A weak default when the bill has to stop itself.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Neon is worth it in 2026 when the database should sleep between deploys and each pull request needs its own Postgres branch. Launch has no monthly minimum: compute is $0.106 per CU-hour, verified on Neon's pricing page in September 2026. Free covers a monthly compute allowance per project, then compute suspends until the next month.

The catch is that an email alert is not a shutoff, so the bill can keep climbing after you have been warned. Anything above 16 CU cannot scale to zero, so that size never sleeps. If you need auth, files, and a cap that refuses more usage, start with the Supabase review instead of treating Neon as a full backend.

Key factsBest for: Preview databases and apps whose Postgres can sleep between…

- Updated: September 25, 2026
- Best for: Preview databases and apps whose Postgres can sleep between deploys
- Price as of September 25, 2026: Launch $0.106/CU-hour, no monthly minimum; Free includes 100 CU-hours.

- Serverless Postgres that branches like a repo and bills compute only while the database is awake.
- Parent: Databricks
- Free compute: 100 CU-hours per project
- Paid plans: Launch and Scale, no minimum
- Alternatives covered: Supabase, Amazon Aurora Serverless, PlanetScale, Railway, Render

Pros

- Free and Launch include 10 branches per project, and a new child branch stores nothing until you write to it.
- Compute suspends after five idle minutes unless you turn that off, so a bursty app pays for hours awake.
- The pooler, pgvector, PostGIS, and TimescaleDB are on every plan, including Free, so you can try them before you pay.

Cons

- Past the compute allowance or 5 GB of egress, Free suspends compute until next month, and writes block past 1 GB, so the app stops instead of billing you.
- Spend emails at 80% and 100% do not stop queries, and computes above 16 CU cannot scale to zero.

ParentDatabricks

Free compute100 CU-hours per project

Paid plansLaunch and Scale, no minimum

Best forBranching Postgres that can sleep

Neon is the Postgres I would attach to a preview app, and the one I would not hand a finance team that wants a ceiling the platform enforces. Launch and Scale bill what ran, which fits a database that sits idle most of the day and gets expensive once you force it to stay awake.

A compute unit is about 4 GB of RAM, so that hourly rate buys a larger machine than an Aurora ACU. Storage, history, and extra branches keep billing after compute has suspended, which is what a sleeping database still costs.

Toolradar data: the [September 2026 managed databases ranking](https://toolradar.com/best/managed-databases) evaluated 23 tools.

Its buyer note names [Neon](https://toolradar.com/tools/neon), [Supabase](https://toolradar.com/tools/supabase), and [PlanetScale](https://toolradar.com/tools/planetscale) beside the hyperscalers.

[Railway](https://toolradar.com/tools/railway) and [Render](https://toolradar.com/tools/render) have directory pages too.

How we compared: each price was read on the vendor's own page on September 24, 2026. Neon was checked on the pricing page and the plans document. Aurora figures are Amazon's US East (Northern Virginia) examples. PlanetScale figures are the AWS us-east-1 catalog.

## What is Neon?

Neon is Databricks' serverless Postgres, sold on the pricing page as Lakebase Postgres: a project, branches, and compute that can suspend. It speaks ordinary Postgres, so your ORM still works and a dump is still the exit if you need to leave.

Branching, the pooler, autoscaling, and extensions such as pgvector, PostGIS, and TimescaleDB are on every plan, including Free, so a side project can test a vector column or a second branch without upgrading. Auth, object storage, and functions are still marked beta, so do not leave a working backend to chase them.

Managed Better Auth caps Free at 60,000 monthly active users, enough for a demo and the wrong cap for a consumer app. Storage and functions are not charged during beta, so they are not the invoice to forecast.

If the job is auth plus files plus a database, that is a backend-as-a-service decision, and Supabase is the closer product.

## How Neon works

You create a project, which arrives with a root branch. Child branches are copy-on-write: storage stays at zero until that branch is written, then bills those writes, capped at the logical size of the data.

Compute is CU-hours, size times hours running, so a 0.25 CU box for four hours is one CU-hour and a suspended compute adds nothing. Sleeping previews stay cheap, and warm production databases do not.

On Free you cannot turn suspend off, and the default is five idle minutes. Neon says the next query wakes the compute in a few hundred milliseconds, which a preview can absorb and a user-facing request should not.

Launch can disable suspend, and Scale can set the window from one minute through always on.

Computes larger than 16 CU never suspend, and Scale's fixed sizes go to 56 CU, so those sizes never get the idle rate. Logical replication keeps compute active while a subscriber is connected, so that pipeline stays on the meter and a sync can erase the savings.

A read replica is a second compute bill, not a free feature. The database clients list is the SQL tool you keep beside this.

## Neon key features

Database branchingEssential
Free and Launch include 10 branches per project, and Scale includes 25, so count pull requests before you call branching free. Extra branches on paid plans are a branch-month fee, and child storage stays at zero until that branch is written.

Scale to zeroEssential
Inactive compute suspends after five minutes on Free and Launch. Free cannot disable it, while Launch can, and Scale can choose from one minute through always on. Anything above the suspend ceiling stays up and keeps billing.

Instant restore historyEssential
Restore runs from root branches only, so a child branch is not your production backup. Free keeps six hours, capped at 1 GB, with no history charge. Paid plans default to one day, up to seven days on Launch or 30 days on Scale.

Autoscaling with a ceiling
Free scales up to 2 CU, about 8 GB of RAM, enough for a preview and tight for a busy API. Launch scales up to the suspend ceiling, and Scale can fix a size past it. Capping the max CU is what bounds the compute line.

Pooling and extensions
Connection pooling, pgvector, PostGIS, and TimescaleDB are included on every plan, so Free is enough to try them before you pay. A read replica is still another compute you pay for.

Beta auth and storage
Managed Better Auth, object storage, and functions sit beside Postgres and are marked beta, with no beta charge for storage and functions. Treat them as a preview of a backend, not as the reason to leave a stack that already has auth.

## Neon pricing

Neon publishes no seat price on Launch or Scale, so headcount is not the meter. You pay CU-hours while compute is running, then storage, history, and branches past the included count. Invoices under $0.50 are not collected, so a tiny preview may never charge the card.

Neon's own light Launch example totals $2.31: about 10 CU-hours, 2 GB on the root, 1 GB on a child branch, and 1 GB of history. That total is the bill when the database mostly sleeps. Their Scale example, 4 CU for roughly 17 billable days plus storage and history, is $431.15 for a database left warm for most of the month.

Free is an allowance, then a wall, not a plan that starts charging. Past the compute allowance or 5 GB of egress, compute suspends until the next billing period, so the demo goes dark instead of invoicing you. Past the storage allowance, writes block and the data is not deleted.

Paid egress includes 500 GB per project before the table's per-GB rate, and a branch can grow to 16 TB.

Supabase Pro is an organization fee, and the compute credit covers one Micro, so a second always-on project is another line. Spend caps are on by default there, which is the brake Neon does not offer.

Aurora Standard in US East (Northern Virginia) bills $0.12 per ACU-hour, and an ACU is about 2 GiB, half the memory of a Neon CU. PlanetScale's cheapest single-node Postgres SKU in AWS us-east-1 is a monthly cluster, paid even when idle.

Railway does not bill CPU while a service is stopped, though the volume keeps charging. Render sells a named size that stays on, so you pay for the reserved month rather than hours queried.

The longer host comparison is Postgres hosting, and if the app host is the purchase, use serverless platforms and the Vercel review.

Plan | Price | Best for |

Neon Free | No monthly fee | 100 CU-hours, 1 GB, and 5 GB egress per project |

Neon Launch compute | $0.106/CU-hour | No monthly minimum. Autoscaling up to 16 CU. |

Neon Scale compute | $0.222/CU-hour | SLA, private networking, SOC 2. Fixed sizes to 56 CU. |

Neon storage | $0.35/GB-month | Launch and Scale. Free stops writes at 1 GB. |

Neon extra branch | $1.50/branch-month | Past 10 branches on Launch, past 25 on Scale |

Neon instant restore | $0.20/GB-month | Up to 7 days on Launch, 30 days on Scale |

Neon public egress | $0.10/GB | After 500 GB per project on Launch and Scale |

Neon snapshot storage | $0.09/GB-month | Manual and scheduled snapshots on paid plans |

Supabase Pro | $25/mo | Org fee. $10 compute credit covers one Micro. |

Supabase extra Micro | $10/mo | Each project beyond the included credit |

Supabase branching | $0.01344/branch-hour | Pro and above. Not on the free plan. |

Supabase Team | $599/mo | SOC 2 and ISO 27001. HIPAA is an add-on. |

Aurora Serverless, Standard | $0.12/ACU-hour | US East (Northern Virginia). 1 ACU is about 2 GiB. |

Aurora Standard storage | $0.10/GB-month | US East example. I/O is a separate Standard charge. |

Aurora I/O-Optimized compute | $0.156/ACU-hour | Same region. Read and write I/O charges drop to zero. |

PlanetScale PS-5 | $5/mo | Single-node Postgres, 512 MiB, AWS us-east-1 |

PlanetScale PS-5 HA | $15/mo | Three nodes, same size, same region |

Railway Hobby | $5/mo | Includes a matching usage credit. Not per seat. |

Railway memory | $10/GB-month | Per second. Stopped services are not billed for RAM. |

Railway CPU | $20/vCPU-month | Same meter. A warm database pays CPU and memory. |

Railway volume | $0.15/GB-month | Disk keeps billing when the service is stopped |

Render 0.1c-256mb | $6/mo | 256 MB RAM, 1 GB storage included, stays on |

Render Postgres disk | $0.30/GB | Expandable storage on top of the included gigabyte |

## Neon pros and cons

### What we like

- Free and Launch include 10 branches per project, and a new child branch stores nothing until you write to it.
- Compute suspends after five idle minutes unless you turn that off, so a bursty app pays for hours awake.
- The pooler, pgvector, PostGIS, and TimescaleDB are on every plan, including Free, so you can try them before you pay.

### What could be better

- Past the compute allowance or 5 GB of egress, Free suspends compute until next month, and writes block past 1 GB, so the app stops instead of billing you.
- Spend emails at 80% and 100% do not stop queries, and computes above 16 CU cannot scale to zero.

## Who Neon is for

Neon fits a developer who wants a branch per pull request and can tolerate a few hundred milliseconds when a preview wakes. Side projects fit Free until they outgrow the allowance and have to move. Launch fits a startup whose traffic is bursty enough that the database still sleeps.

It is the wrong default when the database must stay warm, because disabling suspend is a Launch switch and a large compute cannot suspend at all. It is also the wrong default when a compliance report has to be in hand before Scale, or when auth and file storage are the product rather than a beta beside Postgres.

Skip Neon when you need a cap that refuses traffic. Launch and Scale email you at 80% and 100% of a threshold you set, and the query still runs, so the alert does not stop the bill.

Supabase's spend cap is the closer brake, and a phone client that needs offline sync belongs on the Firebase review, not on a Postgres branch.

A VM you patch yourself is a different invoice, covered in the DigitalOcean review. The Supabase vs Firebase note is the backend fork if Neon was only in the running because it said Postgres.

Devshot data: this newsletter goes to 27,887 developers, and the audience file updated September 20, 2026 shows a 30.3% average open rate. Subscribe free if you want the next meter change in the inbox.

## Best Neon alternatives

If Neon is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Neon | Preview databases and apps whose Postgres can sleep between deploys | Launch $0.106/CU-hour, no monthly minimum | Visit → |

Supabase | Teams that want auth, storage, and a spend cap on Postgres | From $25/mo (Pro) | Visit → |

Amazon Aurora Serverless | AWS accounts that already run Aurora and can set the ACU floor | From $0.12/ACU-hour in US East (Northern Virginia) | Visit → |

PlanetScale | A sized Postgres cluster in Virginia, billed as a monthly SKU | PS-5 single node $5/mo in AWS us-east-1 | Visit → |

Railway | Apps that should stop paying CPU when the service is stopped | Hobby $5/mo with a matching usage credit | Visit → |

Render | A named Postgres size that stays on, with storage you can expand | 0.1c-256mb $6/mo | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 6 do not publish a comparable monthly price and are left out rather than estimated.

Supabase

Postgres plus auth, files, and edge functions, with a Docker install if the hosted plan stops fitting.

Visit →

Amazon Aurora Serverless

Aurora capacity that scales in ACU-hours, with storage and Standard I/O billed beside the compute.

Visit →

PlanetScale

Postgres sold as a cluster size in us-east-1, with a three-node HA row above the single node.

Visit →

Railway

Usage-priced compute where a stopped service drops CPU and memory, and the volume does not.

Visit →

Render

Managed Postgres at a fixed monthly size, with a gigabyte included and a 30-day free row.

Visit →

## The bottom line

Neon is worth paying for when branches and suspend are why you are here, and when someone will watch CU-hours. Start on Free, then move to Launch before a customer depends on it, because the free wall suspends compute instead of billing you.

Choose Supabase when auth and a spend cap matter more than copy-on-write branches. Choose Aurora when the account is already AWS and you can accept a separate storage line.

Choose [PlanetScale](https://toolradar.com/tools/planetscale) for a monthly cluster in us-east-1. Choose Railway when a stopped service should drop CPU, and [Render](https://toolradar.com/tools/render) for a Postgres plan that stays up.

The Devshot brief is the short version of this bill. Subscribe free if you want the next pricing note in the inbox.

Cite this: Devshot, "Neon Review 2026", September 2026.

## Frequently asked questions

Is Neon worth it in 2026?

Yes, when you want Postgres branches for previews and compute that can suspend after it sits idle. Launch bills by the CU-hour while the database is awake, which suits a quiet month and punishes one you leave warm. It is the wrong default when the database must stay warm, when you need a spend cap that refuses further usage, or when SOC 2 has to be on the contract before Scale. Free stops serving once the allowances are used up, rather than charging you. Prices were verified on Neon's pricing page in September 2026.

How much does Neon cost?

Free has no monthly fee and includes the compute, storage, and egress allowances on that plan. Launch compute is $0.106 per CU-hour and Scale compute is $0.222 per CU-hour, with no monthly minimum on either plan. Storage on both paid plans is $0.35 per GB-month, instant restore history is $0.20 per GB-month, and an extra branch past the allowance is $1.50 per branch-month. Neon's light Launch example totals $2.31, which is a sleeping preview rather than a production month. Invoices under $0.50 are not collected, so a bill that small never hits the card.

Does Neon have a free plan?

Yes, Free is not a trial and does not require a card. Each project gets 100 CU-hours, which Neon says covers a 0.25 CU compute for 400 hours, plus 1 GB of storage, 5 GB of egress, and 10 branches. Scale to zero is fixed at five minutes and cannot be disabled, so a free database sleeps even if you wanted it warm. Hitting the CU-hour or egress cap suspends compute until the next billing period, and hitting the storage cap blocks writes. Neither case deletes the data, so the project survives and the app does not.

How does Neon compare with Supabase?

Neon wins when the purchase is the database: branches on the free plan, compute that can sleep, and a bill that follows hours awake. Supabase wins when the purchase is the backend, because auth, storage, and edge functions are the hosted product, and Pro puts a spend cap on by default. Supabase Pro is $25 a month and includes a compute credit for one Micro instance. Its free projects pause after a week idle, and database branching on Pro is $0.01344 per branch-hour, so a preview you leave open keeps metering.

How does Neon compare with Aurora Serverless?

Both scale compute with load and bill by the hour. A Neon CU is about 4 GB of RAM and suspends when idle unless you disable it. Aurora PostgreSQL serverless in US East (Northern Virginia) prices Standard at $0.12 per ACU-hour, and an ACU is about 2 GiB, so the stickers are not the same size of machine. Aurora's feature table lists a starting capacity of 0 ACU, while the published US East example still starts its math at 0.5 ACU, so idle may not be free. Storage in that example is $0.10 per GB-month on Standard, with I/O charged separately unless you move to I/O-Optimized.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Neon pricing](https://neon.tech/pricing)checked Sep 2026
- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026
- [PlanetScale pricing](https://planetscale.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026

Related guides

Supabase ReviewPostgres HostingBackend As A ServiceServerless PlatformsFirebase ReviewVercel ReviewDatabase ClientsDigitalocean ReviewHeroku ReviewFree Database HostingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# OpenAI Codex Review

URL: https://devshot.co/reviews/openai-codex-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: OpenAI Codex review 2026: what Plus, Pro, and Go actually include, why the top Pro tier stopped taking new sign-ups, and when Claude Code, Cursor…

Review

## OpenAI Codex Review

Worth it once a ChatGPT seat should also cover the agent in your terminal and IDE. A weaker default when you want an editor built for the job or a cheaper standalone plan.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

OpenAI Codex is worth it in 2026 if you already pay for ChatGPT and want that same seat to run an agent in your terminal, IDE, and browser. Plus includes it for $20/mo, Go covers light tasks for $8/mo, and Pro starts from $100/mo for five times the Plus usage. OpenAI paused new sign-ups and upgrades to its top Pro tier on September 10, 2026, verified on OpenAI's help center this month.

Codex shares one allowance with ChatGPT Work, so a long coding run can eat into chat too. If the agent should live beside a code editor built around it instead of a chat app, Cursor or [Claude Code](https://toolradar.com/tools/claude-code) are the closer fit.

Key factsBest for: ChatGPT subscribers who want one agent across CLI, IDE…

- Updated: September 25, 2026
- Best for: ChatGPT subscribers who want one agent across CLI, IDE, web, and iOS.
- Price as of September 25, 2026: Free on ChatGPT Free; Go is $8/mo; Plus $20/mo; Pro from $100/mo (top tier sign-ups paused).

- OpenAI's coding agent, bundled into every ChatGPT plan instead of sold as its own SKU.
- Vendor: OpenAI
- Free plan: Yes, capped usage
- Cheapest paid entry: Go, $8/mo
- Alternatives covered: Claude Code, Cursor, GitHub Copilot, Antigravity CLI (formerly Gemini CLI)

Pros

- Plus covers the CLI, IDE extension, web app, and iOS, plus cloud code review and Slack, in one seat, no separate coding SKU.
- Pro offers 5x or 20x Plus usage, and eligible users can buy extra credits instead of upgrading plans outright.
- An API key still runs the CLI for CI work, and Go is the cheapest paid entry among all five tools compared in this review.

Cons

- OpenAI paused new sign-ups and upgrades to its top $200 Pro tier on September 10, 2026, with no reopening date announced.
- Plus and Pro carry no annual discount, unlike Claude Pro's cheaper annual rate for a comparable single seat.
- GPT-6 Astra costs 250 credits per million input tokens against Luna's 2.5, so reasoning-heavy sessions burn Pro credits fastest.

VendorOpenAI

Free planYes, capped usage

Cheapest paid entryGo, $8/mo

Best forExisting ChatGPT subscribers

Codex is not a separate purchase. It rides inside whichever ChatGPT plan you already hold, so the real decision is which tier gives you enough Codex usage for the week you actually work, not whether to buy a new tool.

OpenAI does not publish a fixed message count for any tier, because a one-line fix and a repository migration draw down the allowance at very different rates.

Toolradar data: the [September 2026 AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 153 tools, and [Codex](https://toolradar.com/tools/openai-codex) sits alongside [Claude Code](https://toolradar.com/tools/claude-code) and [Cursor](https://toolradar.com/tools/cursor) in that directory. The closer, head-to-head numbers below are checked against each vendor directly.

How we compared: pricing and product pages for OpenAI Codex, Claude Code, Cursor, GitHub Copilot, and Antigravity CLI, read on September 25, 2026, in USD. A pricing-only breakdown of Codex's closest rival lives at Claude Code pricing.

## What is OpenAI Codex?

OpenAI Codex is OpenAI's coding agent, reachable from the ChatGPT desktop app, the web, a command-line tool, an IDE extension, and iOS. It reads a repository, edits files, runs commands in a sandbox, and can open a pull request, so the work happens against the codebase you already have open rather than in a separate chat window. OpenAI was founded in December 2015 and is headquartered in San Francisco.

Every ChatGPT tier from Free through Enterprise gets some access: the two cheapest tiers run on GPT-6 Luna at standard speed, while the two priciest personal tiers add GPT-6 Sol for heavier reasoning tasks. A worktree system lets several agent sessions run in parallel on the same repo, and cloud environments keep a task going after the laptop closes.

Skills let a team encode a repeatable workflow, such as a pull-request review checklist, so Codex applies it the same way every time instead of being re-explained each session.

Background automation covers issue triage, CI monitoring, and alert response on a schedule, which is the part that turns Codex from a chat helper into something closer to a junior teammate with recurring duties.

An API key runs the CLI outside any ChatGPT plan, billed at standard token rates, but that path drops the cloud features: no automatic GitHub review, no Slack integration. Business and Enterprise workspaces add SAML SSO, SCIM, and audit logs, none of which sit on a personal seat below them.

## How OpenAI Codex works

Install the CLI or the IDE extension, sign in with your ChatGPT account, and describe the change in plain language. Codex searches the repository, edits the files it needs, and can run your existing test suite, so the check that gates the change is the one your team already trusts.

Tag @Codex on a GitHub pull request and it reviews the diff without you opening a terminal at all.

Usage resets locally over a five-hour window, and OpenAI's own dashboard shows estimated local messages per period rather than a hard count: entry-level paid users see roughly 5 to 45 GPT-6 Astra turns or 350 to 3,000 GPT-6 Luna turns per window, and the two Pro tiers scale that range up by five and twenty times respectively.

Model choice, context size, and tool use all change how fast a session burns the allowance, so prompt length alone will not predict your week.

When you hit the ceiling mid-task, Codex finishes the turn you are already on under a fair-use allowance, then paid personal users can buy extra credits rather than being forced to upgrade a plan outright. Business and Enterprise workspaces on flexible pricing draw down shared workspace credits instead of a fixed per-seat cap.

For the review-focused half of that workflow, the AI code review list is the narrower buy, and how to use Claude Code is the closest terminal-agent setup guide for comparison.

## OpenAI Codex key features

One agent across five surfacesEssential
The web app, the CLI, an IDE extension, the ChatGPT desktop app, and iOS all sign in with the same ChatGPT account and draw from the same usage pool, so a fix started on the train can finish at a desk without switching tools.

Cloud tasks and worktreesEssential
Built-in worktrees let several Codex sessions work the same repository in parallel, and cloud environments keep a scheduled task running after the local machine goes to sleep, which is what turns a single agent into several simultaneous ones.

GitHub review and Slack integration
Tagging @Codex on a pull request triggers an automatic review without opening a terminal, and a Slack mention can hand a bug report straight to the agent. Both integrations are cloud features and drop off entirely on an API-key-only setup.

Skills for repeatable workflows
A skill packages a workflow, such as a security review checklist or a release note format, so a team codes the process once and Codex repeats it the same way on every task instead of drifting between sessions.

Background automation
Scheduled runs cover issue triage, CI monitoring, and alert response without a person kicking off the session, so recurring maintenance work moves off a developer's plate instead of piling into a backlog.

Credits instead of a forced upgrade
Paid users who exhaust their usage window can buy additional ChatGPT credits and keep working at the current plan rather than being pushed to the next tier, useful for a single heavy week rather than a permanently heavier habit.

## OpenAI Codex pricing

Codex, Claude Code, Cursor, GitHub Copilot, and Antigravity CLI each bundle their agent into a broader subscription rather than selling it alone, so the comparison below is which host plan gives you enough coding usage for the least money.

Codex Free and Go run on the lightest model tier; Plus is the first level with the CLI, IDE extension, cloud review, and Slack together, and Pro starts at five times that Plus allowance, or four times that again on the top tier.

OpenAI paused new sign-ups and upgrades to that top Pro tier on September 10, 2026, so anyone who needs that ceiling today has to already hold it or wait for the pause to lift, per OpenAI's help center.

Claude Code carries no price of its own either. Annual Pro is $17/mo with Claude Code included, undercutting Codex Plus for a broadly similar single seat, though Claude's monthly Pro rate matches Codex Plus exactly.

Claude's two Max tiers mirror Codex's two Pro tiers almost dollar for dollar, but Anthropic sells no annual discount on either Max tier, where Codex never offered an annual option to begin with. Claude's Team Standard seat matches Codex Business on the annual per-seat rate, and Team Premium runs five times that, both capped per person rather than pooled.

Cursor is the one tool here priced as its own product rather than folded into a chat plan. Hobby is free with limited agent requests, Pro matches the entry rate that Codex and Claude share, and Anysphere recommends Pro+ for daily agent use, a middle tier neither Codex nor Claude Code offers at that exact price point.

Ultra tops out at the same ceiling as the higher Pro-style tiers above, and Teams runs a flat $40/user/mo regardless of Standard or Premium seat type.

GitHub Copilot is the cheapest paid entry of the five: Pro includes 1,500 AI credits worth $15 at the standard per-credit rate, Pro+ is $39/mo with 7,000 credits, and the top individual tier matches the mid-range Pro figure shared by Codex, Claude, and Cursor, but with 20,000 credits included.

Business seats run about half the Pro+ rate per user, and Enterprise costs roughly double Business, both cheaper per seat than Claude's Team Premium or Cursor's Teams plan. Inline completions stay outside the credit meter on every paid Copilot tier, which none of the four chat-first agents here separate out from their general usage pool.

Google moved free and consumer users from Gemini CLI to Antigravity CLI on June 18, 2026, and Antigravity is the only tool here with a real free tier of its own: the $0 individual plan includes Gemini Flash models, Claude Sonnet and Opus, unlimited Tab completions and basic weekly rate limits.

Raising those limits runs through a Google AI subscription rather than a developer-specific plan: Google AI Pro is $19.99/mo, and Google AI Ultra is $99.99/mo, with a $199.99/mo tier for the highest usage.

Organizations reach Antigravity through Google Cloud with consumption-based pricing, so there is no simple seat price to quote against the other four tools.

Plan | Price | Best for |

Codex on ChatGPT Free | No-cost | GPT-6 Luna at standard speed, light usage only |

Codex on ChatGPT Go | $8/mo | Lightweight coding tasks, same standard-speed model |

Codex on ChatGPT Plus | $20/mo | CLI, IDE extension, web, iOS, cloud review, Slack |

Codex on ChatGPT Pro 5x | From $100/mo | 5x Plus usage across the same surfaces |

Codex on ChatGPT Pro 20x | $200/mo | 20x Plus usage; new sign-ups paused since Sept 10, 2026 |

Codex on ChatGPT Business | $25/mo monthly (per user) | Minimum 2 seats, drops to Plus's rate per seat if billed annually |

Codex Enterprise / Edu | Custom quote | SCIM, audit logs, data residency controls |

Claude Free | No-cost | Claude Code is not included on this plan |

Claude Pro, annual | $17/mo | Billed up front, Claude Code included |

Claude Pro, monthly | Matches Codex Plus | Same access, billed month to month |

Claude Max 5x | Matches Codex Pro 5x | 5x Pro session allowance, monthly only |

Claude Max 20x | Matches Codex Pro 20x | 20x Pro session allowance, monthly only |

Claude Team Standard | Matches Codex Business, annual | Minimum two seats, usage capped per person |

Claude Team Premium | Matches Codex Pro 5x, per seat annual | 6.25x Pro allowance per seat, $125 if billed monthly |

Claude Enterprise | Matches Codex Business's per-seat rate, plus API usage | Annual billing, SCIM, audit logs |

Cursor Hobby | Free | Limited agent requests, no credit card required |

Cursor Pro | Matches Codex Plus and Claude Pro monthly | Extended agent limits, frontier model access |

Cursor Pro+ | $60/mo | Anysphere's recommended tier for daily agent use |

Cursor Ultra | Matches Codex Pro 20x | Top individual tier for agent power users |

Cursor Teams | $40/user/mo | SAML/OIDC SSO, Standard or Premium seats |

GitHub Copilot Free | No-cost | Limited chat and agent usage, capped completions |

GitHub Copilot Pro | $10/mo | 1,500 AI credits worth $15, unlimited completions |

GitHub Copilot Pro+ | $39/mo | 7,000 AI credits worth $70 |

GitHub Copilot Max | Matches Codex Pro 5x | 20,000 AI credits worth $200 |

GitHub Copilot Business | $19/user/mo | Org-managed seats, pooled org controls |

GitHub Copilot Enterprise | Matches Copilot Pro+ | Priced per user, pooled credits, repository-aware chat |

Antigravity, individual | No-cost | Gemini Flash and Claude models, unlimited Tab, basic weekly rate limits |

Antigravity, organization | Consumption-based | Through Google Cloud and select Gemini Enterprise subscriptions |

Google AI Pro | $19.99/mo | More generous rate limits and a flexible AI credit pool |

Google AI Ultra | From $99.99/mo | Ultra 5x; the Ultra 20x tier is $199.99/mo |

## OpenAI Codex pros and cons

### What we like

- Plus covers the CLI, IDE extension, web app, and iOS, plus cloud code review and Slack, in one seat, no separate coding SKU.
- Pro offers 5x or 20x Plus usage, and eligible users can buy extra credits instead of upgrading plans outright.
- An API key still runs the CLI for CI work, and Go is the cheapest paid entry among all five tools compared in this review.

### What could be better

- OpenAI paused new sign-ups and upgrades to its top $200 Pro tier on September 10, 2026, with no reopening date announced.
- Plus and Pro carry no annual discount, unlike Claude Pro's cheaper annual rate for a comparable single seat.
- GPT-6 Astra costs 250 credits per million input tokens against Luna's 2.5, so reasoning-heavy sessions burn Pro credits fastest.

## Who OpenAI Codex is for

Codex fits a developer who already pays for ChatGPT and wants that subscription to cover an agent instead of adding a second bill. Plus is the practical entry for a few focused coding sessions a week, and the next Pro tier up is the move once a single evening of migration work burns through that allowance before you are done.

Teams already inside a ChatGPT Business workspace get SSO and admin controls without standing up separate billing for the coding agent.

It is a weaker default when the daily habit is unlimited Tab-style autocomplete, since GitHub Copilot keeps completions off its credit meter on every paid plan, and Cursor bundles unlimited completions into an editor built around that workflow.

GitHub Copilot is the cheaper seat if IDE completions matter more than a terminal agent, and Cursor is the pick when the editor itself should be AI-native rather than a plugin bolted onto VS Code.

Skip a personal seat on Codex's highest tier if that is the one you actually need right now. It has taken no new sign-ups or upgrades since September 10, 2026, so anyone starting from scratch is limited to the lower Pro tier until OpenAI lifts that pause.

Anyone who wants the largest free allowance before paying anything should start on Antigravity's free tier instead, then compare the paid tiers here once its weekly limits start to bind.

Devshot covers pauses like this one as they happen. Subscribe free for the next plan or usage change across these five tools.

## Best OpenAI Codex alternatives

If OpenAI Codex is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

OpenAI Codex | ChatGPT subscribers who want one agent across CLI, IDE, web, and iOS. | Free on ChatGPT Free | Visit → |

Claude Code | Developers who want an agent in the terminal, VS Code, or JetBrains without buying a second SKU. | Included from $17/mo on annual Pro, matching Codex Plus if billed mont | Visit → |

Cursor | Developers who want unlimited Tab completions inside a dedicated AI-native editor. | Hobby is free | Visit → |

GitHub Copilot | Teams that want IDE completions and a GitHub-native seat without a second billing relationship. | Free plan available | Visit → |

Antigravity CLI (formerly Gemini CLI) | Individual developers who want a free tier before paying for any coding agent. | Free individual plan with weekly limits | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Claude Code

Anthropic's repo agent, bundled with every paid Claude plan and left off the free one.

Visit →

Cursor

A standalone AI code editor with its own model pool, priced apart from any chat subscription.

Visit →

GitHub Copilot

The cheapest paid entry among the five, with credits topped up separately from completions.

Visit →

Antigravity CLI (formerly Gemini CLI)

Google's terminal agent, which replaced Gemini CLI for free and Google AI subscribers, with higher limits sold through Google AI plans.

Visit →

## The bottom line

Pay for Plus if the ChatGPT seat you already have should also run the agent, and move to the 5x Pro tier only once a real week of migrations and refactors runs past what Plus covers. Do not plan around Codex's top tier for a new account: it has taken no sign-ups or upgrades since September 10, 2026, and OpenAI has not said when that changes.

Choose [Claude Code](https://toolradar.com/tools/claude-code) when the same seat should also be your daily chat subscription and annual Pro beats Codex Plus on price. Choose Cursor when unlimited Tab completions inside a dedicated editor matter more than a terminal-first agent.

Choose GitHub Copilot when a $10/mo GitHub-native seat covers the job. Choose Antigravity CLI when a genuinely free tier should come before any card gets charged.

The wider field sits in the AI coding assistants list, and the editor-shaped version of this same choice is the vibe coding tools ranking. Subscribe free for the next pricing change across all five.

Cite this: Devshot, "OpenAI Codex Review 2026", September 2026.

## Frequently asked questions

Is OpenAI Codex worth it in 2026?

Yes, if you already pay for ChatGPT Plus or a higher tier and want that same seat to run an agent in the CLI, IDE, web app, and iOS instead of buying a second product. Plus covers a few focused coding sessions a week, and Pro raises that to 5x or 20x the usage; see the exact prices below. It is a weaker buy if you mainly want unlimited autocomplete, which GitHub Copilot and Cursor both offer on their paid plans without touching a shared chat allowance.

How much does OpenAI Codex cost?

There is no separate Codex price: Free costs nothing, Go is priced for light use, Plus is $20/mo, and Pro starts from $100/mo for 5x Plus usage or twice that for 20x. Business runs $25/mo billed monthly per seat, dropping to match Plus's rate annually, with a minimum of two seats, and Enterprise or Edu pricing is a custom quote. None of the personal tiers offer an annual discount. Prices were verified on OpenAI's ChatGPT and Codex pricing page in September 2026.

Is there a free version of Codex?

Yes. The Free ChatGPT plan includes limited Codex access on the GPT-6 Luna model at standard speed, enough to explore the tool on quick coding tasks. The Go tier raises that for lightweight work without adding the cloud features that arrive on Plus, such as automatic GitHub review and Slack integration. An API key also runs the Codex CLI outside any ChatGPT plan, billed at standard token rates with no monthly minimum.

Why did OpenAI pause the top Codex Pro tier?

OpenAI's help center says it temporarily paused new sign-ups and upgrades to the $200 Pro 20x plan on September 10, 2026, citing demand after the GPT-6 Astra release. Existing subscriptions on that tier keep renewing as normal, and eligible customers who lose access during the pause can use a one-time return within 30 days. New customers and anyone upgrading from a lower tier are limited to the lower Pro 5x plan until OpenAI lifts the pause, with no date announced as of this review.

How does Codex compare with Claude Code?

Both agents run from a CLI, an IDE, and the browser without requiring a dedicated editor, so the choice comes down to which chat subscription you would rather own. Annual Claude Pro undercuts Codex Plus for a broadly similar seat, while Codex Plus adds automatic GitHub pull-request review and a Slack integration Claude Code does not match at that tier. Codex Pro and Claude Max mirror each other closely at their matching 5x and 20x-equivalent tiers, but Codex's top tier is currently closed to new sign-ups while Claude's is not.

How does Codex compare with Cursor?

Codex rides inside a ChatGPT plan and works in whichever editor you already use; Cursor is its own editor with a model pool built in. Cursor Pro matches Codex Plus on price, but Cursor also ships unlimited Tab completions on every paid individual tier, something Codex does not offer as a separate feature. Anysphere recommends Cursor Pro+ at $60/mo for daily agent use, a middle tier Codex has no direct equivalent for between its Plus and Pro plans.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [OpenAI Codex pricing](https://openai.com/pricing)
- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026

Related guides

Cursor ReviewGithub Copilot ReviewAi Coding AssistantsVibe Coding ToolsAi For Code ReviewAi For DebuggingCodex vs Claude Code 2026: Plans, Limits and Real CostsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Railway Review

URL: https://devshot.co/reviews/railway-review
Type: review
Published: 2026-09-24
Updated: 2026-09-25
Summary: Railway review 2026: how the Hobby credit actually works, why an idle service still bills, and how Render, Fly.io, Northflank, App Platform, and Heroku compare.

Review

## Railway Review

Worth it when a stopped service should leave the bill, and a weak default when the process stays running with no cap.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Railway is worth it in 2026 when you want the invoice to follow CPU and memory the service actually holds, and a stopped process to cost nothing. Hobby is $5 a month and that fee is the usage credit, while Pro is $20 per workspace with seats included, verified on Railway's pricing page on September 24, 2026.

A service left running still bills with zero requests, because sleep is a setting you have to turn on. Compare [Render](https://toolradar.com/tools/render) when you want a sized instance you can read before you deploy, and [Fly.io](https://toolradar.com/tools/flyio) when the region is why you are choosing a host. Choose Heroku when you want a named dyno instead of a usage meter.

Key factsBest for: Backends that should pay for CPU and RAM in use, and drop…

- Updated: September 25, 2026
- Best for: Backends that should pay for CPU and RAM in use, and drop compute when stopped.
- Price as of September 25, 2026: Hobby $5/mo includes the usage credit; Pro $20 per workspace, seats included.

- Containers billed per second, where the subscription is the usage credit for what stays running.
- Founded: 2020
- Meter: Per second
- Free domains: None on Free
- Alternatives covered: Render, Fly.io, Northflank, DigitalOcean App Platform, Heroku

Pros

- The paid subscription is the usage credit, so a month under that credit adds no second line.
- A stopped service drops CPU and memory, and Serverless can sleep a quiet app in about 5 to 10 minutes.
- Pro is one workspace fee with a zero seat line, so a teammate does not add a seat charge.

Cons

- Sleep is off by default, so an idle service that is still running keeps billing RAM and CPU.
- Free stops at one project and the memory cap on the Free row, so a production app has to leave it.

Founded2020

MeterPer second

Free domainsNone on Free

Best forUsage-billed backends

Railway fits a backend that sits quiet most of the day, and it is a poor fit when staging is left running. The subscription is a floor on the invoice rather than a ceiling, so a quiet month still costs the plan.

Toolradar data: the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) we publish scored 297 products in September 2026. The median cheapest paid plan among the 10 at the top is $5.50/mo, and [Railway](https://toolradar.com/tools/railway) is outside that ten. The decision is whether a stopped service leaves the invoice.

How we compared: Railway, Render, Fly.io, Northflank, DigitalOcean App Platform, and Heroku, in USD, on September 24, 2026. We did not deploy a test app. A frontend with no long-running process belongs on the Vercel review.

## What is Railway?

Railway is a usage-billed platform for services you deploy from GitHub, a Docker image, the CLI, or a template. Nixpacks can build the image, or you bring a Dockerfile, and build CPU, memory, and image storage are not charged, so a rebuild does not add a line.

The canvas is services rather than dyno types, so Postgres, Redis, MongoDB, and ClickHouse templates are a service plus a volume and the disk line follows that volume. A public URL and a private URL both exist, and only the private one is free to talk to, so the public database hostname is the one that adds a network charge.

Enterprise is a contract, while Hobby and Pro are the self-serve bills you can open without a sales conversation.

## How Railway works

You connect a repo and Railway runs a service, and CPU, memory, and disk bill per second while it runs. A stopped service costs nothing, while an idle process that stays up still bills RAM and CPU.

Serverless is the off switch, and it stays off until you enable it, so a quiet app does not sleep on its own. The docs call a service inactive about 5 minutes after the last outbound packet, and sampling means sleep usually lands between 5 and 10 minutes. The toggle hits every replica and starts on the next deploy.

A database pool, Next.js telemetry, NTP, and private-network calls all count as outbound, so any of them keeps the meter running. The metrics graph hides private traffic, so a service can look idle and still stay awake, and the first request after sleep can return a 502. A webhook that does not retry will drop that event.

Five open pull requests can burn the credit five times over, because each preview is another running copy. The private hostname is not egress, and the public database URL is, so database calls belong on the private hostname.

## Railway key features

Hobby and Pro creditsEssential
Each paid subscription is a usage credit of the same size, and it resets every cycle without rolling over, so a quiet month cannot bank the leftover. Stay under it and the invoice is the subscription, and cross it and the invoice equals total usage, so the fee is a minimum rather than a cap. Railway's bill guide shows a $3 month charged as the subscription only, and an $8 month charged as that usage total.

Serverless sleepEssential
Sleep is a per-service toggle, not the default, so a quiet HTTP app drops compute only after you enable it and a few minutes pass with no outbound packets. A pool that stays connected keeps the meter running. Wake-up can 502 the first caller, so a webhook that does not retry will drop that event.

Volumes, templates, and egressEssential
A database template is still a service plus disk, so you pay compute plus the volume rather than a named database plan. Hobby caps the volume, and Pro can self-serve up to the larger ceiling in the table. Service egress is the per-GB rate in the table, object storage is cheaper per GB, and its egress is waived. A public database URL is how a small app grows a network line.

Replicas, regions, and previews
Hobby allows 6 replicas, with a per-service ceiling of 48 vCPU and 48 GB, and Pro allows 42 replicas, up to 1,000 vCPU and 1 TB, with ceilings that include replica multiplication. Each copy bills, and so does every open preview, so review apps can cross the credit while production is quiet.

Usage limits and the agent
Compute and agent spend have separate caps, so a runaway agent and a runaway service are different bills. A compute hard limit shuts workloads down, while an agent hard limit disables the agent and leaves the app up. The agent cap defaults to the plan fee and cannot be removed, the compute cap can, and the smallest one equals one GB of memory for a full month.

Sandboxes
Sandboxes are ephemeral VMs for agents and untrusted code, billed on the VM rows in the table rather than the container rates. Memory stays billable while the VM waits, destroyed sandboxes stop, and Hobby allows 50 running sandboxes per environment.

## Railway pricing

Read the credit before the resource rates, because the Hobby fee and the Pro fee are minimums that wipe the first slice of CPU, memory, disk, and egress. Unused credit dies at the end of the cycle, and a quiet month still costs the subscription, which the plans FAQ states outright, so low usage does not zero the bill.

The pricing page meters resources per second and prints a monthly constant beside each rate, for comparison only. One full gigabyte of RAM for the whole month is the memory row, and that single line spends half the Pro credit before any vCPU time.

A second replica doubles that draw, CPU is its own row, and a full vCPU for the month equals one Pro fee by itself.

Free is a $1 monthly credit, one project, three services, one member, and no custom domain, so it cannot carry a production hostname. The per-service cap is 1 vCPU and the memory on the Free row.

Trial is a one-time credit equal to one Hobby month, with no card, then it expires, and upgrading carries any leftover trial balance onto the paid plan.

Hobby is the solo plan: 50 projects, 50 services, 3 members per project, 2 custom domains, and the volume ceiling in the table. Pro is the team plan: one fee per workspace, a zero seat line, unlimited members, and 20 custom domains.

Logs run 3 days on Free, 7 on Hobby, and 30 on Pro, so a longer history means a higher plan. Rollback of a removed deploy lasts 24 hours on Free, 72 on Hobby, and 120 on Pro.

After you cancel, volume data is deleted in 30 days on Free, 60 on Hobby, and 90 on Pro, so a cancelled volume is not a backup. Invoices under $0.50 are carried forward, and a partial charge can hit mid-cycle. A post-paid card is required, a Hobby fee waiver exists and is automatic, and Railway will not review requests.

SSO and RBAC are tied to committed spend or Enterprise, not to the Pro fee, so the self-serve plan does not include SSO. Business Class Support has no published price, and Enterprise publishes no list price, so neither can be budgeted from the plan fee.

The other hosts in the table sell a size or a region, not a credit. Render's paid instances do not spin down, Fly.io has no workspace fee, and Northflank's sandbox stays awake. App Platform's smallest shared container matches the Hobby sticker, and Heroku's metrics dyno does not add RAM.

Prices below were read on each vendor's pricing page on September 24, 2026.

Plan | Price | Best for |

Railway Free | $0, plus $1/mo credit | 1 project, 3 services, 0 custom domains, 0.5 GB cap per service. |

Railway Trial | $0 for 30 days | One-time credit equal to one Hobby month. No card. Balance can carry over. |

Railway Hobby | $5/mo | Matching credit. 2 custom domains, 5 GB volume cap, 3 members. |

Railway Pro | $20/workspace/mo | Credit matches the fee. Unlimited seats. 20 custom domains. 30-day logs. |

Railway CPU | $0.00000772/vCPU/s | A full month of one vCPU equals the Pro fee. Stopped services drop it. |

Railway memory | $10 per GB | Per second while running. One always-on GB spends half the Pro credit. |

Railway volume | $0.15 per GB | Hobby ceiling is 5 GB. Pro can self-serve to 1 TB. |

Railway service egress | $0.05 per GB | Public database URLs count. Private networking inside the project does not. |

Railway object storage | $0.015 per GB-month | Egress from object storage is waived. Service egress is not. |

Railway sandbox VM | $50/vCPU and $50/GB | While running. Waiting still holds memory. Draws on the same credit. |

Render Starter | $7/mo | 512 MB, 0.5 CPU. Paid instances do not spin down. |

Render Pro workspace | $25/mo | Unlimited seats, 25 GB bandwidth, then the per-GB rate below. |

Render bandwidth overage | $0.15 per GB | Hobby includes 5 GB. Free web services count against that pool. |

Render Postgres 0.1c-256mb | $6/mo | 256 MB, 100 connections, 1 GB storage. Free Postgres expires in 30 days. |

Fly.io shared-cpu-1x | $1.94 to $3.14/mo | 256 MB. The spread is the region tables on the pricing page. |

Fly.io support Standard | $29/mo | Separate from the Machine. A card is required on the organization. |

Northflank nf-compute-10 | $2.70/mo | 0.1 shared vCPU, 256 MB, prorated to the second. Egress is $0.06/GB. |

App Platform shared 512 MiB | Matches the Hobby fee | 1 vCPU, 50 GiB transfer. Extra outbound is $0.02/GiB. |

App Platform dev database | $7 for 512 MiB | No default backups. Destroyed with the app. Postgres only. |

Heroku Basic | $7/dyno/mo | Always on, 0.5 GB. Eco is personal-only and sleeps after 30 minutes. |

Heroku Standard-1X | $25/dyno/mo | Metrics and preboot. RAM stays 0.5 GB. |

## Railway pros and cons

### What we like

- The paid subscription is the usage credit, so a month under that credit adds no second line.
- A stopped service drops CPU and memory, and Serverless can sleep a quiet app in about 5 to 10 minutes.
- Pro is one workspace fee with a zero seat line, so a teammate does not add a seat charge.

### What could be better

- Sleep is off by default, so an idle service that is still running keeps billing RAM and CPU.
- Free stops at one project and the memory cap on the Free row, so a production app has to leave it.

## Who Railway is for

Railway fits a developer who will turn Serverless on, set a compute cap, and keep databases on the private hostname. Hobby covers a side project under the credit, and Pro is the switch when the volume ceiling or a teammate needs the workspace.

Skip it when the app must stay warm, because a running service bills with no users. Skip it when Postgres should be a named plan: see Postgres hosting, the Neon review, and Supabase if auth ships with the database.

A page-only Next.js app is a closer call on the Vercel review and the serverless platforms roundup. Next.js telemetry is outbound traffic, and it blocks Railway sleep.

For a sized instance, start with the Heroku alternatives and the Heroku review. For a VM you administer, use VPS hosting for developers and the DigitalOcean review.

A box you patch yourself is the self-hosting tools list, a cluster you already operate is the Kubernetes tools decision, and subscribe free for the next credit note.

## Best Railway alternatives

If Railway is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Railway | Backends that should pay for CPU and RAM in use, and drop compute when stopped. | Hobby $5/mo includes the usage credit | Visit → |

Render | Teams that want a sized web instance and a Postgres plan with a sticker. | Starter is 512 MB and 0.5 CPU on the table | Visit → |

Fly.io | Long-running Machines in a chosen region, billed while the Machine stays started. | shared-cpu-1x at 256 MB lists from $1.94 to $3.14/mo by region | Visit → |

Northflank | Containers and previews for teams that need a sandbox that stays awake. | From $2.70/mo (nf-compute-10) | Visit → |

DigitalOcean App Platform | Git deploys when the rest of the stack is already on DigitalOcean. | Smallest shared container matches Hobby | Visit → |

Heroku | Teams that want a named dyno and a Postgres add-on on one invoice. | Basic is the always-on dyno in the table | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 6 do not publish a comparable monthly price and are left out rather than estimated.

Render

Instance sizes plus a workspace fee, and a free web service that sleeps on a timer.

Visit →

Fly.io

Firecracker Machines priced by region, and CPU accrues only while the Machine is started.

Visit →

Northflank

Per-second containers on Northflank's cloud or yours, with a sandbox that stays up.

Visit →

DigitalOcean App Platform

Fixed shared containers, a static-only free tier, and a dev database at a flat rate.

Visit →

Heroku

Salesforce's PaaS, where a dyno stays on the bill while it is scaled up, traffic or not.

Visit →

## The bottom line

Pay for Railway when the workload can sleep, someone will set a hard limit, and services talk on the private hostname. Start on Trial with no card, move to Hobby before a production hostname, and move to Pro when the volume or the replica ceiling is the blocker.

Leave it when the process must stay warm, because a running service spends the memory row with no users and a full vCPU spends a Pro fee of compute on its own. That ceiling is a sized instance on [Render](https://toolradar.com/tools/render), or a dyno on [Heroku](https://toolradar.com/tools/heroku).

Choose [Fly.io](https://toolradar.com/tools/flyio) when the region table matters and a workspace fee does not, [Northflank](https://toolradar.com/tools/northflank) when the sandbox should stay awake, and [App Platform](https://toolradar.com/tools/digitalocean-app-platform) when the account already pays DigitalOcean.

Subscribe free if you want the next credit change in the Devshot brief.

Cite this: Devshot, "Railway Review 2026", September 2026.

## Frequently asked questions

Is Railway worth it in 2026?

Yes, when the app can stop and the CPU and memory it held should leave the bill with it. The paid subscription is the usage credit, so a small month adds no surcharge. It is a weak buy for a process that must stay warm, because a running service bills with no traffic until you enable Serverless. It is also a weak buy for a database that needs a named plan and more than the 5 GB Hobby volume cap.

How much does Railway cost?

Hobby is $5 a month and Pro is $20 per workspace, each with a matching usage credit that resets every cycle. Under the credit you pay the subscription, and over it you pay the usage total, so a busy month replaces the fee instead of stacking on it. Memory is $10 per GB for a full month, service egress is $0.05 per GB, and volumes are $0.15 per GB. A full vCPU for the month equals the Pro fee, and Enterprise publishes no list price. Rates were verified on Railway's pricing page on September 24, 2026.

Does Railway have a free plan?

Yes, Free is a $1 monthly credit, one project, three services, no custom domain, and a 0.5 GB cap, so it cannot carry a production hostname. Trial is a one-time credit equal to one Hobby month, for 30 days, with no card. Render's $0 web instance sleeps after 15 minutes, which is the closer free web tier if a custom domain matters on day one.

How does Railway compare with Render?

Railway meters what the service holds and lets a stopped service leave the compute bill. Render sells a size: Starter is 512 MB and 0.5 CPU at $7/mo, and that paid instance does not spin down, so the size bills with or without traffic. Render's free web service does spin down, after 15 minutes, inside 750 hours. Render also sells Postgres as a plan, 0.1c-256mb at $6/mo, while Railway runs the database as a service plus a volume. Pick Render when you want a ceiling you can read before you deploy, and pick Railway when stopped time should cost nothing.

Does Railway charge when the app is idle?

Yes, if the service is still running, because you pay for CPU and RAM held, not for requests. Serverless stops an inactive service after about 5 to 10 minutes without outbound packets, and it is off until you toggle it and redeploy. A connection pool or Next.js telemetry blocks that sleep, and a hard usage limit can shut the workspace down, with a floor of one month on the memory row.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Railway pricing](https://railway.app/pricing)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026

Related guides

Heroku ReviewHeroku AlternativesVercel ReviewDigitalocean ReviewNeon ReviewSupabase ReviewPostgres HostingServerless PlatformsVps Hosting For DevelopersSelf Hosting ToolsKubernetes ToolsFly Io ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Render Review

URL: https://devshot.co/reviews/render-review
Type: review
Published: 2026-09-24
Updated: 2026-09-25
Summary: Render review 2026: the workspace fee is flat, each instance is a fixed size, and bandwidth replaced the old per-seat bill.

Review

## Render Review

Worth it when the web process, the worker, and Postgres each have a size you can budget. A poor fit when that worker should cost nothing overnight.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Render is worth it in 2026 when the app is a web process, a worker, and Postgres, and you want each size fixed before the month starts. A Starter instance is $7/mo for 512 MB and 0.5 CPU. The Pro workspace is $25/mo with unlimited seats, checked on Render's pricing page in September 2026.

Hobby includes 5 GB of outbound bandwidth, then a per-gigabyte rate, so every service in the workspace shares that pool. A paid instance keeps billing when nobody is calling it. Compare Railway when a stopped service should leave the compute line, and [Fly.io](https://toolradar.com/tools/flyio) when the region table sets the price.

Key factsBest for: Teams that will name a web process, a worker, and Postgres…

- Updated: September 25, 2026
- Best for: Teams that will name a web process, a worker, and Postgres before the month starts.
- Price as of September 25, 2026: Starter is the 512 MB row in the table. Pro is a flat workspace fee with unlimited seats.

- Fixed sizes for each service, plus a workspace plan, and a free web process that sleeps on a timer.
- Founded: 2019
- Free web idle: 15 min
- Pro seats: Unlimited
- Free Postgres: 30 days

Pros

- The smallest paid web plan is half a CPU and 512 MB, so the ceiling is visible before the first deploy.
- Pro dropped the per-member charge on April 23, 2026, and unlimited seats sit on one workspace fee.
- Suspend a paid service and compute billing stops at that second, which an idle paid instance will not do on its own.

Cons

- Paid web services and workers do not spin down, so a worker left running bills its plan through every quiet hour.
- Free Postgres expires 30 days after creation, so the free database is a clock and not a home.

Founded2019

Free web idle15 min

Pro seatsUnlimited

Free Postgres30 days

Render is the host I would give a Rails or Django app that still runs a web process, a queue worker, and Postgres. Those three pieces are a bill you can name in advance. I would not give it a frontend that never stays warm, or a worker that should vanish from the bill at night.

The workspace plan sets who can log in, how much outbound traffic is included, and whether a pull request can boot the whole stack. That line moves when a teammate needs the dashboard, not when the app needs more RAM.

The compute plan sets RAM and CPU for one service, and it keeps charging while that service is running.

Toolradar data: the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) we publish scored 297 tools in September 2026. [Render](https://toolradar.com/tools/render) is not among the 10 named at the top.

Heroku and DigitalOcean App Platform sit in the group ranked 11 through 30. The median cheapest paid plan among those 10 is $5.50/mo.

How we compared: Render, Railway, Fly.io, Northflank, DigitalOcean App Platform, and Heroku, priced in USD on September 24, 2026. We did not deploy a test app. A Next.js site with no long-running process belongs on the Vercel review and the serverless platforms roundup.

## What is Render?

Render is a Git-connected platform where you pick a size for each service, so the month has a ceiling you can read first. A web service gets a public URL, HTTP/2, and managed TLS, and that URL is what visitors open.

A private service stays off that URL, and a background worker runs with no HTTP port, so the queue is not a public site.

Cron, static sites, Render Postgres, and Key Value (compatible with Redis clients) are their own service types, so each one is its own line. Native builds cover Node.js, Bun, Python, Ruby, Go, Rust, and Elixir, so those stacks ship without a Dockerfile.

Java, .NET, and anything else ship as a Docker image, and that image is still a sized instance.

A blueprint file, render.yaml, describes the web process, the worker, and the database in one pull request. That file is how a team stops staging from drifting away from production. Hobby, Pro, and Scale are self-serve, so you can open the account without a sales call.

Enterprise is a contract and publishes no list price, so you cannot budget it from the page. Custom compute, up to 64 CPU and 512 GB of RAM, is quote-only, so a large box is a sales conversation.

## How Render works

You connect a repo, choose a compute plan, and Render runs the build, so the size is the decision. Build commands and pre-deploy commands spend pipeline minutes, and Hobby includes 500 of those minutes. That allowance fits a modest solo app and gets tight if you rebuild all day.

Past it, new builds can be blocked if you have no card, or billed if you do. A free account then stops rather than running up a charge.

A spend cap on pipeline minutes stops new builds while the last good artifact keeps serving. The live site stays up, and the next release waits. A Free web service spins down after 15 minutes without inbound HTTP or WebSocket traffic, then takes about a minute to wake. Browsers see a loading page during that wake, so the first visitor on a cold link waits.

The workspace gets 750 free instance hours each calendar month. Spun-down time does not spend those hours, and leftover hours do not roll over, so a quiet week cannot be saved for later. A second free web service running beside the first will exhaust the pool before the month ends.

Paid web services and workers do not spin down, so sleep is a free-tier behavior only.

The disk on a service is ephemeral, so uploads and a local SQLite file disappear on redeploy, restart, or spin-down, and rows belong in Postgres. A persistent disk keeps those files, and attaching one disables zero-downtime deploys for that service.

Free web services may send private-network requests to a database in the same region, and they cannot receive private traffic.

Hobby can preview a single service, which covers one pull request and not the API beside its database. Full-stack preview environments, the pull request that boots the API and the database together, start on Pro, and so does horizontal autoscaling.

Vertical scaling, moving one service to a larger plan, is available on Hobby, so you can buy RAM without a teammate. Suspend a paid service and the compute clock stops at that second.

## Render key features

Workspace plan versus compute planEssential
The workspace plan sets seats, the outbound pool, and whether a pull request can preview every service, and that fee stays flat when you add people. RAM is not included, so each web service, worker, and private service has its own compute plan. Every running copy bills, including copies a preview creates, so a quiet pull request is not free.

Free sleep, paid uptimeEssential
Free web services stop after 15 idle minutes and take about a minute to wake, which a webhook that must answer now cannot accept. Each workspace gets 750 free instance hours a month, and spun-down time does not spend them, while paid services never get that sleep.

Postgres and Key ValueEssential
Free Postgres is one database per workspace, capped at 1 GB, with no backups, so it fits one prototype and not a second app. It expires 30 days after creation, then you have 14 days to move it to a paid plan before the data is deleted. Free Key Value is 25 MB, keeps nothing on disk, and drops its data on restart, so treat it as a cache.

Bandwidth after the seat fee
On April 23, 2026 Render replaced per-member Professional pricing with a flat Pro fee, so a new teammate does not multiply the platform line. Hobby and Pro each include an outbound pool, and each extra gigabyte is its own charge, which is the meter that replaced the seat. Inbound is free, including traffic Render attributes to a DDoS attack, so a flood aimed at you is not an outbound bill.

Previews, autoscaling, blueprints
A blueprint can declare the web service, the worker, and the database together, so one pull request describes the stack. Hobby previews one service, while Pro previews the whole environment and can autoscale on CPU or memory when the API depends on the database.

Disks, cron, and workflows
Persistent disk is billed per gigabyte and turns off zero-downtime deploys on that service, so the add-on changes the release. Cron bills only while the job runs, at a per-minute rate, so a nightly task should not be an always-on worker. Workflows bill active task compute instead, which keeps a short task off a paid instance all month.

## Render pricing

Hobby has no monthly fee and allows one member, 25 services, two custom domains, and two environments per project, which keeps it a solo account. A second person in the dashboard means Pro, the flat fee in the table, and seats stop multiplying.

Twenty people cost the same workspace line as two, until the team needs SSO, SCIM, or a HIPAA workspace, which sit on Scale.

A Starter web service is $7/mo and 0.1c-256mb Postgres is $6/mo, so the pair is $13/mo on Hobby before bandwidth. A worker cannot use the free plan, so it is a second Starter and stays billed when idle. Five people add the Pro workspace on top of that web service, worker, and database. The sum is $45/mo before outbound traffic, and the headcount is not inside that sum.

The April 23, 2026 changelog is why a 2025 quote will mislead you, because Professional was $19 per member per month. Five members were $95 before any instance, and twenty were $380. Pro replaced that bill with one workspace charge and unlimited members.

Scale replaced Organization, which had been $29 per member per month. Render's post the same day says 75% of paying customers would see the bill fall or stay flat. A small team with heavy bandwidth or many domains could pay more.

An editor's note says every workspace had been moved by August 1, 2026, and compute prices did not change in that rollout. The instance menu is the part of an old quote you can still trust.

Outbound bandwidth is now the seat fee, then $0.15 per GB after the included pools, shared by every service, including static sites and free web services. A marketing site on the same workspace can eat the API's allowance.

Scale includes 1 TB, inbound is free, and extra custom domains and persistent disk are both $0.25 a month, per domain or per GB.

Postgres storage past the included gigabyte is $0.30 per GB, and a paid disk also disables zero-downtime deploys, so storage changes the release. Build pipeline overage, dedicated outbound IPs, and AWS PrivateLink are separate rows in the table.

Adding a card runs a $1 USD authorization that Render refunds, so the card check is not money you keep. If a Pro or Scale workspace has no services, live or suspended, and no activity all month, Render waives that month's workspace fee. Enterprise publishes no list price, so a contract buyer has to ask.

Railway's subscription is a usage credit, and a stopped service drops CPU and memory, unlike an instance that bills while it runs. Fly.io has no workspace fee, and the small shared machine changes with the region, so the city sets the price.

Northflank's sandbox stays awake, so a preview that must not sleep still costs. App Platform's smallest container is a fixed size on an account that may already pay for Droplets. Heroku's always-on dyno matches Starter and does not grow RAM when you buy metrics.

Prices below were read on each vendor's pricing page on September 24, 2026.

Plan | Price | Best for |

Render Hobby workspace | $0/mo | 1 seat, 25 services, 5 GB outbound, 2 custom domains, 500 build minutes. |

Render Pro workspace | $25/mo | Unlimited seats and services, 25 GB outbound, 15 domains, full-stack previews. |

Render Scale workspace | $499/mo | SSO, SCIM, HIPAA at 20% of compute, 1 TB outbound, multiple workspaces. |

Render Enterprise | Publishes no list price | Contract, uptime SLA, technical account manager, support response targets. |

Render web Free | $0/mo | 512 MB, 0.1 CPU. Spins down after 15 idle minutes. 750 hours per workspace. |

Render web Starter | $7/mo | 512 MB, 0.5 CPU. Same specs for a private service or a background worker. |

Render web Standard | $25/mo | 2 GB, 1 CPU. The jump when 512 MB is no longer enough for the process. |

Render web Pro | $85/mo | 4 GB, 2 CPU. This is compute, not the workspace plan with the same word. |

Render Postgres 0.1c-256mb | $6/mo | 0.1 CPU, 256 MB, 100 connections, 1 GB storage included. Free expires in 30 days. |

Render Key Value Starter | $10/mo | 256 MB, 250 connections, disk persistence. Free is 25 MB and not durable. |

Render bandwidth overage | $0.15/GB | After 5 GB on Hobby or 25 GB on Pro. Inbound traffic is not billed. |

Render disk or extra domain | $0.25/mo | Per GB of persistent disk, or per custom domain past the included count. |

Render Postgres storage | $0.30/GB | Expansion past the 1 GB included with the database plan. |

Render build overage | $5 per 1,000 min | After 500 minutes on Hobby or 1,000 on Pro. A spend cap can halt new builds. |

Render cron Starter | $0.00016/min | 512 MB, 0.5 CPU, billed only while the job runs. |

Render workflow flex CPU | $0.20/active CPU hr | Memory is a separate active rate. Provisioned task sizes are a different row. |

Render dedicated IP set | $100/mo | Pro and above. Outbound addresses unique to the workspace. |

Render AWS PrivateLink | $30/mo | Includes up to 3 links on Pro. Extra outbound on a link is a per-GB rate. |

Railway Hobby | $5/mo | The fee is the usage credit. A stopped service drops CPU and memory. |

Railway Pro | $20/workspace/mo | Matching credit. Seats included. Memory on a running service is the next row. |

Railway memory | $10/GB | Per second while running. Service egress is $0.05/GB. Enterprise is a quote. |

Fly.io shared-cpu-1x | $1.94 to $3.14/mo | 256 MB. The spread is the region tables. No workspace subscription. |

Fly.io NA/EU egress | $0.02/GB | Public internet, granular rates. A card is required on the organization. |

Northflank nf-compute-10 | $2.70/mo | 0.1 shared vCPU, 256 MB. Egress is $0.06/GB. Charges are in USD. |

App Platform shared 512 MiB | Matches Railway Hobby | 1 shared vCPU, 50 GiB transfer. Extra outbound is $0.02/GiB, pooled. |

App Platform extra static app | $3/mo | Three static-only apps are free. A fourth app with a static site is this fee. |

Heroku Eco | Matches Railway Hobby | Personal accounts only. Sleeps after 30 minutes. Two process types. |

Heroku Basic | Matches Render Starter | Always on, 0.5 GB, 10 process types. Team accounts cannot use Eco. |

Heroku Standard-1X | Matches the Pro workspace fee | Per dyno. Metrics and preboot. RAM stays 0.5 GB. |

## Render pros and cons

### What we like

- The smallest paid web plan is half a CPU and 512 MB, so the ceiling is visible before the first deploy.
- Pro dropped the per-member charge on April 23, 2026, and unlimited seats sit on one workspace fee.
- Suspend a paid service and compute billing stops at that second, which an idle paid instance will not do on its own.

### What could be better

- Paid web services and workers do not spin down, so a worker left running bills its plan through every quiet hour.
- Free Postgres expires 30 days after creation, so the free database is a clock and not a home.

## Who Render is for

Render fits a developer who will name the instance and keep the worker and Postgres as separate services. Watch outbound traffic after the included pool, because that is what the invoice assumes. Hobby is the solo account, and Pro is the switch when a teammate needs the dashboard. Pro is also the plan when a pull request has to boot more than one service.

Skip it when the process should cost nothing while it is stopped, which is the comparison on the Railway review. Skip it when the city is the reason you are picking a host, which is [Fly.io](https://toolradar.com/tools/flyio).

Skip it when the rest of the account already pays DigitalOcean, and read the DigitalOcean review before you add App Platform.

A database that has to branch for every preview is a different product. Start with the Neon review and Postgres hosting, and with Supabase if login ships next to the tables.

Leaving dynos is Heroku alternatives and the Heroku review. A static site that should not treat bandwidth as the meter is the Netlify review.

A machine you patch yourself is VPS hosting for developers.

Subscribe free for the next change to this bill.

## Best Render alternatives

If Render is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Render | Teams that will name a web process, a worker, and Postgres before the month starts. | Starter is the 512 MB row in the table | Visit → |

Railway | Backends that should pay for CPU and RAM in use, and drop compute when the process stops. | Hobby is $5/mo and that fee is the usage credit | Visit → |

Fly.io | Machines that have to sit in one city, with CPU charged only while started. | shared-cpu-1x at 256 MB is $1.94 to $3.14/mo across region tables | Visit → |

Northflank | Containers and previews when the sandbox has to stay awake. | From $2.70/mo on nf-compute-10 | Visit → |

DigitalOcean App Platform | Git deploys when the rest of the stack already lives on DigitalOcean. | Smallest shared container matches Railway Hobby | Visit → |

Heroku | Teams that still want a dyno type and managed Postgres on the same invoice. | Eco matches Railway Hobby and is personal-only | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 2 of 6 do not publish a comparable monthly price and are left out rather than estimated.

Railway

Per-second containers whose subscription is also the monthly usage credit.

Visit →

Fly.io

Region-priced Machines whose CPU line exists only while that Machine is started.

Visit →

Northflank

Second-level container billing on their cloud or yours, and a sandbox that stays awake.

Visit →

DigitalOcean App Platform

Shared containers sold at a fixed size, plus a free tier limited to static sites.

Visit →

Heroku

Salesforce's platform, where the dyno type is the line item and sleep is a personal-plan feature.

Visit →

## The bottom line

Pay for Render when you want the instance size written down before the month starts, and someone will suspend staging instead of hoping it sleeps. Start on Hobby with a free web service if a one-minute wake is acceptable.

Move Postgres off the free plan before it expires, and move the workspace to Pro when a teammate needs access.

Leave it when outbound traffic is the product, because the included pool on Hobby is small and every service shares it. Leave it when the worker is idle most of the day, since a paid instance does not spin down.

The closer fit is a credit that follows CPU and RAM, on the [Railway](https://toolradar.com/tools/railway) page and the Railway review.

Choose [Fly.io](https://toolradar.com/tools/flyio) when the region table is acceptable and a workspace fee is not, and choose [Northflank](https://toolradar.com/tools/northflank) when the sandbox must stay awake.

Choose [App Platform](https://toolradar.com/tools/digitalocean-app-platform) when the rest of the account is already on DigitalOcean, and choose [Heroku](https://toolradar.com/tools/heroku) when you still want a dyno name on the invoice.

Subscribe free if you want the next bandwidth note in the Devshot brief.

Cite this: Devshot, "Render Review 2026", September 2026.

## Frequently asked questions

Is Render worth it in 2026?

Yes, when you want a web process, a worker, and Postgres as named sizes, and a second teammate should not multiply the platform fee. Pro is a flat workspace charge with unlimited seats, so the platform line does not grow with the team. It is a weak buy when the worker must stay warm and you wanted idle time to cost nothing. That worker will not spin down on its own, so the quiet hours stay on the invoice. It is also a weak buy when outbound traffic will blow past a few gigabytes, because Hobby's included pool is small and every service shares it.

How much does Render cost?

Hobby has no workspace fee, and Starter compute is $7/mo for 512 MB and 0.5 CPU. 0.1c-256mb Postgres is $6/mo, so that pair is $13/mo before bandwidth, and a worker is another Starter. Five people on Pro, with that web service, worker, and database, come to $45/mo before outbound traffic. Pro itself is $25/mo, and past 5 GB on Hobby or 25 GB on Pro, outbound is $0.15 per GB. Scale is $499/mo, and Enterprise publishes no list price, so the public menu stops at Scale. Rates were checked on Render's pricing page in September 2026.

Does Render have a free plan?

Yes, with hard walls that show up in the first week rather than on a feature grid. A free web service is $0, spins down after 15 idle minutes, and draws on 750 instance hours per workspace each month. One free database and one free Key Value instance are allowed, and the database expires 30 days after creation. You then have 14 days to upgrade before Render deletes it, so move the rows before that clock ends. Static sites have no compute charge and still count against bandwidth and build minutes. Upgrading the workspace to Pro does not remove those free-instance limits, so seats are not free compute.

How does Render compare with Railway?

Render charges for a size you pick, and a paid web process keeps that charge while it is running, requests or not. Railway charges for resources the service holds, and Hobby is $5/mo because that payment is the credit. Pro is $20 per workspace, and stopping the process takes CPU and memory off the bill. Render's small paid database is 0.1c-256mb at $6/mo, with storage as its own line, while Railway's database is a service plus a volume. Choose Render when you want the month's ceiling before you deploy, and choose Railway when hours with the process stopped should not accrue compute.

What changed in Render pricing in 2026?

On April 23, 2026 Render replaced Professional at $19 per member per month with Pro at a flat rate and unlimited members. Five members on the old rate were $95 before compute, twenty were $380, and Organization at $29 per member became Scale. Bandwidth moved to a per-gigabyte overage, and the old $250 a month HIPAA minimum was removed in favor of 20% of compute on Scale. Render says every workspace was on the new plans by August 1, 2026, and instance prices were not part of that change.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Render pricing](https://render.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026

Related guides

Railway ReviewHeroku ReviewHeroku AlternativesVercel ReviewDigitalocean ReviewNeon ReviewSupabase ReviewNetlify ReviewPostgres HostingServerless PlatformsVps Hosting For DevelopersFly Io ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Replit Review

URL: https://devshot.co/reviews/replit-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Honest Replit review for developers: what the Agent actually builds, how effort-based credit pricing works, real prices checked on Replit's pricing page…

Review

## Replit Review

A browser-based IDE with a built-in AI Agent, database and one-click hosting, priced by effort-based credits instead of a flat monthly fee, checked against Lovable, Bolt, Cursor and GitHub Codespaces.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Replit is worth it in 2026 if you want one workspace that builds, hosts and stores data for an app, not just a chat window that writes code. The free Starter plan gets you one 30-day published app and daily Agent credits; Core starts at $20/month ($18/month billed annually), with matching monthly credit toward Agent and cloud usage, and Pro jumps to $100/month ($90/month billed annually) for 10 parallel agents and 28-day database rollback.

Its edge is that the database, hosting and Agent live in one bill, so there is no separate Supabase or Vercel account to wire up. The catch is Pro's price: it costs four to five times what Lovable, Bolt or Cursor charge for their entry paid tier. If you mainly want a fast prototype and already have a database and host you like, Lovable or Bolt start cheaper.

Key factsBest for: Solo builders and small teams who want Agent, a database…

- Updated: September 25, 2026
- Best for: Solo builders and small teams who want Agent, a database and hosting in one workspace and one bill.
- Price as of September 25, 2026: From $20/mo (Core, $18/mo billed annually); free Starter plan capped at one 30-day published app.

- An all-in-one AI coding workspace with a built-in database and hosting, billed by effort, not a flat fee.
- Free plan: Yes, Starter plan with one 30-day published app
- Top plan: Pro, billed monthly or annually
- Alternatives covered: Lovable, Bolt, Cursor, GitHub Codespaces

Pros

- Database, hosting and Agent live in the same workspace, so there is no separate provider to configure.
- Checkpoints snapshot code, database contents and Agent context together, with rollback up to 28 days on Pro.
- Effort-based credits mean small fixes cost less than full feature builds, instead of one flat price per message.

Cons

- Pro costs $100/mo ($90/mo billed annually), four to five times the entry paid tier at Lovable, Bolt or Cursor.
- The free Starter plan limits you to one published app that expires after 30 days, not a lasting free tier.
- A demanding week of Full build work can burn through Core's entire included credit before the month resets.

Free planYes, Starter plan with one 30-day published app

Starting priceCore plan, billed monthly or annually

Pricing modelEffort-based credits, not per seat

Top planPro, billed monthly or annually

Replit started as a browser code playground and has turned into a full AI app-building platform: type a prompt into Agent and it scaffolds a working app, provisions a Postgres database, and publishes it to a live URL, all inside the same tab.

Built by Replit, Inc., it now sits alongside Lovable, Bolt and v0 in the vibe coding conversation developers keep having about which tool actually ships a working product.

The real question for a developer, not a hobbyist testing a demo, is whether Replit's all-in-one workspace beats stitching together a chosen database, a chosen host and a separate AI coding tool.

This review covers what Replit Agent builds, how its effort-based credit pricing actually charges you, and where the database and hosting bundle helps or gets in the way. It compares Replit against Lovable, Bolt, Cursor and GitHub Codespaces so a developer can see where the all-in-one workspace pays off and where a specialized tool wins on price or control.

Solo builders and small teams who want fewer accounts to manage are the audience Replit fits best.

## What is Replit?

Replit is a cloud development platform built by Replit, Inc. that runs an entire coding environment, an AI Agent, a Postgres database and app hosting inside a browser tab, no local setup required.

You open a workspace, describe an app to Agent in plain language, and it writes the code, wires up a database when the app needs one, and can publish the result to a replit.app subdomain or a custom domain.

Unlike a chat-only app builder, the workspace stays open after Agent finishes, so you can read and edit every file it wrote in a real code editor.

Beyond the chat-to-app flow, Replit adds a visual Design Canvas for exploring layout ideas before Agent builds them, a Plan Mode for brainstorming a feature without touching code, and Connectors for wiring in services like Google and Stripe without hand-rolled API setup.

Checkpoints snapshot the whole project state, including database contents and Agent conversation context, so a bad change rolls back cleanly. Teams get collaboration seats, a shared Kanban board and, on Pro, database rollback reaching back 28 days.

It sits closer to a full development platform than a single-purpose code generator.

## How Replit works

A Replit project starts the same way on every plan: describe what you want to build, and Agent scaffolds the app, choosing a Lite build for quick edits or a Full build for autonomous, multi-step work.

Free Mode covers everyday chat and small changes at no extra cost, within a daily allowance; once you move to Power or Max mode for heavier work, Agent starts drawing from your plan's monthly credit balance.

Cost follows effort, not a per-message price: Replit bundles a request's changes into one checkpoint, so a small bug fix typically costs less than a checkpoint that adds a full feature with a new integration.

Every checkpoint is a complete snapshot, covering code, the database and Agent's conversation context, and you can roll forward or backward through that history from the Agent tab.

The Starter plan restricts you to Lite build and one published app that expires after 30 days; Core unlocks Full build, unlimited published apps, Plan Mode and third-party connectors; Pro adds parallel agents and a much longer rollback window.

The rough edge is the same credit meter every effort-based tool shares: a productive week of Full build work on Core can burn through its entire included model credit before the month resets, pushing you toward a credit pack or Pro.

## Replit key features

Agent with Lite and Full build modesEssential
Agent writes code from a plain-language prompt, using Lite build for quick, targeted edits on the free Starter plan or Full build for autonomous, multi-step work once you upgrade to Core. It is the core workflow the rest of the platform supports.

Effort-based checkpoint pricingEssential
Every completed request becomes one checkpoint, billed by how much work it took: a simple fix costs less than a checkpoint that bundles a new feature and an integration. Core and Pro subscribers get monthly credit toward this usage, and Free Mode covers everyday work at no cost within a daily allowance.

Built-in Postgres database and hostingEssential
Apps that need one get a Postgres database provisioned automatically inside the same workspace, and one-click publishing puts the app on a replit.app subdomain or a custom domain without a separate hosting account.

Checkpoints and rollback
Every checkpoint snapshots the full project state, including database contents and Agent's conversation context, so you can roll back a bad change with one click. Pro extends this window to 28 days, well beyond Core's shorter history.

Design Canvas and Plan Mode
Design Canvas lets you explore visual layouts and convert one into a working artifact on Core and Pro; Plan Mode lets you brainstorm and scope a feature with Agent before any code changes, which Replit's own docs recommend as a way to cut effort-based costs.

Collaboration and Kanban planning
Core includes up to 5 collaborator seats and Pro up to 15, plus a shared Kanban board for tracking work across a small team, all inside the same workspace rather than a bolted-on project management tool.

## Replit pricing

Every price below comes straight off each vendor's own pricing page, read on the day this review was written, September 2026.

All five list a public price; only the Enterprise tier at Replit, Lovable, Bolt and Cursor hides behind a sales call. Replit and Lovable share a similar mechanic: a credit balance that drains faster on effort-heavy work, so the plan price is a floor, not a ceiling, on either platform.

The cheapest paid entry among the app builders is Cursor Pro, tied with Replit Core's monthly rate. Replit Pro is the most expensive plan in this review, well above Lovable Business at $50/month or Bolt Teams at $30 per member.

GitHub Codespaces is not priced like the other four: instead of a flat plan fee, it bills compute at $0.18 an hour on a 2-core machine and storage at $0.07 per GB-month, layered on top of a GitHub account plan.

Devshot data: [Toolradar's AI app builders guide](https://toolradar.com/guides/best-ai-app-builders), refreshed in May 2026 with pricing checked through September 2026, tracks 144 tools in the category and lists Replit, Lovable and Bolt among its top picks.

See each tool's full profile on Toolradar: [Replit](https://toolradar.com/tools/replit), [Lovable](https://toolradar.com/tools/lovable), [Bolt](https://toolradar.com/tools/bolt), [Cursor](https://toolradar.com/tools/cursor) and [GitHub Codespaces](https://toolradar.com/tools/github-codespaces).

Plan | Price | Best for |

Replit Starter | $0 | Daily Agent credits, Lite build only, one published app for 30 days |

Replit Core | $20/mo | Monthly credit matching the plan price, Full build, up to 5 collaborators, unlimited published apps |

Replit Core (annual) | $18/mo billed annually | Same as Core, billed yearly |

Replit Pro | $100/mo | Monthly credit matching the plan price, 10 parallel agents, up to 15 collaborators, 28-day database rollback |

Replit Pro (annual) | $90/mo billed annually | Same as Pro, billed yearly |

Replit Enterprise | Custom quote | SSO/SAML, custom seat limits, single-tenant environments |

Lovable Pro | $25/mo | 100 credits a month, no cap on workspace members, custom domain support |

Lovable Business | $50/mo | Same credit count as Pro, plus SSO, role-based access and priority support |

Lovable Enterprise | Custom platform fee | Volume-based pricing, SCIM, audit logs, custom SLA |

Bolt Pro | $25/mo | Token ceiling lifts entirely per day, 10M-token starting allowance each month |

Bolt Teams | $30/mo per member | Shared org billing with per-seat access controls |

Bolt Enterprise | Custom quote | Compliance support plus a named account contact |

Cursor Pro | $20/mo | Wider Agent usage pool and access to top-tier models |

Cursor Pro Plus | $60/mo | Recommended tier for using the Agent every day |

Cursor Ultra | $200/mo | Biggest usage pool, built for constant agent workloads |

Cursor Teams (Standard) | $40/user/mo | Shared billing, single sign-on, automated code review |

Cursor Enterprise | Custom quote | Usage pooled org-wide with SCIM provisioning and audit trails |

GitHub Codespaces compute | $0.18/hour (2-core) | Scales with machine size, up to $2.88/hour on a 32-core machine |

GitHub Codespaces storage | $0.07/GB-month | Billed for usage beyond your plan's free quota |

GitHub Team | $4/user/mo | First 12 months; unlocks Codespaces for organizations, no free quota included |

GitHub Enterprise | $21/user/mo | First 12 months; additional security, compliance and deployment controls |

## Replit pros and cons

### What we like

- Database, hosting and Agent live in the same workspace, so there is no separate provider to configure.
- Checkpoints snapshot code, database contents and Agent context together, with rollback up to 28 days on Pro.
- Effort-based credits mean small fixes cost less than full feature builds, instead of one flat price per message.

### What could be better

- Pro costs $100/mo ($90/mo billed annually), four to five times the entry paid tier at Lovable, Bolt or Cursor.
- The free Starter plan limits you to one published app that expires after 30 days, not a lasting free tier.
- A demanding week of Full build work can burn through Core's entire included credit before the month resets.

## Who Replit is for

Replit is a strong fit for solo builders and small teams who want one bill and one workspace instead of stitching together a database provider, a host and a coding assistant.

It suits developers building a full-stack prototype who would rather let Agent provision Postgres and publish the result than configure that infrastructure themselves, and teams that value the 28-day database rollback on Pro as a real safety net for a production-adjacent app.

Educators and teams that already run classes or workshops inside Replit's browser IDE get extra mileage from the platform beyond just Agent.

Who should skip it: budget-conscious solo developers who only need fast UI generation, since Lovable and Bolt's entry paid tiers cost the same as Replit Core with a gentler credit meter on Bolt's side.

Teams that have already settled on Supabase, Vercel or another dedicated host get less value from Replit's bundling and more control from a tool built around that stack. Engineers who want AI woven into an existing large codebase, not a new app generated from scratch, are better served by an editor like Cursor.

Heavy Agent users should budget past the sticker price: on Core, a demanding week can outrun its included credit well before Pro's monthly rate starts to look proportionate.

## Best Replit alternatives

If Replit is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Replit | Solo builders and small teams who want Agent, a database and hosting in one workspace and one bill. | From $20/mo (Core, $18/mo billed annually) | Visit → |

Lovable | Non-technical founders and developers who want a full-stack app from a chat prompt, backed by Supabase. | From $25/mo (Pro, 100 monthly credits) | Visit → |

Bolt | Developers who want a fast browser prototype with full code control and no daily usage cap once paid. | From $25/mo (Pro, no daily token cap) | Visit → |

Cursor | Engineers who already own a codebase and want AI help inside it, rather than a hosted app-generation tool. | From $20/mo (Pro) | Visit → |

GitHub Codespaces | Teams already on GitHub who want a cloud dev environment tied to their existing repos and billing. | Pay-as-you-go: $0.18/hour compute (2-core), $0.07/GB-month storage | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Lovable

An AI app builder that outputs a working product with a Supabase-backed database, not just a UI.

Visit →

Bolt

StackBlitz's AI builder, fast in the browser with token-metered pricing and no daily ceiling on Pro.

Visit →

Cursor

A code editor that adds an AI Agent to work you already own, instead of generating a new hosted project.

Visit →

GitHub Codespaces

GitHub's own cloud dev environment, billed by the hour on top of your existing GitHub plan.

Visit →

## The bottom line

Replit earns its price for a developer who wants Agent, a database and hosting under one login instead of three separate accounts. The checkpoint system, which snapshots code and database state together, and Pro's 28-day rollback window are genuine safety features none of the other four match in depth.

Core's entry price ties Cursor Pro's monthly rate for a very different kind of tool: a hosted workspace instead of an editor plugin.

The trade-off is Pro's price and the credit meter underneath every tier: its monthly rate is steep next to Lovable Business at $50 or Bolt Teams at $30 per member, and Core's included credit does not stretch through a demanding week of Full build work.

Choose Lovable if you want a cheaper full-stack entry point with unlimited workspace members, Bolt if you want no daily token cap once paid, and GitHub Codespaces if your team already lives in GitHub repos and just needs a cloud dev box. Pick Cursor instead if you are extending an existing codebase rather than generating a new app from scratch. Buy Replit when the bundled database and hosting save more setup time than the credit meter costs you.

Methodology note: every figure in this review, plan names, prices and the effort-based billing mechanics, was pulled from Replit's own pricing and docs pages and each rival's pricing page on the date of publication.

Lovable, Bolt, Cursor and GitHub Codespaces were picked because they are the tools Devshot readers ask us to weigh against Replit in the vibe coding and cloud dev environment space.

Cite this: Devshot, "Replit Review 2026: Pricing, Pros, Cons and Alternatives," September 2026.

## Frequently asked questions

How much does Replit cost?

Replit's free Starter plan gives daily Agent credits and one published app that expires after 30 days. Core is the cheapest paid plan, priced the same per month as Cursor Pro, or $18/month billed annually, and includes monthly credit matching its price toward Agent and cloud usage plus up to 5 collaborator seats. Pro costs $100/month, or $90/month billed annually, with monthly credit matching that price, up to 15 collaborators and database rollback up to 28 days. Enterprise is a custom quote with SSO, custom seat limits and single-tenant environments.

Is Replit worth it in 2026?

For a developer who wants Agent, a Postgres database and hosting in one workspace, yes: the checkpoint system and Pro's 28-day rollback are real safety features the other tools in this review do not match. For a budget-conscious solo builder who only needs fast UI generation, Lovable or Bolt's entry paid tier costs the same as Core with a less aggressive credit meter.

Does Replit have a free plan?

Yes, the Starter plan, but it is limited: daily Agent credits within a monthly cap, Lite build only (not the more autonomous Full build), and one published app that automatically goes down after 30 days. Lovable and Bolt's free plans are more generous for ongoing use, since neither expires a published app on a fixed timer the way Replit's Starter plan does.

What are the best Replit alternatives?

Lovable is the closest match on output, generating a full-stack app with a Supabase-backed database at a lower entry price. Bolt suits developers who want no daily token cap once they pay. Cursor is a different category, an AI-first editor for extending an existing codebase rather than generating a new hosted app. GitHub Codespaces fits teams that already work in GitHub repos and just need a billed-by-the-hour cloud dev environment.

How does Replit's effort-based pricing work?

Instead of charging a flat rate per message, Replit bundles the changes from one request into a single checkpoint and bills it by how much work it took: a small fix like a copy change typically costs less than a checkpoint that adds a full feature with a new integration, per Replit's own billing docs. Free Mode covers everyday chat and small edits at no cost within a daily allowance, and Core and Pro subscribers draw from their monthly credit balance once they move to a paid Agent mode. Replit's docs recommend using Plan Mode to scope a change before it runs, since planning conversations use less of that credit than jumping straight into a build.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Replit pricing](https://replit.com/pricing)checked Sep 2026
- [Lovable pricing](https://lovable.dev/pricing)checked Sep 2026
- [Bolt pricing](https://bolt.new/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026

Related guides

Vibe Coding ToolsLovable ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Supabase Review

URL: https://devshot.co/reviews/supabase-review
Type: review
Published: 2026-08-13
Updated: 2026-09-25
Summary: An honest Supabase review for 2026: what the free tier really gives you, where the Pro plan's usage billing bites, and how it compares to Firebase…

Review

## Supabase Review

Postgres with the boring parts done for you, and no proprietary query language to unlearn later.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Supabase is the best backend-as-a-service for teams that want to keep a real Postgres database, bundling auth, storage, edge functions and realtime around it. Free tier covers 50,000 monthly active users, 500 MB of database and 2 projects; Pro is $25 a month; Team is $599. The catch is usage-based overage on top of the base fee, which you should model before launch.

Key factsBest for: Full-stack applications that want to keep real Postgres

- Updated: September 25, 2026
- Best for: Full-stack applications that want to keep real Postgres
- Price as of September 25, 2026: From $25/mo (Pro); free plan ($0).

- Firebase-speed development on an actual Postgres database, open source and self-hostable if the hosted plan stops fitting.
- Founded: 2020
- Core: PostgreSQL, open source
- Alternatives covered: Firebase, Neon, PlanetScale, Appwrite

Pros

- It is real Postgres: full SQL, extensions, and your data is portable by design
- Row-level security puts authorisation in the database where it is hardest to bypass
- Free tier is genuinely enough to build and demo a real product

Cons

- The $25 Pro fee is a base, and egress overage is the line that surprises image-heavy apps
- Free projects pause after a week idle, so they are unsuitable for anything customer-facing
- SOC 2 and ISO 27001 sit on the $599 Team tier, a steep jump for a small company that needs compliance

Supabase

Founded2020

CorePostgreSQL, open source

Entry priceFree, then $25/mo

Best forFull-stack apps on Postgres

The pitch that made Supabase popular is simple and mostly true: you get Firebase's speed of setup without Firebase's lock-in, because underneath it is just Postgres. Your data lives in tables you can query with SQL, dump with pg_dump and move somewhere else.

That single design decision is why a lot of teams who once reached for Firebase now reach for this instead. This review covers what you actually get at each tier and where the bill surprises people.

## What is Supabase?

Supabase is an open-source backend platform built on PostgreSQL.

Every project provisions a real Postgres database, then wraps it in the services an application needs: an auto-generated REST and GraphQL API, authentication with social and enterprise providers, row-level security policies enforced in the database itself, file storage, realtime subscriptions over websockets, edge functions running Deno, and vector search for AI features.

The whole stack is open source and self-hostable, which is the practical exit route if the hosted pricing stops working for you.

## How Supabase works

You create a project and get a Postgres instance plus a dashboard with a table editor and SQL editor. Client libraries for JavaScript, Flutter, Python, Swift and others talk to the auto-generated API, so a working CRUD app needs no backend code.

Access control is the part worth understanding: instead of writing an authorisation layer, you write row-level security policies in SQL that the database enforces on every query, no matter which client sends it.

Migrations are managed through the CLI and versioned in your repo, and edge functions handle the logic that does not belong in the client.

## Supabase key features

Real PostgreSQL
Not a Postgres-like layer: an actual instance with extensions, foreign keys, transactions and full SQL. Your schema knowledge and your data both remain portable.

Row-level security
Authorisation rules live in the database as policies, enforced on every query. Fewer places for an access bug to hide than a hand-rolled middleware layer.

Auth with social and SSO
Email, magic links, OAuth providers, phone and SAML on higher tiers, wired to the same database as your application data.

Realtime subscriptions
Listen to inserts, updates and deletes over websockets, plus broadcast and presence for collaborative features.

Edge functions and vector search
Deno functions at the edge for server-side logic, and pgvector for embeddings, which makes AI features a schema decision rather than a second database.

Open source and self-hostable
The whole platform can run on your own infrastructure. Even if you never use it, this materially changes the negotiation.

## Supabase pricing

Supabase publishes list prices: a free plan at $0, Pro at $25 per month, and Team at $599 per month. Enterprise is a custom quote. Firebase Blaze is pay-as-you-go above the free Spark tier, Neon Launch bills $0.106 per CU-hour with no monthly minimum, and Appwrite Pro charges the same entry fee as Supabase, per project. PlanetScale has no free tier, and its single-node Postgres starts at $5 a month.

The cheapest entry is Supabase's free plan, then Pro at $25 a month, but that fee is a base, not a cap: overage is roughly $0.00325 per extra monthly active user, $0.125 per GB of disk, $0.09 per GB of egress, and $0.0213 per GB of file storage.

Costs jump on egress for image-heavy apps, and again on Team, where SOC 2 and ISO 27001 sit.

Plan | Price | Best for |

Supabase Free | $0 | 50k MAU, 500 MB database, 1 GB storage, 2 projects |

Supabase Free | $0 | Shared CPU, 500 MB RAM, pauses after a week idle |

Supabase Free | $0 | 5 GB egress included |

Supabase Pro | $25/mo, then usage-based | 100k MAU, 8 GB disk, 250 GB egress, 100 GB storage |

Supabase Pro | $25/mo | Email support and 7-day backups |

Supabase Pro extra MAU | roughly $0.00325 per extra MAU | Overage past included monthly active users |

Supabase Pro extra disk | roughly $0.125 per GB of disk | Overage past included disk |

Supabase Pro extra egress | roughly $0.09 per GB of egress | Overage past included egress |

Supabase Pro extra storage | roughly $0.0213 per GB of file storage | Overage past included file storage |

Supabase Team | $599/mo | SOC 2, ISO 27001, SLAs, and 14-day backups |

Supabase Team | $599/mo | Same quotas, priority support, project-scoped access controls |

Supabase Enterprise | Custom quote | Custom enterprise plan |

Firebase Spark | Free | Free tier before Blaze pay-as-you-go |

Firebase Blaze | Pay-as-you-go | Paid plan above the free Spark tier |

Neon Free | Free | Free tier before usage-based paid plans |

Neon Launch | $0.106/CU-hour | Usage-based, no monthly minimum, storage $0.35/GB-month |

PlanetScale Postgres single node | From $5/mo | Development and low-traffic workloads; no free tier |

Appwrite Free | Free | 2 projects, 75K MAU, pauses after a week idle |

Appwrite Pro | $25/mo | One project with 2 TB bandwidth, 150 GB storage, 200K MAU |

Appwrite Enterprise | Custom quote | Custom limits, premium support and security features |

## Supabase pros and cons

### What we like

- It is real Postgres: full SQL, extensions, and your data is portable by design
- Row-level security puts authorisation in the database where it is hardest to bypass
- Free tier is genuinely enough to build and demo a real product
- Open source and self-hostable, which removes the usual lock-in fear
- pgvector makes AI features a schema change rather than a second system

### What could be better

- The $25 Pro fee is a base, and egress overage is the line that surprises image-heavy apps
- Free projects pause after a week idle, so they are unsuitable for anything customer-facing
- SOC 2 and ISO 27001 sit on the $599 Team tier, a steep jump for a small company that needs compliance
- Self-hosting is real work, not a weekend, despite being possible

## Who Supabase is for

Supabase suits startups and product teams building a normal web or mobile application on relational data, particularly anyone who values SQL and wants an exit path. It is excellent for AI products, since pgvector lives in the same database as everything else.

It is also a strong choice for solo developers, because the free tier genuinely supports a real prototype.

It is a weaker fit for applications with unpredictable heavy egress on a tight budget, for teams needing hard multi-region write latency guarantees, and for organisations that require SOC 2 without paying $599 a month, since compliance sits on the Team tier.

## Best Supabase alternatives

If Supabase is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Supabase | Full-stack applications that want to keep real Postgres | From $25/mo (Pro) | Visit → |

Firebase | Mobile apps needing offline sync and Google infrastructure | Free Spark tier | Visit → |

Neon | Teams wanting serverless Postgres with database branching | Free plan (100 CU-hours per project) | Visit → |

PlanetScale | MySQL teams needing serious horizontal scale | Postgres single node from $5/mo | Visit → |

Appwrite | Teams that want a self-hosted open-source backend | Free plan | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

Firebase

The original backend-as-a-service, unmatched for mobile offline sync, at the cost of a NoSQL model you cannot easily leave.

Visit →

Neon

Serverless Postgres with instant branching, so every pull request can have its own copy of the database.

Visit →

PlanetScale

MySQL built on Vitess with schema branching, aimed at applications that will outgrow a single database node.

Visit →

Appwrite

The closest open-source alternative in spirit, with auth, database, storage and functions, designed to be self-hosted easily.

Visit →

## The bottom line

Supabase is the default choice in 2026 for a new application on relational data, and the reason is portability: you get the development speed of a backend-as-a-service without betting your data model on a proprietary store.

Start on the free tier, but move to Pro before anything customer-facing, because idle projects pause. Model your egress before launch, since that is where the bill grows. Choose Neon instead if you only want the database and love branching, and Firebase if offline mobile sync is the core requirement.

## Frequently asked questions

How much does Supabase cost?

The free tier costs nothing and includes 50,000 monthly active users, a 500 MB database and 2 projects. Pro is $25 per month with much larger quotas plus usage-based overage, and Team is $599 per month adding SOC 2, ISO 27001 and SLAs. Budget for overage on egress and storage rather than treating $25 as a ceiling.

Is Supabase free tier good enough for production?

No, for one specific reason: free projects pause after a week of inactivity, so a low-traffic production app can go offline. The free tier is excellent for prototypes, demos and side projects. Anything with real users belongs on Pro.

Is Supabase better than Firebase?

For relational data, usually yes, because Supabase is real Postgres with SQL, joins and portable data, while Firebase is a document store that is hard to leave. Firebase still wins decisively for mobile offline sync and has a more mature mobile SDK ecosystem. Pick on your data model and whether offline-first matters.

Can you self-host Supabase?

Yes, the entire platform is open source and there are official Docker configurations for running it yourself. Be realistic about the effort: you take on upgrades, backups, scaling and monitoring for several services. Most teams never self-host, but the option meaningfully reduces lock-in risk.

What is row-level security in Supabase?

It is Postgres's own authorisation mechanism, exposed as the main way you control access. You write policies in SQL that decide which rows each user can read or modify, and the database enforces them on every query regardless of which client made it. It is harder to bypass than checking permissions in application code, but it must be enabled deliberately on each table.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Supabase pricing](https://supabase.com/pricing)checked Sep 2026
- [Firebase pricing](https://firebase.google.com/pricing)checked Sep 2026
- [Neon pricing](https://neon.tech/pricing)checked Sep 2026
- [PlanetScale pricing](https://planetscale.com/pricing)checked Sep 2026
- [Appwrite pricing](https://appwrite.io/pricing)checked Sep 2026

Related guides

Self Hosting ToolsGraphql ToolsVercel ReviewFree Database HostingAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Vercel Review

URL: https://devshot.co/reviews/vercel-review
Type: review
Published: 2026-08-04
Updated: 2026-09-25
Summary: Honest Vercel review for engineers: git-push deploys, preview URLs, serverless and edge functions, v0, and the AI SDK.

Review

## Vercel Review

The Frontend Cloud from the team behind Next.js. Git-push deploys, instant preview URLs, and managed edge infrastructure, free to start and usage-based as you scale.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Vercel is the managed Frontend Cloud built by the team behind Next.js, and for a large slice of the web it is the default way to ship a React app. The workflow is the whole pitch: connect a Git repo, push a commit, and get a production URL plus a unique preview URL for every branch and pull request in seconds. Pricing starts free on the Hobby plan, then Pro is $20 per user per month with $20 of included usage credit, and Enterprise is custom, with everything above the included limits billed as usage. Its biggest strength is developer experience: deploys, previews, edge and serverless functions, and now v0, the AI SDK, and Fluid compute all work with almost no setup. The biggest catch is cost predictability: usage-based bills can spike on a traffic surge or a busy serverless endpoint, Pro is per seat, and you build on Vercel's primitives. The closest alternatives are Netlify, Railway, Render, and Cloudflare Pages.

Key factsBest for: Frontend and full-stack teams on Next.js or modern JS…

- Updated: September 25, 2026
- Best for: Frontend and full-stack teams on Next.js or modern JS frameworks who want git-push deploys, preview URLs, and managed edge infrastructure.
- Price as of September 25, 2026: From $20/user/mo (Pro); free Hobby plan.

- The Frontend Cloud from the makers of Next.js, with the best deploy-and-preview workflow in the business and usage-based pricing to match.
- Founded: 2015
- Headquarters: San Francisco, CA
- Alternatives covered: Netlify, Railway, Render, Cloudflare Pages

Pros

- Best-in-class git-push deploys with a preview URL for every branch and PR.
- Zero-config Next.js, edge and serverless functions, and strong performance defaults.
- Fast-growing AI stack: v0, the AI SDK, AI Gateway, and Fluid compute.

Cons

- Usage-based bills can spike on a traffic surge or a busy serverless endpoint.
- Pro is per seat, and v0 plus some features cost extra on top.
- Best for Next.js and frontend, not heavy backends, and you build on Vercel's primitives.

Founded2015

HeadquartersSan Francisco, CA

Est. priceFree to $20/user/mo + usage

Best forNext.js and frontend teams

Vercel is the company behind Next.js, and for a huge slice of the frontend world it has become the default answer to a simple question: I have a repo, how do I get it live. You connect a Git provider, push a commit, and a few seconds later you have a production URL, a global CDN, and a unique preview URL for every branch and pull request.

That workflow is the reason Vercel shows up in so many engineering stacks, and it is why the company markets itself as the Frontend Cloud.

This review is written for frontend and full-stack engineers deciding whether to run production on Vercel or self-host on a raw cloud like AWS.

We look at what you actually get, how deploys and preview environments work, where serverless and edge functions fit, what the v0 AI builder and the AI SDK add, and the honest trade-offs around usage-based bills, per-seat pricing, and lock-in.

We also cover who should look elsewhere, and five alternatives worth pricing before you commit.

## What is Vercel?

Vercel is a managed deployment and hosting platform for web applications, founded in 2015 (originally as ZEIT) and headquartered in San Francisco.

It is best known as the maker of Next.js, the React framework a large share of modern web apps are built on, though Vercel also deploys Astro, SvelteKit, Nuxt, Remix, Vite, and plain static sites.

The core idea is that you should not have to assemble your own infrastructure to ship a fast web app.

Vercel gives you Git-connected continuous deployment, a global edge network, serverless and edge functions, image optimization, caching and incremental static regeneration, analytics, and a web application firewall, all behind one dashboard and one CLI.

You write the app, Vercel runs the build, distributes the output, and handles scaling from zero to millions of requests without you provisioning servers.

More recently Vercel has leaned hard into AI. It ships v0, a prompt-to-app builder that generates React and Tailwind UI, the open-source AI SDK for wiring models into apps, an AI Gateway for routing between providers, and Fluid compute, a function model designed to keep AI and backend workloads efficient.

The positioning has shifted from pure frontend hosting toward infrastructure for AI-driven apps and agents.

## How Vercel works

The everyday loop is git push. You link a GitHub, GitLab, or Bitbucket repo, and every push triggers a build. Merges to your production branch deploy to your live domain, and every other branch and pull request gets its own preview URL on a stable subdomain.

Those preview deployments are one of Vercel's best features: reviewers, designers, and QA can click a link and see the exact change running in production-like conditions before it merges, which kills a whole class of "works on my machine" arguments.

Dynamic code runs as functions. Serverless functions handle API routes and server rendering in a chosen region, while edge functions run lighter logic close to users across the network.

With Fluid compute, functions can stay warm and reuse compute across concurrent requests instead of cold-starting one instance per call, which matters a lot for AI endpoints that sit idle waiting on a model response.

Around that sit the managed extras: image optimization, ISR for regenerating static pages on demand, edge caching, per-environment variables, and observability. Everything is configured through the dashboard, the vercel CLI, or a vercel.json file.

The trade-off for that polish is that you work in Vercel's primitives. When you need something the platform does not expose, such as a long-running worker, a persistent socket, or a custom container, you either wait for Vercel to ship it or drop down to a rawer cloud.

## Vercel key features

Git-push deploys and preview URLsEssential
Connect a repo and every push builds and deploys automatically, with a unique preview URL for each branch and pull request. Production ships on merge to your main branch. This is the core loop, and it is fast, reliable, and the single biggest reason teams pick Vercel.

Serverless and edge functionsEssential
Run API routes and server rendering as serverless functions in a region, or push light logic to edge functions near users. Fluid compute keeps functions warm and shares compute across requests, cutting cold starts and making it viable to host AI and backend endpoints alongside the frontend.

Next.js and framework supportEssential
Vercel makes Next.js, so features like the App Router, server components, ISR, and image optimization work with zero config. It also deploys Astro, SvelteKit, Nuxt, Remix, Vite, and static output, but the deepest, first-day support is always for Next.js.

Global edge network and caching
Static assets and cached responses are served from a global CDN, with incremental static regeneration to rebuild pages on demand instead of a full redeploy. Built-in image optimization and edge caching mean strong performance defaults without you tuning a cache layer yourself.

AI SDK and AI Gateway
The open-source AI SDK gives a unified TypeScript API for calling models, streaming responses, and building tool-using agents, while the AI Gateway routes and load-balances across providers. Together they make Vercel a genuinely convenient place to build AI features, not just host static sites.

v0 AI app builder
v0 turns a text prompt or a Figma import into working React and Tailwind components and full-stack scaffolding you can deploy in one click. It is a paid product with its own credit-based pricing, useful for prototyping and design-to-code, though generated code still needs review.

## Vercel pricing

Vercel, Netlify, Railway, Render, and Cloudflare Pages all publish entry prices, and Enterprise is a custom quote at Vercel, Netlify, and Cloudflare Pages. Vercel, Netlify, Render, and Cloudflare Pages include a free plan, while Railway's free plan is a 30-day trial followed by $1 of monthly credit.

The cheapest credible paid entry is $5 a month, on Railway Hobby and on Cloudflare Workers Paid. Costs jump on usage past included credits and allotments, on per-seat plans, and on v0, which Vercel bills separately from hosting, even with the default spend cap and alerts.

Plan | Price | Best for |

Vercel Hobby | Free forever | Personal projects, non-commercial fair use |

Vercel Pro | $20/user/mo, includes $20 credit | Faster builds, cold-start prevention, teams |

Vercel Pro edge requests | 10 million/mo, then $2/million | Edge requests above the included allotment |

Vercel Pro data transfer | 1 TB included, then $0.15/GB | Fast data transfer above the allotment |

Vercel Pro Active CPU | About $0.128 per CPU-hour | Compute billed on Active CPU |

Vercel Pro invocations | $0.60 per million invocations | Function invocation overage on Pro |

Vercel Pro memory | $0.0106 per GB-hour | Provisioned memory for functions |

Vercel image, blob, ISR | Metered | Image optimization, blob storage, and ISR |

Vercel Enterprise | Custom quote | SSO, SCIM, managed WAF, multi-region failover, guaranteed support, 99.99% SLA |

Vercel v0 | Free, $5/mo of credits | Free tier, monthly generation credits |

Vercel v0 Premium | $20/mo of credits | Premium tier, monthly generation credits |

Vercel v0 Team | $30/user/mo | Team tier, billed per user |

Netlify Free | Free | 300 usage credits a month, one member |

Netlify Pro | From $20/mo | Unlimited members, 3,000 credits a month and up |

Netlify Enterprise | Custom quote | Quote-only enterprise plan |

Railway Hobby | $5/mo, includes $5 usage | Free plan is a 30-day trial, then $1 of monthly credit |

Railway Pro | $20/mo minimum usage | Fee returns as usage credit; unlimited workspace seats |

Render free | Free | Static sites and small services |

Render instances | From about $7/mo | Paid instances above the free tier |

Render Pro | $25/mo + compute | Flat workspace fee with unlimited seats and services |

Cloudflare Pages free | Free | Unlimited static requests, 500 builds a month |

Cloudflare Workers Paid | $5/mo | Higher function limits than the free plan |

Cloudflare Enterprise | Custom quote | Quote-only enterprise plan |

## Vercel pros and cons

### What we like

- Best-in-class git-push deploys with a preview URL for every branch and PR.
- Zero-config Next.js, edge and serverless functions, and strong performance defaults.
- Fast-growing AI stack: v0, the AI SDK, AI Gateway, and Fluid compute.

### What could be better

- Usage-based bills can spike on a traffic surge or a busy serverless endpoint.
- Pro is per seat, and v0 plus some features cost extra on top.
- Best for Next.js and frontend, not heavy backends, and you build on Vercel's primitives.

## Who Vercel is for

Vercel is a strong fit for frontend and full-stack teams building on Next.js or another modern JavaScript framework who want to ship fast and not babysit infrastructure. Startups, agencies, and product teams that value preview deployments, instant rollbacks, and good performance defaults get the most out of it.

If your app is content, dashboards, marketing sites, or an AI feature layered on a React frontend, Vercel is one of the best developer experiences you can buy.

It is a weaker fit in a few clear cases. If your workload is a heavy backend, such as long-running jobs, websockets at scale, big data processing, or containerized services that need full control, a general cloud or a container platform will serve you better and often cheaper.

High-traffic sites with thin margins can find usage-based bandwidth and compute expensive versus a flat VPS. And teams that want to avoid platform lock-in, or that already run deep on AWS, GCP, or Azure, may prefer to keep hosting where the rest of their infrastructure lives.

If cost predictability matters more than developer experience, compare carefully before you commit.

## Best Vercel alternatives

If Vercel is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Vercel | Frontend and full-stack teams on Next.js or modern JS frameworks who want git-push deploys, preview URLs, and managed edge infrastructure. | From $20/user/mo (Pro) | Visit → |

Netlify | Frontend and Jamstack teams who want a Vercel-style deploy workflow that is framework-agnostic rather than Next.js-first. | From $9/mo (Personal) | Visit → |

Railway | Full-stack developers who want to deploy backends, databases, and services, not just a frontend, with minimal DevOps. | From $5/mo (Hobby) | Visit → |

Render | Teams that want a Heroku-style all-in-one for web services, static sites, cron jobs, and managed databases. | From about $7/mo (instances) | Visit → |

Cloudflare Pages | Cost-conscious teams already on Cloudflare who want static hosting plus Workers at the edge with a generous free tier. | From $5/mo (Workers Paid) | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Netlify

The other big frontend host, with a similar git-push and preview experience across many frameworks.

Visit →

Railway

A developer-friendly platform for shipping backends, databases, and full apps with usage-based compute.

Visit →

Render

A unified cloud for web services, static sites, workers, and databases with predictable per-service pricing.

Visit →

Cloudflare Pages

Static hosting plus Workers on Cloudflare's edge, with one of the most generous free tiers around.

Visit →

## The bottom line

Vercel earns its reputation. For a frontend or full-stack team building on Next.js or a modern JavaScript framework, it is the fastest path from a Git repo to a fast, globally distributed, production app, and the preview-URL-per-PR workflow alone is worth a lot.

The AI stack, v0, the AI SDK, AI Gateway, and Fluid compute, makes it a credible place to build AI features rather than just host static pages.

The trade-off is cost and control. Usage-based billing rewards you when traffic is steady and can surprise you when it is not, Pro is per seat, and you are building on Vercel's primitives rather than raw infrastructure.

Choose Vercel if developer experience and speed to ship matter more than squeezing the lowest possible hosting bill. If you need to deploy heavy backends and databases, look at Railway or Render; if you want a framework-agnostic frontend host, compare Netlify; and if you want the most generous free tier on a massive edge network, Cloudflare Pages is the value pick.

## Frequently asked questions

How much does Vercel cost?

Vercel has three plans plus usage. Hobby is free forever for personal projects. Pro is $20 per user per month and includes $20 of monthly usage credit, faster builds, and team features, with anything above the included allotments billed as usage (for example, 10 million edge requests included then $2 per additional million, 1 TB of data transfer included then $0.15 per GB, and compute at about $0.128 per CPU-hour). Enterprise is custom-priced with SSO, an SLA, and multi-region compute. The v0 AI builder is priced separately, starting with a free tier that includes $5 of credits and a $20-per-month Premium tier.

Is Vercel worth it?

For frontend and full-stack teams that value developer experience, yes. The git-push deploys, preview URLs, and zero-config Next.js support save real engineering time, and the performance defaults are strong. It is less worth it if you run heavy backends, need full control over containers, or are optimizing purely for the lowest hosting bill, since usage-based pricing can cost more than a flat VPS at high traffic.

Does Vercel have a free plan?

Yes. The Hobby plan is free forever and is genuinely usable, with git-connected deploys, preview URLs, a global CDN, DDoS mitigation, and the web application firewall. It carries fair-use limits and a non-commercial restriction, so once you ship a real product or add teammates you will move to Pro.

What are the best Vercel alternatives?

The closest frontend host is Netlify, which offers a similar deploy-and-preview workflow across many frameworks. For deploying backends, databases, and full-stack services, Railway and Render are stronger and often cheaper. And if you want the most generous free tier on a huge edge network, Cloudflare Pages paired with Workers is the value option, especially if you already use Cloudflare.

Can Vercel host backends and databases?

Partly. Vercel runs backend logic as serverless and edge functions, and Fluid compute makes those efficient enough for API routes and AI endpoints. It does not run long-lived processes, persistent websockets at scale, or your own containers, and it is not a managed database host, though it integrates with providers and offers its own storage products. For heavy or stateful backends, Railway or Render fit better.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Vercel pricing](https://vercel.com/pricing)checked Sep 2026
- [Netlify pricing](https://netlify.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026
- [Cloudflare Pages pricing](https://pages.cloudflare.com)checked Sep 2026

Related guides

Ai App Development PlatformsAi Coding AssistantsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Windsurf Review

URL: https://devshot.co/reviews/windsurf-review
Type: review
Published: 2026-09-25
Updated: 2026-09-25
Summary: Windsurf review 2026: Devin Desktop pricing for new and grandfathered seats, what the quota replaced, and when Cursor, Copilot, Claude Code…

Review

## Windsurf Review

Worth a new Pro seat when you want the Windsurf editor, now called Devin Desktop, and a daily quota covers the agent work. A weak buy when the job is autocomplete inside the IDE you already have.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 4 alternatives covered

TL;DR

Windsurf is worth it in 2026 when you want that editor, now renamed Devin Desktop, and a new Pro seat at $20/mo covers the agent work. Existing Pro subscribers keep $15/mo. Verified on the Devin pricing page, September 2026. Pick Copilot for autocomplete in the IDE you already use, or Cursor when Tab in that editor is the habit.

A two-person team lands on an $80 monthly minimum before a third full seat, so the duo pays that floor and a third person is a new charge. The free plan still includes unlimited Tab, so autocomplete alone is not a reason to start paying, and the GitHub Copilot review is the completions comparison.

Key factsBest for: Engineers who want one Cognition plan for the IDE and a…

- Updated: September 25, 2026
- Best for: Engineers who want one Cognition plan for the IDE and a local agent manager.
- Price as of September 25, 2026: New Pro is $20/mo; grandfathered Pro keeps its rate; Free includes unlimited Tab.

- The Windsurf IDE, now Devin Desktop, billed as a daily and weekly quota you have to watch.
- Vendor: Cognition
- Product name: Devin Desktop
- New Pro: Monthly individual seat
- Alternatives covered: Cursor, GitHub Copilot, Claude Code, Codex

Pros

- New Pro adds frontier models, Devin Cloud, and extra usage at API list prices, so work can finish after the laptop closes.
- Tab and inline edits stay unlimited on Free, so autocomplete is not why you upgrade.
- Two full seats clear the Teams minimum, so a pair does not fund a third person.

Cons

- No fixed message cap is printed on the live pricing page, so a heavy week shows up on the meter.
- A flex seat excludes the desktop app, so it cannot staff this editor.
- The binary renamed to Devin, so an MDM allowlist for Windsurf alone can block the update.

VendorCognition

Product nameDevin Desktop

New ProMonthly individual seat

Best forIDE plus local agents

The buying question is no longer whether Windsurf is a second editor next to Cursor. Windsurf is Devin Desktop, and the bill is a Cognition quota rather than a pile of prompt credits you can save for later. A new individual seat is the clean entry if you do not already pay, and a grandfathered seat is the one to keep if you do.

Toolradar data: the [September 2026 AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 153 tools. [Windsurf](https://toolradar.com/tools/windsurf) and [Cursor](https://toolradar.com/tools/cursor) both have directory pages there.

The longer head-to-head on that site is [Windsurf versus Cursor](https://toolradar.com/blog/windsurf-vs-cursor-2026).

How we compared: Devin's pricing page, quota docs, and self-serve billing, the June 2, 2026 rename post, plus Cursor's pricing docs, GitHub's Copilot plans, Claude's pricing page, and OpenAI's Codex pricing, read on September 24, 2026.

The wider set on this site is the AI coding assistants list.

## What is Windsurf?

Windsurf is Cognition's desktop IDE, shipped under a new name. On June 2, 2026 the product became Devin Desktop: the same editor, extensions, keybindings, and settings, delivered as an over-the-air update, with the Agent Command Center as the default surface.

Spaces group sessions, pull requests, and files, and a Kanban view shows which local or cloud agent is still running, so the status is visible without opening every session.

Devin Local replaced Cascade as the primary local agent. Cognition's rename post says the new harness is up to 30% more token efficient and adds subagents, so the same task can draw less of the allowance when you run that harness.

That same post kept the legacy Cascade agent available through July 1, 2026, so a Cascade workflow needs a check against Devin Local before the next sprint.

You do not have to run only Cognition's agent. Devin Desktop speaks the Agent Client Protocol, and the launch post lists Codex, Claude Agent, OpenCode, and agents your team builds. Those sessions share the Kanban and a Space, which is why a shop already paying for another agent might still open this editor.

Project rules survive the rename: a root .windsurfrules file is still read, .devin/rules/ wins over .windsurf/rules/, the editor can import .cursor/rules, and AGENTS.md is read too.

The shell command changed from surf to devin-desktop, so a script on the old name will miss the app.

The JetBrains plugin is in maintenance mode, Cascade inside it is deprecated, and newer JetBrains releases can break that plugin, so Cognition points those users at Devin over ACP. The desktop binary also renamed, to Devin.app on macOS and Devin.exe on Windows, so an MDM allowlist that only names Windsurf can block the update.

## How Windsurf works

Install the desktop app, open a repo, and describe the change. Devin Local searches the tree, edits files, and can run the checks you already use, so the edits land in the repo rather than in a side chat.

Supercomplete is the Tab behavior: it tries to predict the next edit, not only the next token. Fast Context pulls the files the agent needs before a long prompt, so that prompt is not searching the tree blind.

Quotas replaced prompt credits in March 2026, and each paid plan includes a daily allowance plus a weekly allowance, except Max, which is weekly only and has no daily cap. A long day on Max does not stop when the date changes.

The daily slice is larger than one seventh of the week, so weekend work does not have to give back the rest of the weekly budget. Free models, including the SWE family Cognition names for routine tasks, do not draw the quota, while frontier requests do.

That split is why a cheap-model week can look idle and a frontier week can still empty the meter. When the meter hits zero, Free waits for the next calendar reset and cannot buy through it. Pro, Max, and Teams can buy extra usage at the model's API list price, and a fast mode (SWE-1.5 Fast, or a fast Opus variant) spends that balance faster. The usage meter in the app, and the plan page, show what is left.

On Teams, review and automations draw the shared on-demand credit pool rather than the full seat's included quota, so a busy review setting can empty the credits flex seats also use.

Anyone can open a public GitHub PR on devinreview.com, or swap github.com for devinreview.com in the URL, with no Devin account and no credit spend. That is the free lane for a public repo, and a private repo is a different bill.

After the agent opens the PR, the AI code review list is the next decision, and CodeRabbit is the specialist rather than a second seat on this plan.

## Windsurf key features

One editor, several agentsEssential
Devin Local, Devin Cloud, and ACP agents (Codex, Claude Agent, OpenCode, or an in-house agent) share the Kanban and Spaces. You can run local-only agents and skip Cognition's cloud agent. A flex seat on Teams omits the desktop app, so that free login cannot do this job.

Daily and weekly quotasEssential
March 2026 retired prompt credits, so Pro and a Teams full seat refresh on a calendar day and a calendar week, and a quiet Monday does not extend Friday. Max refreshes weekly with no daily cap, which is the plan when a daily reset would cut the work off.

Unlimited Tab on every planEssential
Free includes unlimited Tab completions and inline edits, plus a light agent quota and a shorter model list. Paid plans keep that Tab behavior and add frontier models. Tab does not meter, so an autocomplete-only day is a weak reason to pay.

Cloud handoff
Pro adds Devin Cloud, so a task can keep running after the laptop closes. Free and Pro cap concurrent sessions at 10, while Max and Teams lift that cap, and a flex seat has no cloud access.

Rules that survived the rename
Legacy Windsurf rules still load, .devin/rules takes precedence, and .cursor/rules can be imported, so day one does not start with a rewrite. Enterprise admins can also drop rules, workflows, and skills in a system directory.

SWE-2 Free, with a date on it
The pricing page includes SWE-2 Free in Devin Desktop and the CLI through October 10, 2026, while the homepage advertises unlimited SWE-2. Budget against the dated line, not the banner, because the free model can end while the ad is still up.

## Windsurf pricing

Windsurf has no price of its own anymore: the Devin pricing page is the Windsurf price, and the old product pricing URL only sends you there. New Pro is a single user and cannot be shared, so two engineers cannot pass one login around.

It adds frontier models from OpenAI, Claude, Gemini, and SpaceXAI, full model choice, Devin Cloud, and extra usage at API rates.

Max is the individual plan above it, with a larger weekly quota and no daily cap, still one person, so it does not cover a teammate.

Existing subscribers do not move to the new sticker. Quota docs grandfather an existing Pro subscriber at the old rate indefinitely, and an existing Teams subscriber at the old per-seat rate indefinitely. Enterprise contracts, including legacy Windsurf Enterprise, stay on the agreement already signed.

Leftover add-on prompt credits converted at the rate you paid, which Cognition states as $0.04 per credit, so 250 unused credits became $10 of extra usage that you can spend or have refunded.

Teams is the plan that surprises a duo, because the account minimum in the table and each full seat are separate charges. Zero full seats turns the whole minimum into prepaid credit, so the floor is paid before anyone has the editor, and one full seat leaves the remainder as prepaid credit.

Two full seats clear the minimum with nothing extra included, and three or more bill as seat count times the seat price. Flex seats are free, count toward the 200-user cap, draw only shared credits, and do not include Devin Desktop. Past 200 users, the next step is Enterprise, which publishes no list price.

SSO moved off new Teams plans, which is the gap if identity was why five engineers were about to land on the self-serve tier. A team that already bought SSO as an add-on keeps it on the grandfathered plan, and everyone else asks Enterprise for SAML or OIDC, a VPC deploy, and central admin.

Cognition's March 18, 2026 pricing note estimated Pro and Teams at 7 to 27 Premium Plus messages a day (it named Opus 4.6, GPT-5.4, and GPT-5.3-Codex), 8 to 101 on the Premium examples, and 47 to 190 on lightweight models, with a weekly cap on top.

Max was estimated at 42 to 170, 47 to 631, and 291 to 1,190 on those same bands. Those ranges can size a heavy day, and they are not a cap printed on the live pricing page.

The comparison that changes the buy is the seat next to it, not a second Windsurf tier. Cursor's docs price Pro Plus at $60/mo and tell daily agent users to start there if that editor's Tab habit is the one you will keep.

Copilot Pro is $10/mo, and paid plans leave completions outside the credit meter, the smaller bill if you stay in the IDE you have. Claude's annual Pro is $17/mo with Claude Code included, so that login can replace an editor seat, and Max is the monthly step above it.

Codex Go is $8/mo for light tasks on a ChatGPT plan. The rows below are the stickers, verified the same day.

Plan | Price | Best for |

Devin Desktop Free | No-cost | Light agent quota, limited models, unlimited Tab and inline edits. No extra-usage purchase. |

Pro, new subscribers | $20/mo | One person. Frontier models, Devin Cloud, daily and weekly quota, extra usage at API rates. |

Pro, grandfathered | $15/mo | Existing Pro subscribers keep this rate indefinitely on the new quota. |

Max | $200/mo | One person. Larger weekly quota, no daily cap. Same extras as Pro. |

Teams minimum | $80/mo | Floor for the org. Up to 200 users. Fewer than two full seats prepays the rest as credits. |

Teams full seat | $40/mo | Includes Devin Desktop and a Pro-equivalent quota. Three seats bill $120/mo. |

Teams flex seat | No-cost | No desktop app. Spends the shared credit pool only. |

Enterprise | Quote | SSO, VPC, dedicated deploy. Legacy Windsurf Enterprise keeps its contract. |

Cursor Pro | Matches new Pro | Hobby is free. Unlimited Tab on Pro, Pro Plus, and Ultra. |

Cursor Pro Plus | $60/mo | The individual plan Cursor recommends for daily agent users. |

Cursor Ultra | Matches Max | Top individual plan. On-demand usage continues at API rates. |

Cursor Teams | $40/user/mo | Standard seat. Premium is $120/user/mo for 5x agent limits, and SSO is included. |

GitHub Copilot Free | No-cost | 2,000 completions a month and 50 chat requests. No card required. |

GitHub Copilot Pro | $10/mo | USD. Unlimited completions. Monthly credits cover chat and agent, not Tab. |

GitHub Copilot Pro+ | $39/mo | USD. Premium models including Opus, and a larger monthly credit total. |

GitHub Copilot Max | $100/mo | USD. Priority model access and the largest individual credit pool. |

Claude Pro, monthly | Matches new Pro | Includes Claude Code. Free leaves Claude Code off. |

Claude Pro, annual | $17/mo | Stated monthly equivalent of $200 billed once. |

Claude Max | From $100/mo | 5x or 20x Pro per five-hour session. Monthly only. |

Codex Go | $8/mo | Light coding tasks on a ChatGPT plan. |

Codex on Plus | Matches new Pro | CLI, IDE extension, web, and iOS. Cloud review and Slack included. |

Codex Pro | From $100/mo | 5x or 20x the Plus allowance. API-key use drops cloud features. |

## Windsurf pros and cons

### What we like

- New Pro adds frontier models, Devin Cloud, and extra usage at API list prices, so work can finish after the laptop closes.
- Tab and inline edits stay unlimited on Free, so autocomplete is not why you upgrade.
- Two full seats clear the Teams minimum, so a pair does not fund a third person.

### What could be better

- No fixed message cap is printed on the live pricing page, so a heavy week shows up on the meter.
- A flex seat excludes the desktop app, so it cannot staff this editor.
- The binary renamed to Devin, so an MDM allowlist for Windsurf alone can block the update.

## Who Windsurf is for

Windsurf fits a developer who wants the editor and the agent manager in one window, and who will watch a quota instead of a credit counter. Indie hackers get the cleanest new bill on Pro, because one login covers the desktop, the local agent, and Devin Cloud.

Someone already on the old Pro rate should stay there: the grandfathered price is the discount, and cancelling to rebuy would lose it.

Engineering leads have an endpoint task before the next update: allow the Devin binary, and keep ~/.codeium in the policy, because user settings still live there while the app name changes. If the allowlist still says Windsurf, the team never receives the update.

It is the wrong default when the day is Tab-to-accept in VS Code, JetBrains, or Neovim. Paid Copilot leaves those completions off its credit meter inside the IDE already installed, and that is the smaller change if this editor is not where the day happens.

The Cursor review is the editor comparison if Tab inside a dedicated fork is the habit you will not drop. The Cursor versus Copilot note covers that pair without this rename in the middle.

Leave personal Max on the shelf when the company will pay for a full seat, and skip a flex seat if the person needs the desktop app, because that login never includes the IDE.

If the failure is a broken test rather than a missing feature, the debugging tools list is a narrower buy. Devshot's daily brief is where a cutoff like the SWE-2 date belongs. Subscribe free if you want that note in the inbox.

## Best Windsurf alternatives

If Windsurf is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Windsurf | Engineers who want one Cognition plan for the IDE and a local agent manager. | New Pro is $20/mo | Visit → |

Cursor | Developers whose Tab habit lives in Cursor's editor and who will not move it. | Hobby is free | Visit → |

GitHub Copilot | Engineers who want completions in the IDE they already use, at a lower seat cost. | Pro is $10/mo in USD | Visit → |

Claude Code | Engineers who want a repo agent in the terminal or the IDE they already open. | Included on paid Claude plans | Visit → |

Codex | Developers who already pay for ChatGPT and want the CLI and IDE agent on that bill. | Go is $8/mo | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

Cursor

An editor with its own model pool, a free Hobby plan, and a higher tier aimed at daily agents.

Visit →

GitHub Copilot

GitHub's assistant, with paid completions kept off the credit meter inside your current IDE.

Visit →

Claude Code

Anthropic's coding agent on the Claude plan you already pay for, and missing on Free.

Visit →

Codex

OpenAI's coding agent on a ChatGPT plan, with a low tier when the tasks stay light.

Visit →

## The bottom line

Pay the new Pro rate if Devin Desktop is the editor you will actually live in, and keep a grandfathered Pro seat if you already have one. Move to Max when the daily cap is what stops the workday. Put a company on Teams only after you count full seats against the minimum, because fewer than two full seats still prepays that floor as credits.

Choose Cursor when Tab inside that editor is the habit you will not move, and Pro Plus is the daily-agent tier its own docs name.

Choose GitHub Copilot when the IDE you already use should keep completions off a meter. Choose Claude Code when the agent should follow the Claude login you already pay for, including the setup walkthrough.

Choose [Codex](https://toolradar.com/tools/openai-codex) when the ChatGPT plan should be the agent, and check the October 14 retirement before you standardize on GPT-5.5.

The directory records are [Windsurf on Toolradar](https://toolradar.com/tools/windsurf) and [Copilot on Toolradar](https://toolradar.com/tools/github-copilot). Prompt-to-app tools are a different purchase, covered in vibe coding tools.

The chat-model version of the Claude versus OpenAI choice is ChatGPT versus Claude for coding. Subscribe free if you want the next quota change in the Devshot brief.

Cite this: Devshot, "Windsurf Review 2026", September 2026.

## Frequently asked questions

Is Windsurf worth it in 2026?

Yes, when you want the Windsurf editor, now called Devin Desktop, and a quota is an acceptable way to pay for agent work. New Pro is the individual entry and includes frontier models plus Devin Cloud, and existing Pro subscribers keep $15/mo. It is the wrong buy when you mainly want unlimited autocomplete in VS Code or JetBrains, where paid Copilot keeps completions off the meter. It is also the wrong buy when the only seat on offer is a flex login that excludes the desktop app.

How much does Windsurf cost?

New Pro is $20/mo for one person, and Max is $200/mo with a weekly quota and no daily cap, still for one person. Teams starts at $80/mo, then $40/mo per full seat, with flex seats free and without the desktop app. Two full seats cost the minimum, three cost $120/mo, and Enterprise publishes no list price. Existing Pro stays at $15/mo and existing Teams seats stay at $30/mo, both indefinitely. Prices were read on the Devin pricing page and quota docs on September 24, 2026.

Is there a free Windsurf plan?

Yes, Free includes a light agent quota, a limited model list, unlimited inline edits, and unlimited Tab, enough to learn the editor and not a week of frontier models. There is no way to buy extra usage, so an empty meter waits for the daily or weekly reset, and free models do not consume the quota. Devin Cloud and full model choice start on Pro. Claude's free plan is a different product and does not include Claude Code.

Did Windsurf shut down?

No: on June 2, 2026 Cognition renamed it Devin Desktop and shipped the change as an update, so plans, extensions, and in-progress work carry over. The IDE remains, with the Agent Command Center in front, Cascade's replacement is Devin Local, and the June 2 post kept the old Cascade agent available through July 1, 2026. The JetBrains plugin is the piece in maintenance mode, not the desktop IDE, so a plugin failure is not a shutdown of the product.

How does Windsurf compare with Cursor?

Both are AI editors with unlimited Tab on the plans that include it, and the editor you keep open matters more than a matching monthly sticker. Cursor's docs put Pro Plus at $60/mo for daily agent users, with Ultra at the same sticker as Windsurf Max. Cursor Teams includes SSO at $40/user/mo while new Windsurf Teams does not, and Cursor also charges Teams $0.25 per million tokens on third-party models, so the team bill can move after the seat price.

What happens when Windsurf hits its limit?

Free waits for the next daily or weekly reset, while Pro, Teams, and Max can buy extra usage at the model's API list price, and fast modes cost more per token. Max has no daily cap, only a weekly one, whereas Pro and a Teams full seat have both, so Pro can stop at midday with weekly allowance still left. Cognition does not print a fixed message allowance on the live pricing page, and its March 18, 2026 note estimated 7 to 27 Premium Plus messages a day on Pro, which is a planning range, not a contract.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Windsurf pricing](https://codeium.com/windsurf/pricing)checked Sep 2026
- [Cursor pricing](https://cursor.com/pricing)checked Sep 2026
- [GitHub Copilot pricing](https://github.com/features/copilot/plans)checked Sep 2026
- [Claude Code pricing](https://code.claude.com/pricing)checked Sep 2026

Related guides

Cursor ReviewGithub Copilot ReviewClaude Code ReviewAi Coding AssistantsAi For Code ReviewVibe Coding ToolsAi For DebuggingCoderabbit ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Residential Proxy Providers in 2026

URL: https://devshot.co/reviews/best-residential-proxies
Type: review
Published: 2026-09-24
Updated: 2026-09-24
Summary: Residential proxy providers ranked for 2026 on the rate you still pay after the coupon, and on what a US exit costs.

Expert Guide

## The Best Residential Proxy Providers in 2026

Bright Data's 141 GB coupon card is $499/mo for the discounted window. SOAX Builder adds a plan fee before the US per-GB rate.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 7 tools compared

TL;DR

The best residential proxy provider in 2026 is Bright Data when the target blocks datacenter IPs and you will pass KYC. Pay as you go lists $8/GB, and code RESIGB50 cuts that to $4/GB for 3 months. [Oxylabs](https://toolradar.com/tools/oxylabs) Starter is $6/GB ($30/mo for 5 GB), and [Decodo](https://toolradar.com/tools/decodo-formerly-smartproxy)'s 3 GB plan is $3.75/GB ($11.25/mo).

Skip Bright Data if a compliance review would stall the crawl. IPRoyal's 1 GB subscription is $7.00/GB, a small-test rate, and the $1.75/GB floor starts at 10 TB, so a normal month never reaches it. SOAX Builder is the plan when US traffic sits on the expensive tier, with prices checked on vendor pages, September 2026.

Key facts7 tools compared: Bright Data, Oxylabs, Decodo, SOAX, IPRoyal, NodeMaven, Rayobyte

- Updated: September 24, 2026
- Top pick: Bright Data (best for: Production scrapes of sites that block datacenter IPs, if you can pass KYC)
- Top pick price as of September 24, 2026: Bright Data: List $8/GB pay as you go; coupon RESIGB50 halves it for 3 months.

- 7 tools compared: Bright Data, Oxylabs, Decodo, SOAX, IPRoyal, NodeMaven, Rayobyte
- Oxylabs (best for: Teams that want a self-serve enterprise pool and can accept blocked targets): From $6/GB on Starter (5 GB, $30/mo); VAT may apply.
- Decodo (best for: Developers who want a small USD monthly total and can add tax after): From $3.75/GB (3 GB, $11.25/mo, ex-VAT); pay as you go costs more per GB.
- SOAX (best for: Teams that need a published US rate, not a single global average): Sandbox is free; Builder is $200/mo, and US traffic is tier 1 at $3.00/GB.

Buy Bright Data when the site rejects a datacenter address and you can wait for a compliance review.

The coupon lasts 3 months, and the invoice after that is the higher card, so a yearly forecast on the discount alone comes up short.

Oxylabs is the self-serve pool with a buy button, if you can accept the categories it will not serve.

Decodo is for a USD total before a sales call, and the wrong page if tax must already be inside it.

IPRoyal fits a paused scrape because unused traffic does not expire, and mid-size months have no printed price.

SOAX fits a US exit that needs that country's rate, because the cheap column is other countries.

Toolradar data: the [directory homepage](https://toolradar.com/) lists 10,680+ tools across 401+ categories, checked 24 September 2026.

Residential bandwidth is not its own shelf there, so open [Bright Data](https://toolradar.com/tools/bright-data), [Oxylabs](https://toolradar.com/tools/oxylabs), [Decodo](https://toolradar.com/tools/decodo-formerly-smartproxy), and [NodeMaven](https://toolradar.com/tools/nodemaven), then use the ladders below for the gigabyte rate.

How we ranked: seven providers, each read on the vendor's own pricing page on 24 September 2026.

The order follows the published dollar rate for a production scrape, what a gigabyte includes, and whether the cheap line needs a coupon, a 10 TB commit, or a sales call.

A managed fetch is a different bill from these per-GB pools.

That job is the web scraping tools guide and the Firecrawl review.

## Top Picks

Based on features, real-world fit, and value for money.

Best Residential Proxy Providers in 2026: 7 tools compared, updated Sep 2026

Tool | Pricing | Best for |

[Bright Data](https://toolradar.com/tools/bright-data) | List $8/GB pay as you go; coupon RESIGB50 halves it for 3 months. | Production scrapes of sites that block datacenter IPs, if you can pass KYC |

[Oxylabs](https://toolradar.com/tools/oxylabs) | From $6/GB on Starter (5 GB, $30/mo); VAT may apply. | Teams that want a self-serve enterprise pool and can accept blocked targets |

[Decodo](https://toolradar.com/tools/decodo-formerly-smartproxy) | From $3.75/GB (3 GB, $11.25/mo, ex-VAT); pay as you go costs more per GB. | Developers who want a small USD monthly total and can add tax after |

SOAX | Sandbox is free; Builder is $200/mo, and US traffic is tier 1 at $3.00/GB. | Teams that need a published US rate, not a single global average |

IPRoyal | Subscription $7.00/GB at 1 GB (5% off); pay as you go is $7.35/GB. | Intermittent jobs where unused gigabytes should not expire |

NodeMaven | From $4.25/GB monthly at 2 GB; pay as you go at that size is $5.00/GB. | Buyers who want every residential rung listed in USD |

Rayobyte | From $3.50/GB on Starter (1 to 49 GB); enterprise as low as $0.50/GB. | High volume past the Starter band, once you price the band and not the floor |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 4 of 7 do not publish a comparable monthly price and are left out rather than estimated.

1

### Bright Data

Top Pick

Best for: Production scrapes of sites that block datacenter IPs, if you can pass KYC

PricingList $8/GB pay as you go; coupon RESIGB50 halves it for 3 months.

+400M+ IPs in 195 countries, with country, state, city, ZIP, and ASN targeting included, so narrow geos are covered.

+A zone cap in bytes or dollars, plus trial traffic shown in the dashboard but not billed, stops a runaway job.

+Pay as you go has no monthly commitment, which suits a spiky scrape, and anything above 1 TB is a custom quote.

−The coupon lasts 3 months, so forecast the list card or later invoices come up short.

−Residential and mobile use waits on a KYC review, sometimes a short video call, so the pool stays closed.

Visit Bright Data →

2

### Oxylabs

Best for: Teams that want a self-serve enterprise pool and can accept blocked targets

PricingFrom $6/GB on Starter (5 GB, $30/mo); VAT may apply.

+Starter, Basic, Advanced, and Corporate are buy-now plans with geo targeting included, so you skip the sales wait.

+Sticky sessions, HTTP, HTTP/3, and SOCKS5, with a stated 99.95% success rate to test on your own targets.

+Corporate adds IPv4 or IPv6 selection, OS settings, and a dedicated account manager, which the cheaper cards lack.

−Starter caps you at 3 proxy users and top-ups stop at 100 GB, so a larger team has to move up a card.

−Apple, banking, streaming, ticketing, government, and some Google domains are restricted by policy, so confirm your use case is allowed before you buy.

Visit Oxylabs →

3

### Decodo

Best for: Developers who want a small USD monthly total and can add tax after

PricingFrom $3.75/GB (3 GB, $11.25/mo, ex-VAT); pay as you go costs more per GB.

+115M+ IPs in 195+ locations, including 7.1M+ in the United States, with ZIP and ASN targeting for broad US work.

+The pricing FAQ gives the monthly USD total, not only the per-GB line, so finance can approve without a call.

+14-day money-back on eligible subscriptions, and ISO/IEC 27001:2022 on proxies and the Scraping API, which helps a security review.

−Every total excludes VAT, so an EU address adds tax on top of the USD figure.

−The cheapest per-GB line is the 1 TB plan at a $2,000 month, so that headline is not a small-team rate.

Visit Decodo →

4

### SOAX

Best for: Teams that need a published US rate, not a single global average

PricingSandbox is free; Builder is $200/mo, and US traffic is tier 1 at $3.00/GB.

+Payments are in USD unless another currency is shown, and prices are public, so you can forecast without a sales deck.

+Builder, Team, Scale, and Enterprise include 200, 500, 1,500, and 3,000 credits, so the plan is a credit bundle.

+Country, region, city, and ISP filters start below Scale, while ASN and ZIP targeting start on Scale, so fine targeting costs more.

−The United States is tier 1, so a forecast that uses the tier 3 column will miss the US bill.

−Credits on Sandbox or a monthly plan expire after 60 days, even if you change plans, so a pause burns the balance.

Visit SOAX →

5

### IPRoyal

Best for: Intermittent jobs where unused gigabytes should not expire

PricingSubscription $7.00/GB at 1 GB (5% off); pay as you go is $7.35/GB.

+Purchased traffic does not expire and stays until you use it, which is why a sporadic scrape belongs here.

+The subscription toggle is 5% off and pay as you go is higher, so leaving the toggle off costs more.

+The 10 GB card cites 64M+ IPs in 195+ countries, the pool you accept in exchange for traffic that waits.

−Public rungs are 1 GB, 2 GB, 10 GB, and a 10 TB custom floor, so a mid-volume month has no printed price.

−A 1 GB pay-as-you-go test sits at the expensive end, so it is a poor way to sample the network.

Visit IPRoyal →

6

### NodeMaven

Best for: Buyers who want every residential rung listed in USD

PricingFrom $4.25/GB monthly at 2 GB; pay as you go at that size is $5.00/GB.

+Nine monthly rungs and nine pay-as-you-go rungs are priced in USD, so there is no second currency to convert.

+Monthly beats pay as you go from the 2 GB rung through 1,000 GB, so a steady job should leave pay as you go.

+Sticky sessions plus country and city targeting come with the residential product, not as a later add-on.

−No free residential rung is listed, so a test starts on a paid card and there is no standing free pool.

−The 2 GB monthly rate is high next to the 100 GB commitment, so the tiny plan is only a sample.

Visit NodeMaven →

7

### Rayobyte

Best for: High volume past the Starter band, once you price the band and not the floor

PricingFrom $3.50/GB on Starter (1 to 49 GB); enterprise as low as $0.50/GB.

+Starter, Consumer, Professional, and Corporate are priced before the contact tier, so most volumes skip a sales call.

+40M+ IPs, sticky sessions, and city, state, and country targeting are included, a smaller map that still covers broad countries.

+Unlimited threads are listed with the prices, so a parallel scraper does not add a second fee.

−The enterprise floor is a contact tier at 5,000 GB or more, so the lowest rate is not what a pilot pays.

−You stay on the Starter rate from 1 GB through 49 GB, so a small test and a 49 GB month share that price.

Visit Rayobyte →

## What it is

A residential proxy sends your request out through an IP an ISP assigned to a household connection, so the site sees a consumer address instead of a server.

Datacenter proxies cost less and get blocked more often, so they are the wrong buy on defended targets.

ISP proxies are sold per IP, not per GB, so that invoice is different.

The bill is almost always bandwidth in both directions, so headers and the body both count.

Bright Data includes request headers, POST bodies, response headers, and response data, so a retry inflates the gigabyte.

SOAX counts partial responses, retries, and interrupted connections, and defines a GB as one billion bytes.

Sticky sessions keep one IP for a login flow, and rotating sessions change the IP per request for a public crawl. A rotating exit breaks a login that expects the same address.

Country, city, and ZIP targeting is included on the leaders here, so treat it as already in the price.

## Why it matters

A 100 GB month is a different company on each ladder. On Decodo that volume is the $275 plan, a card you can hand to finance.

On Oxylabs it sits between the $100 plan and the $500 plan, so you buy the larger card or you top up.

On IPRoyal the public cards stop at 10 GB and resume at 10 TB, so a 100 GB job has no printed rung.

Devshot data: 28,000 developer readers get this brief, and the audience file updated 20 September 2026 puts the open rate at 30%.

They compare a gigabyte the way they compare a cloud bill, so a coupon with no end date is the month-four mistake. Later sticker changes land in the Devshot brief for the next rate check.

Pair the pool with something that tells you the job failed, because a silent retry is a missed row and extra bandwidth.

The monitoring tools guide and the Datadog review cover the log invoice, which grows when a scraper retries.

A datacenter VPS, covered in the DigitalOcean review, is not a substitute, because a site that blocks servers will block that VPS too.

Geo checks on your own API belong with API testing tools, and the gateway in front is a separate buy in the API gateways guide.

## Key features to look for

The rate after the discountEssential
Coupons, subscription toggles, and 10 TB floors advertise a rate a 20 GB month will never pay, so read the rung your volume lands on.

What a gigabyte includesEssential
Headers, retries, and the response count at Bright Data and at SOAX. A vendor that bills successful results only is a scraping API, which is a different purchase.

Whether US exits are the cheap tierEssential
SOAX puts the United States and Germany on tier 1, so tier 3 is another country. Forecasting a US job from that column will miss the bill.

Traffic that expires
IPRoyal keeps purchased traffic until you use it, which suits a scrape that pauses. Oxylabs top-ups end with that monthly period, and SOAX credits expire after 60 days, so a pause burns the balance.

KYC and blocked targets
Bright Data reviews residential use before the pool opens, so a team that cannot wait should start elsewhere. Oxylabs restricts banking, streaming, ticketing, and some Google domains under its acceptable-use policy, so read that list against your use case before you pay.

## Pricing

Budget the rate that survives month four, because the coupon will not defend the later invoice.

Bright Data bills in USD, RESIGB50 is 50% off for 3 months, and the coupon bills are $499/mo, $999/mo, and $1,999/mo, with anything above 1 TB on a quote.

Decodo's FAQ, excluding VAT, prices 100 GB at $275/mo, 250 GB at $625/mo, and 1 TB at $2,000/mo, which is the cost of a real month.

Oxylabs prints monthly totals of $100, $500, and $2,500 beside Starter, takes wires in dollars or euros, and says VAT may apply.

A US exit on SOAX is tier 1, so the Enterprise line from $0.85/GB is not the Sandbox column.

IPRoyal's floors start at 10 TB, so that headline does not price a pilot.

[NodeMaven](https://dupple.com/reviews/nodemaven)'s 1,000 GB monthly rung is $2.20/GB against $2.45/GB on pay as you go, a gap that matters only at that volume.

Rayobyte's Consumer band is $2/GB, verified on each vendor's pricing page, September 2026.

Plan | Price | Best for |

Bright Data residential, pay as you go, list | $8/GB | No monthly commitment; USD is the binding charge |

Bright Data residential, pay as you go, RESIGB50 | $4/GB | 50% off the list rate for 3 months |

Bright Data residential, 141 GB included, list | $7/GB | Coupon card shows half that rate |

Bright Data residential, 141 GB, coupon bill | $499/mo | Billed monthly next to the halved rate |

Bright Data residential, 332 GB included, list | $6/GB | Coupon card shows half that rate |

Bright Data residential, 332 GB, coupon bill | $999/mo | Billed monthly |

Bright Data residential, 798 GB included, list | $5/GB | Coupon card shows half that rate |

Bright Data residential, 798 GB, coupon bill | $1,999/mo | Billed monthly; above 1 TB is contact |

Oxylabs Residential Starter | $6/GB | 5 GB, $30/mo; top up to 100 GB; VAT may apply |

Oxylabs Residential Basic | $5/GB | 20 GB, $100/mo; top up to 100 GB |

Oxylabs Residential Advanced | $4/GB | 125 GB, $500/mo; top up to 2 TB |

Oxylabs Residential Corporate | $2.50/GB | 1 TB, $2,500/mo; IPv4 or IPv6 selection included |

Decodo residential, 3 GB | $3.75/GB | $11.25/mo, excludes VAT |

Decodo residential, 10 GB | $3.50/GB | $35/mo, excludes VAT |

Decodo residential, 25 GB | $3.25/GB | $81.25/mo, excludes VAT |

Decodo residential, 50 GB | $150/mo | 50 GB plan, excludes VAT |

Decodo residential, 100 GB | $2.75/GB | $275/mo, excludes VAT |

Decodo residential, pay as you go | $4/GB | No monthly commitment, excludes VAT |

Decodo residential enterprise, 250 GB | $2.50/GB | $625/mo, excludes VAT |

Decodo residential enterprise, 1 TB | $2/GB | $2,000/mo, excludes VAT |

IPRoyal residential subscription, 1 GB | $7.00/GB | 5% off the pay-as-you-go card |

IPRoyal residential subscription, 2 GB | $5.95/GB | Subscription toggle |

IPRoyal residential subscription, 10 GB | $5.25/GB | Card cites 64M+ IPs in 195+ countries |

IPRoyal residential subscription, from 10 TB | $1.75/GB | Custom floor, not a small-job rate |

IPRoyal residential pay as you go, 1 GB | $7.35/GB | No monthly minimum |

IPRoyal residential pay as you go, 2 GB | $6.25/GB | Marked most popular on the pay-as-you-go set |

IPRoyal residential pay as you go, 10 GB | $5.51/GB | Same 64M+ IP claim as the subscription card |

IPRoyal residential pay as you go, from 10 TB | $1.84/GB | Custom floor without the 5% toggle |

SOAX Sandbox | Free | Tier 1 (US and Germany) $5.00/GB; prices exclude tax |

SOAX Builder | $200/mo | 200 credits; tier 1 $3.00/GB; tier 2 $2.25/GB; tier 3 $1.20/GB |

SOAX Team | $500/mo | 500 credits; tier 1 $2.20/GB; tier 2 $1.65/GB; tier 3 $0.90/GB |

SOAX Scale | $1,500/mo | 1,500 credits; tier 1 $1.50/GB; tier 2 $1.15/GB; tier 3 $0.60/GB |

SOAX Enterprise | $3,000/mo | 3,000 credits; tier 1 from $0.85/GB; tier 2 from $0.55/GB; tier 3 from $0.35/GB |

SOAX Sandbox tier 2 | $4.00/GB | Not the US tier |

SOAX Sandbox tier 3 | $2.00/GB | Not the US tier |

NodeMaven residential monthly, 2 GB | $4.25/GB | PAYG $5.00/GB |

NodeMaven residential monthly, 8 GB | $4.00/GB | PAYG $4.88/GB |

NodeMaven residential monthly, 20 GB | $3.75/GB | PAYG $4.50/GB |

NodeMaven residential monthly, 40 GB | $3.45/GB | PAYG $4.10/GB |

NodeMaven residential monthly, 55 GB | $3.18/GB | PAYG $3.73/GB |

NodeMaven residential monthly, 100 GB | $3.00/GB | PAYG $3.45/GB |

NodeMaven residential monthly, 200 GB | $2.85/GB | PAYG $3.25/GB |

NodeMaven residential monthly, 450 GB | $2.60/GB | PAYG $2.92/GB |

NodeMaven residential monthly, 1,000 GB | $2.20/GB | PAYG $2.45/GB |

NodeMaven ISP, 30-day plan | $2.99 | Separate from the residential per-GB ladder |

NodeMaven ISP, 90-day plan | $7.79 | Separate from the residential per-GB ladder |

Rayobyte residential Starter | $3.50/GB | 1 to 49 GB |

Rayobyte residential Consumer | $2/GB | 50 to 249 GB |

Rayobyte residential Professional | $1.50/GB | 250 to 999 GB |

Rayobyte residential Corporate | $0.70/GB | 1,000 to 4,999 GB |

Rayobyte residential Enterprise | $0.50/GB | As low as this, at 5,000 GB or more; contact |

Mistakes to avoid

×Budgeting Bright Data at the coupon for a year, when RESIGB50 is printed as 3 months and the list card is what you pay after that window.

×Quoting SOAX tier 3 for a US crawl misses the bill. The United States and Germany sit on tier 1, and the bands reset each billing period.

×Treating a 10 TB floor as a pilot price. IPRoyal's $1.75/GB and $1.84/GB lines start at 10,000 GB, and a 1 GB pay-as-you-go test is $7.35/GB, a sample rather than scale.

Expert tips

→Put a zone cap on Bright Data before the first large job. The dashboard can limit a zone by bytes or by dollars, on a roughly 15-minute check, so the job can overshoot by that window.

→Oxylabs publishes a restricted list (streaming, ticketing, banking, Apple, some Google domains); check any provider's acceptable-use policy against your project before you buy a TB. The block is on the residential product page, so it will not show up as a credit at month end.

→For a scraper you do not want to operate, skip the per-GB pool and read the <a href="/reviews/bright-data-review">Bright Data review</a> next to the <a href="/reviews/best-error-tracking-tools">error tracking tools guide</a>. Failed requests are an engineering problem and a bandwidth problem at the same time.

## The bottom line

Bright Data is the buy when the target is defended and KYC is acceptable. Keep the list rate in the forecast and treat RESIGB50 as a 3-month cut, not the permanent card.

Move to Oxylabs for four self-serve cards if you can live with the blocked-target list.

Move to Decodo for a small USD plan with tax called out as extra, and skip it when finance needs tax inside the number.

Move to SOAX when the US rate must be explicit, because a blended average will miss the invoice.

Move to IPRoyal when gigabytes must survive a pause, and skip it when you need a printed mid-volume rung. Move to NodeMaven to compare the monthly rung and the pay-as-you-go rung in USD.

Move to Rayobyte when volume is past the Starter band and the contact tier is a real conversation.

NetNut is off this list, because netnut.io displays an FBI seizure notice and there is no pricing page to buy from. Cite this: Devshot, "Best Residential Proxy Providers in 2026", September 2026.

## Frequently asked questions

What is the best residential proxy provider in 2026?

Bright Data is the buy if the site blocks datacenter IPs and you will complete KYC. Pay as you go lists $8/GB, and RESIGB50 cuts that rate for 3 months, checked on Bright Data's residential pricing page in September 2026, so the list figure survives the coupon. Oxylabs is the self-serve Starter pick when your targets are not restricted. Decodo is the smaller USD entry card when you want a monthly total and can add tax later.

How much do residential proxies cost?

On the pages checked in September 2026, a small month is Decodo's 3 GB plan at $11.25, Oxylabs' 5 GB Starter, or SOAX Builder at $200/mo. Bright Data's list pay-as-you-go card is the forecast after the coupon ends, and that is the yearly number. Rayobyte's enterprise contact tier and IPRoyal's $1.75/GB floor sit on 5,000 GB and 10 TB commits, so they do not price a pilot.

Is there a free residential proxy?

SOAX Sandbox is a free account, but the United States tier 1 rate is still a paid per-GB charge, so the free label is the account, not the traffic. Bright Data and Decodo offer a trial or a money-back window, not an ongoing free pool. NodeMaven lists no free residential rung, and a free proxy list is a different product, not one of these seven networks.

Bright Data vs Oxylabs: which should a developer buy?

Bright Data when you need the larger published pool (400M+ IPs) and you can pass KYC, which fits a defended target if the review will not stall you. Oxylabs when you want buy-now cards at $100, $500, and $2,500 per month beside Starter and your targets are not restricted. Oxylabs states an average 99.95% success rate and a 175M+ IP pool on the residential pricing page. VAT may apply there, while Bright Data bills the binding amount in USD.

Why is my per-GB bill higher than the page size?

Bright Data adds request headers, the POST body, response headers, and the response, so the meter is the whole exchange, not the HTML alone. SOAX adds partial responses, retries, and interrupted connections, and calls one GB one billion bytes, so a failed attempt still counts. A retry multiplies that sum, and Bright Data checks a zone cap about every 15 minutes, so an uncapped loop becomes the bill.

Which residential proxy is cheapest for US traffic?

There is no single cheapest US rate, because these vendors do not publish one comparable US-only card. SOAX is the only one here that prices the United States as its own tier: tier 1 is $5.00/GB on Sandbox, $3.00/GB on Builder, $2.20/GB on Team, $1.50/GB on Scale, and from $0.85/GB on Enterprise. Decodo lists 7.1M+ US IPs on the global plans, from the 3 GB card, with no US surcharge, so US traffic pays the global rung.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Bright Data pricing](https://brightdata.com/pricing)checked Sep 2026
- [Oxylabs pricing](https://oxylabs.io/pricing)checked Sep 2026
- [Decodo pricing](https://decodo.com)checked Sep 2026

Related guides

Bright Data ReviewWeb Scraping ToolsFirecrawl ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Netlify Review

URL: https://devshot.co/reviews/netlify-review
Type: review
Published: 2026-09-24
Updated: 2026-09-24
Summary: Netlify review 2026: credit billing fits a frontend that does not want a fee per developer, while Legacy Pro still bills every Git contributor.

Review

## Netlify Review

Worth it when the account bills in credits and the app is a frontend, because headcount does not multiply the subscription. A weak default if you are still on Legacy Pro and every Git push counts as a seat.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 alternatives covered

TL;DR

Netlify is worth it in 2026 when the app is a frontend and the account bills in credits. Credit Pro starts at $20/mo with unlimited members and 3,000 credits, verified on Netlify's pricing page in September 2026. Free is a 300-credit hard stop, and a zero balance pauses every project on the team.

Personal is $9/mo for 1,000 credits, the solo tier if the account will not be shared. Accounts opened before September 4, 2025 can still sit on Legacy Pro at $19 per member, Git contributors included, so a contractor's push raises the bill.

Price [Vercel](https://toolradar.com/tools/vercel), Cloudflare Pages, [Render](https://toolradar.com/tools/render), Firebase Hosting, and Railway before you treat the base fee as the whole invoice.

Key factsBest for: Front-end teams on credit billing who do not want a fee per…

- Updated: September 24, 2026
- Best for: Front-end teams on credit billing who do not want a fee per developer.
- Price as of September 24, 2026: Credit Pro from $20/mo (unlimited members); Personal $9/mo.

- A CDN, previews, and functions billed in credits, with unlimited seats only on credit Pro.
- Founded: 2014
- Headquarters: San Francisco
- New accounts: Credits since Sep 4, 2025
- Alternatives covered: Vercel, Cloudflare Pages, Render, Firebase Hosting, Railway

Pros

- Credit Pro includes unlimited members, so a twenty-person team does not multiply the subscription the way Legacy Pro does.
- Deploy Previews and branch deploys cost zero credits, so only a successful production publish spends 15.
- Forms are free and unlimited on credit plans, including Free, so a form is not a reason to upgrade.

Cons

- A zero credit balance pauses every project on the team, and Free cannot buy a pack to stay up.
- Legacy Pro still charges per member, Git contributors included, and the move to credits is irreversible.
- Free and Personal share one concurrent build, and an extra build is a $40 add-on.

Founded2014

HeadquartersSan Francisco

New accountsCredits since Sep 4, 2025

Best forFront ends on credit Pro

Netlify still fits the job it invented, a Git-connected frontend on a global CDN, and in 2026 the invoice depends on which plan generation the team is stuck on. A new signup cannot buy the old member price, so credit tiers are the only offer a fresh account will see.

An older team can keep Legacy billing, and the switch to credits cannot be reversed, so model both invoices before you move.

This review is for developers, engineering leads, and indie hackers who need the real monthly cost, not a feature tour. It prices the credit tiers, the meters that spend them, and the Legacy allowances that still surprise a team that never migrated.

The fuller cross-site sheet is the [Netlify pricing page on Dupple](https://dupple.com/pricing/netlify).

Toolradar data: the [September 2026 hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) evaluated 297 tools. [Netlify](https://toolradar.com/tools/netlify) has its own page in that catalog and is not one of the 10 names at the top of the list.

How we compared: Netlify, Vercel, Cloudflare Pages, Render, Firebase Hosting, and Railway, with USD figures read on each vendor's pricing page or billing docs on September 24, 2026. No hands-on deploy was run for this piece.

If you open Netlify through a link here, Devshot may earn a commission, and that does not change the verdict.

## What is Netlify?

Netlify is a composable web platform: you connect a repo, it builds, and the CDN serves it with HTTPS on a custom domain. The same account can run serverless and edge functions, scheduled and background functions, Database, and Blob storage.

That covers light backend work, and it is the wrong host if you need to size a long-running process.

Forms ship with the platform, so a contact form does not add another vendor. Agent Runners, announced on October 1, 2025, meter model spend separately from the CPU that hosts the run, so an agent can draw credits while the page is quiet. Firewall rules and basic rate limiting are on every credit plan, including Free.

New accounts from September 4, 2025 onward are credit plans only: Free, Personal, Pro, or a custom Enterprise contract. Older Free, Starter, and Pro teams can keep Legacy plans, which is why an older write-up can still match an old dashboard.

Enterprise lists a 99.99% SLA, SSO and SCIM, log drains, and unlimited credits for teams that need a contract.

## How Netlify works

The daily loop is a Git push, with an API and an AI deploy path beside it, and the charge lands on a successful production publish rather than a pull request.

That publish spends 15 credits, while Deploy Previews, branch deploys, failed builds, and rollbacks spend zero, so review work does not drain the month.

Credits expire in date order, monthly first, then rollover, then packs that do not expire, so an older pack is the last buffer. Auto recharge stays off until a Team Owner enables it for the whole team, not for one site, so a single project cannot refill the shared balance.

When the balance hits zero, every project pauses and visitors get a Site not available page, a team-wide outage rather than a warning on one site. One concurrent build on Free means the whole team builds one at a time, so a second site waits on a launch day.

Edge functions draw web requests rather than compute, on the same meter as pages and assets, while compute covers serverless work, preview servers, Agent Runners, and database compute. A background job hits that second meter, which is the one that surprises a quiet frontend.

## Netlify key features

Credits for deploys, bandwidth, and computeEssential
A production publish spends 15 credits, while previews, branch deploys, failed builds, and rollbacks spend none, so review traffic stays off the invoice. Bandwidth is 20 credits per GB, requests are 2 credits per 10,000, and compute is 10 credits per GB-hour, which is what a busy month actually spends.

Credit Pro seats versus Legacy membersEssential
Credit Pro includes unlimited members and private organization repos, so adding a developer does not raise the base fee. Legacy Pro bills per member and counts active Git contributors, and the move to credits is permanent, so a seat saving cannot be undone if the meters cost more.

Functions, edge, and Agent Runners
Serverless, scheduled, and background functions share the compute meter, while edge functions count as web requests, so a cron and a page-view route draw different pools. Agent Runners, launched October 1, 2025, bill CPU on that compute meter and model tokens as AI inference, at 180 credits per $1 of provider usage.

Forms, Database, and Blobs
Form submissions are free and unlimited on every credit plan, while Legacy plans still start at 100 submissions per site each month, so an old contact form can still be a package. Netlify Database is included, with 3 databases on Free, 5 on Personal, and 50 on Pro, and Blob storage is on all three credit plans.

Builds and Live Preview Servers
Free includes one concurrent build for the whole team, Personal is capped at three, and Pro includes three before the same add-on with no cap. Every credit plan includes one Live Preview Server, and each extra server is its own monthly charge, so several live reviews stack fees.

Analytics window and private repos
Analytics cover the past day on Free, 24 hours on Personal, and 30 days on Pro, so a Free dashboard cannot explain last week's traffic. Shared env vars, 7-day audit logs, password protection, and private organization repos are Pro, while Free and Personal private projects stay owner-only.

## Netlify pricing

Read the plan generation before the sticker, because a credit subscription and a Legacy seat fee are different products under the same name.

Credit prices below were checked on [Netlify's pricing page](https://www.netlify.com/pricing/) and the [credit-plan docs](https://docs.netlify.com/manage/accounts-and-billing/billing/billing-for-credit-based-plans/credit-based-pricing-plans) on September 24, 2026.

Legacy rates are from the Legacy plan docs the same day, and Enterprise is a quote on Netlify, Vercel, Render, and Railway.

Free includes 300 credits and will not sell you more, so the ceiling never turns into a bill. Spend that allotment only on production deploys and you get 20 publishes, because each one is 15 credits. Spend it only on bandwidth and you get 15 GB, and a real month mixes meters, so the wall arrives sooner than either extreme.

Personal adds a 500-credit pack with auto recharge, off unless an owner enables it for the whole team. Pro's pack is 1,500 credits, and Netlify's own calculator prices a production deploy at $0.10, bandwidth at $0.13 per GB, and compute at $0.07 per GB-hour at that pack rate.

Those pack dollars sit on top of the subscription, so a tier-only budget misses a busy month.

Rollover starts at the $33 tier, 5,000 credits, and unused monthly credits last one extra cycle. The 10,000, 15,000, and 20,000 tiers roll over too, while the 3,000-credit tier does not, so the cheapest Pro forfeits what you do not spend.

A downgrade deletes leftover monthly credits and rollover credits, while packs stay, so spend the monthly balance before you step down.

An extra concurrent build is $40/mo, and each Live Preview Server past the included one is $15/mo, both on top of the tier. Twenty Legacy Pro members are twenty seat fees, at $19 per member including active Git contributors, before bandwidth, while the same twenty on credit Pro do not multiply the subscription.

Legacy bandwidth is 100 GB on Free and Starter. Free stops serving at that ceiling instead of billing, and Starter pays $55 per extra 100 GB, while Legacy Pro includes 1 TB before the same package.

Build minutes over the Legacy Pro pool are $7 per 500, and edge functions past 2 million are $2 per million. Forms past 100 per site are a paid Legacy package, while credit plans leave forms unmetered.

Vercel Pro is per user at the same monthly rate as credit Pro, with a usage credit equal to the seat fee, so each added developer still costs a seat. Hobby is free, a solo plan rather than a team plan.

Cloudflare Pages does not bill static asset requests, and Pages Functions draw Workers. The paid plan has a $5 minimum, 10 million requests included, then $0.30 per million, with no egress fee, which favors a site made of files.

Render static sites are free, and a Starter instance is 512 MB and 0.5 CPU at the flat price in the table. The Pro workspace is $25/mo with unlimited seats and 25 GB of bandwidth, then the per-GB rate in the table.

Firebase Spark needs no card, with 10 GB stored and 360 MB of transfer a day, then Blaze storage at $0.026 per GB. Cloud Functions are not on Spark, so any function forces the paid Google plan.

Railway's Hobby fee equals its usage credit, CPU and memory stop when the service stops, and egress is $0.05 per GB.

Plan | Price | Best for |

Netlify Free (credit) | No subscription, 300 credits | Hard stop. No auto recharge. Sites pause at a zero balance. |

Netlify Personal | $9/mo, 1,000 credits | One owner. Secret detection. Analytics for the past 24 hours. |

Netlify Pro, 3,000 credits | $20/mo | Unlimited members. No rollover. Private org repos and 30-day analytics. |

Netlify Pro, 5,000 credits | $33/mo | First tier that rolls unused monthly credits for one extra cycle. |

Netlify Pro, 10,000 credits | $63/mo | Rollover on. Same Pro features as the base tier. |

Netlify Pro, 15,000 credits | $95/mo | Rollover on. Downgrade deletes leftover monthly credits. |

Netlify Pro, 20,000 credits | $126/mo | Top self-serve credit tier. Enterprise above this is a quote. |

Personal credit pack | $5 for 500 credits | Auto recharge, off by default. Packs do not expire. |

Pro credit pack | $10 for 1,500 credits | Calculator rate: deploy $0.10, bandwidth $0.13/GB, compute $0.07/GB-hour. |

Extra concurrent build | $40/mo each | Personal cap is 3. Pro includes 3, then this add-on with no cap. |

Extra Live Preview Server | $15/mo each | One server included on every credit plan. |

Legacy Pro member | $19/member/mo | Pre-Sep 4, 2025 accounts. Active Git contributors count as members. |

Legacy bandwidth package | $55 per 100 GB | Starter after 100 GB. Legacy Pro after 1 TB. Free is a hard limit. |

Legacy Pro build minutes | $7 per 500 minutes | After 25,000 minutes. Starter's pool is 300 minutes. |

Legacy Pro edge functions | $2 per extra million | 2 million included. Credit plans meter edge as web requests instead. |

Vercel Pro | Per user, credit matches the fee | Hobby is free. Viewer seats included. Enterprise is a quote. |

Cloudflare Workers Paid | $5/mo minimum | 10 million requests, then $0.30 per million. No egress fee. Static requests free. |

Render Starter web | $7/mo | 512 MB, 0.5 CPU. Static sites are free. Hobby workspace has no fee. |

Render Pro workspace | $25/mo | Unlimited seats. 25 GB bandwidth, then $0.15/GB. |

Firebase Hosting Spark | No-cost, 360 MB/day | 10 GB storage. Blaze transfer after the allowance is $0.15/GB. |

Firebase Hosting storage | $0.026/GB | Blaze, after 10 GB. Functions require Blaze, not Spark. |

Railway Hobby | Fee equals included credit | Stopped services drop CPU and memory. Egress is $0.05/GB. |

Railway memory | $10 per GB-month | Metered per second. Pro seats are unlimited and the fee is per workspace. |

## Netlify pros and cons

### What we like

- Credit Pro includes unlimited members, so a twenty-person team does not multiply the subscription the way Legacy Pro does.
- Deploy Previews and branch deploys cost zero credits, so only a successful production publish spends 15.
- Forms are free and unlimited on credit plans, including Free, so a form is not a reason to upgrade.

### What could be better

- A zero credit balance pauses every project on the team, and Free cannot buy a pack to stay up.
- Legacy Pro still charges per member, Git contributors included, and the move to credits is irreversible.
- Free and Personal share one concurrent build, and an extra build is a $40 add-on.

## Who Netlify is for

Netlify fits a frontend team that wants previews, a CDN, and forms without a server, on credit Pro so headcount does not multiply the subscription. A solo project inside 1,000 credits belongs on Personal. A hobby repo that can go offline belongs on Free, but a customer URL does not.

Skip it for a long-running process, a websocket server, or a database you size yourself. Those jobs want a dyno or a VM, which is a Heroku or Railway call, or a VM on the VPS hosting list and the DigitalOcean review.

A Next.js app that wants the framework's own host should read the Vercel review first.

Stay off Legacy Pro if you are about to add contractors, because they count as members and each one adds a seat fee. Stay off credit Free when a spike would be worse than a bill, because the plan stops the site instead of charging you.

Password protection and shared env vars wait for Pro, so a locked client preview is not a Free feature.

If the product is Postgres and the frontend is only the shell, start with Postgres hosting and the Supabase review.

Pipelines past deploy previews belong on the CI/CD tools list, and the function-versus-container choice is serverless platforms.

Credit tiers moved again in July 2026, so subscribe free if you want the next change in the inbox.

## Best Netlify alternatives

If Netlify is not the right fit, these are the closest options.

Tool | Best for | Starts at | |

Netlify | Front-end teams on credit billing who do not want a fee per developer. | Credit Pro from $20/mo (unlimited members) | Visit → |

Vercel | Next.js teams that will pay per developer for previews and included usage. | Pro is per user at credit Pro's monthly rate, with a usage credit equa | Visit → |

Cloudflare Pages | Static sites that want unlimited asset requests and a low Workers minimum. | Workers Paid from $5/mo | Visit → |

Render | Teams that want a sized web service plus a static site on one account. | Starter web $7/mo (512 MB, 0.5 CPU) | Visit → |

Firebase Hosting | Static hosting on a Google project, with Blaze only if functions must run. | Spark is no-cost | Visit → |

Railway | Services that should stop paying CPU and memory when they are stopped. | Hobby fee equals its usage credit | Visit → |

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 3 of 6 do not publish a comparable monthly price and are left out rather than estimated.

Vercel

The Next.js host: a per-user Pro fee, a matching usage credit, and a free Hobby plan.

Visit →

Cloudflare Pages

Pages on Cloudflare's edge: static requests are free, and Functions bill as Workers.

Visit →

Render

A workspace fee plus a sized instance, with static sites at no charge.

Visit →

Firebase Hosting

Google's host: a no-card Spark allowance, then Blaze rates for storage and transfer.

Visit →

Railway

Per-second CPU and memory, with a plan fee that counts as usage credit.

Visit →

## The bottom line

Buy Netlify on credit Pro when the product is a frontend, previews matter, and you want one subscription for the whole team. Use Personal for a solo app inside 1,000 credits, and use Free only for a project that can sit offline between cycles.

Do not approve a budget from a 2024 post that still quotes a per-member Pro. That rate is Legacy, it includes Git contributors, and leaving it is a one-way door. If the meter, not the seat, is the risk, raise the credit tier before you add a pack on autopilot, since rollover starts above the base tier.

Choose [Vercel](https://toolradar.com/tools/vercel) when the app is Next.js and a per-user fee is acceptable, and [Cloudflare Pages](https://toolradar.com/tools/cloudflare) when static traffic should not be a line item.

Choose [Render](https://toolradar.com/tools/render) for a sized process, Firebase Hosting for a no-card static allowance, and [Railway](https://toolradar.com/tools/railway) when a stopped service should cost nothing.

Devshot's newsletter carries the next credit-tier change, so join the free list if that note should land in the inbox.

Cite this: Devshot, "Netlify Review 2026", September 2026.

## Frequently asked questions

Is Netlify worth it in 2026?

Yes, for a frontend on the credit plans, where the base Pro tier includes unlimited members and 3,000 credits, so a growing team does not pay per seat. It is a poor default for an old team still on Legacy Pro's per-member rate, because active Git contributors are seats, and for any workload that needs a long-running server. Free is a sandbox rather than a customer URL, because the balance is a hard stop, and prices were read on Netlify's pricing docs in September 2026.

How much does Netlify cost?

Credit Free has no subscription and 300 credits, with no way to buy more, so it ends when the allotment ends rather than as an overage. Personal is $9/mo for 1,000 credits, and Credit Pro starts at $20/mo for 3,000 credits and unlimited members, then $33, $63, $95, and $126 for the higher monthly tiers. Legacy Pro, for accounts opened before September 4, 2025, is $19 per member, extra bandwidth on Legacy is $55 per 100 GB, and Enterprise publishes no list price.

Does Netlify have a free plan?

Yes, Credit Free is a standing plan with 300 credits, unlimited deploy previews, custom domains, and a global CDN, and it will not auto recharge. When the credits are gone, every site on the team pauses until the next cycle. Legacy Free, if the team still has it, caps bandwidth at 100 GB and build minutes at 300, also as a hard limit, and forms on credit plans do not draw credits.

How does Netlify compare with Vercel?

Both sell a free personal tier and a paid team tier, and the split that matters is seats versus credits. Vercel Pro bills per user and includes a usage credit equal to that fee, so five developers are five fees, the invoice credit-plan Netlify does not send. Netlify's credit Pro does not multiply by members, and Deploy Previews spend no credits, so a wide team pays one subscription while reviews stay off the meter. Vercel is the closer fit for Next.js, Netlify fits when the framework varies and you would rather buy credits than seats, and both Enterprise plans are a quote.

What happens if I switch from Legacy pricing to credits?

The switch is permanent, and Netlify's docs say you cannot return to a Legacy plan after moving to credits, so run the numbers before you confirm. You leave per-member billing, Git contributors included, and land on credits that pause sites at zero unless Personal or Pro has auto recharge on. Rollover exists only on Pro tiers of 5,000 credits and up, from $33/mo, so the base tier will not keep unused credits. Check Usage and billing in the team dashboard before you click through, because that screen shows which generation you are on.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Netlify pricing](https://netlify.com/pricing)checked Sep 2026
- [Vercel pricing](https://vercel.com/pricing)checked Sep 2026
- [Cloudflare Pages pricing](https://pages.cloudflare.com)checked Sep 2026
- [Render pricing](https://render.com/pricing)checked Sep 2026
- [Railway pricing](https://railway.app/pricing)checked Sep 2026

Related guides

Vercel ReviewServerless PlatformsFirebase ReviewHeroku ReviewDigitalocean ReviewVps Hosting For DevelopersPostgres HostingCi Cd ToolsSupabase ReviewAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Kubernetes Tools in 2026

URL: https://devshot.co/reviews/best-kubernetes-tools
Type: review
Published: 2026-09-01
Updated: 2026-09-01
Summary: The Kubernetes tools a platform team actually standardises on in 2026, compared on what is free, what a seat costs…

Expert Guide

## The Best Kubernetes Tools in 2026

k9s and Helm have no price list. Argo CD and Rancher are free until you buy a control plane or a support contract. Lens is $25 a seat once the company crosses $10 million.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Sep 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

k9s and Helm are free, Apache 2.0, with no paid edition from the project. Argo CD is the same if you host it; Akuity's managed platform starts at $495 a month for 50 applications. Rancher community is free; SUSE Rancher Prime is quote-only; the only public SUSE number is Rancher for AWS at $25 per vCPU per month. Lens Personal is free under $10 million revenue or funding; Plus and Pro are $25 per user per month on annual ($300 billed up front), Enterprise is $50.

Key facts5 tools compared: Lens, k9s, Argo CD, Helm, Rancher

- Updated: September 1, 2026
- Top pick: Lens (best for: Engineers who want a desktop Kubernetes IDE, and companies over $10 million that have to pay for it anyway)
- Top pick price as of September 1, 2026: Lens: From $25/user/mo (Plus or Pro, billed yearly); free under $10 million, 30-day evaluation above

- 5 tools compared: Lens, k9s, Argo CD, Helm, Rancher
- k9s (best for: People who already live in a terminal and do not want a license conversation): Free (Apache 2.0); no paid edition, no seat and no support SKU
- Argo CD (best for: Teams that want Git as the desired state, and will self-host until the control plane is the on-call problem): From $495/month (Akuity Pro); free to self-host; qualifying startups get 6 months free
- Helm (best for: Anyone installing third-party software on Kubernetes, which is almost everyone): Free (Apache 2.0); no license or seat; private chart hosting is a separate bill

Kubernetes tooling is five different jobs sold as one shopping list. A desktop IDE, a terminal UI, a GitOps controller, a package manager and a fleet manager all show up in the same "best tools" roundup. They do not replace each other.

A team that "picks Kubernetes tools" is really picking a cluster UI, a way to ship YAML, and (sometimes) a control plane for every cluster.

The invoice is the part that surprises people. Three of these projects are $0 forever.

The money appears when someone sells you seats (Lens), a hosted Argo CD (Akuity), or a support contract on Rancher (SUSE).

## Top Picks

Based on features, real-world fit, and value for money.

Best Kubernetes Tools in 2026: 5 tools compared, updated Sep 2026

Tool | Pricing | Best for |

Lens | From $25/user/mo (Plus or Pro, billed yearly); free under $10 million, 30-day evaluation above | Engineers who want a desktop Kubernetes IDE, and companies over $10 million that have to pay for it anyway |

k9s | Free (Apache 2.0); no paid edition, no seat and no support SKU | People who already live in a terminal and do not want a license conversation |

Argo CD | From $495/month (Akuity Pro); free to self-host; qualifying startups get 6 months free | Teams that want Git as the desired state, and will self-host until the control plane is the on-call problem |

Helm | Free (Apache 2.0); no license or seat; private chart hosting is a separate bill | Anyone installing third-party software on Kubernetes, which is almost everyone |

Rancher | Community is free; Prime is quote-only; Rancher for AWS is $25/vCPU/month, 30-day trial | Teams running more than a couple of clusters who want one place for access, provisioning and upgrades |

Pricing read from each vendor's own published pricing page, checked Sep 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Sep 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### Lens

Top Pick

Best for: Engineers who want a desktop Kubernetes IDE, and companies over $10 million that have to pay for it anyway

PricingFrom $25/user/mo (Plus or Pro, billed yearly); free under $10 million, 30-day evaluation above

+Personal is a real client: multi-cluster, logs, metrics, Helm, port-forward, no tab cap

+Pro at $25/seat/year-billed is cheaper than most observability seats for the same humans

+RBAC stays on the cluster; Lens uses your kubeconfig, it does not install a privileged agent

−The $10 million rule is a hard cutoff, not a polite suggestion. Past it, Personal is a 30-day trial

−Plus and Pro buy Ask AI, EKS/AKS discovery and GitOps views. You can live without those if you already have kubectl

Visit Lens →

2

### k9s

Best for: People who already live in a terminal and do not want a license conversation

PricingFree (Apache 2.0); no paid edition, no seat and no support SKU

+Genuinely free for company use. No revenue clause, no 'Pro' paywall on logs

+Fast on a small machine and over SSH, which is how a lot of production access still works

+One binary. No cluster-side install, no control plane to patch

−No vendor to page. Bugs and upgrades are GitHub and your own packaging

−It will not onboard someone who has never used a terminal UI. Lens exists for that person

Visit k9s →

3

### Argo CD

Best for: Teams that want Git as the desired state, and will self-host until the control plane is the on-call problem

PricingFrom $495/month (Akuity Pro); free to self-host; qualifying startups get 6 months free

+The open-source install is the product. SSO, RBAC, ApplicationSets, Helm and Kustomize are not paywalled

+Drift detection is the point. kubectl-apply-and-forget is how clusters rot

+Akuity's $495 floor is a published number, which is more than SUSE gives you for Prime

−Self-hosting is free and is also a cluster you now patch, back up and page on

−$495 a month for 50 apps is cheap at 40 apps and silly at 8. Count applications before you talk to sales

Visit Argo CD →

4

### Helm

Best for: Anyone installing third-party software on Kubernetes, which is almost everyone

PricingFree (Apache 2.0); no license or seat; private chart hosting is a separate bill

+It is the distribution format. Datastores, ingress controllers and operators ship as charts

+No revenue threshold and no 'Team' tier. Finance has nothing to approve

+Works with Argo CD. Standardising on Helm does not lock you out of GitOps

−A chart is a templating language. Bad charts produce YAML nobody can read

−helm install by a human is still a snowflake. Pair it with Argo CD or you will not know what is live

Visit Helm →

5

### Rancher

Best for: Teams running more than a couple of clusters who want one place for access, provisioning and upgrades

PricingCommunity is free; Prime is quote-only; Rancher for AWS is $25/vCPU/month, 30-day trial

+Community is a full fleet manager, not a demo. Plenty of teams never buy Prime

+Talks to EKS, AKS, GKE, RKE2 and K3s. You are not buying a distro to get a UI

+The $25/vCPU AWS SKU is a real public number if you want SUSE on the hook and you are all-in on EKS

−Prime pricing is a form. Budget with a quote, not a blog post that guesses per-node dollars

−$25 per vCPU per month adds up on large node pools. Count vCPUs before that looks cheaper than self-hosting

Visit Rancher →

## What it is

These are the add-ons a platform team standardises on after the cluster exists. Lens and k9s are how humans look at pods, logs and events. Helm is how you install other people's applications as a versioned chart.

Argo CD is how Git becomes the desired state and the cluster is reconciled to it. Rancher is how you provision and govern many clusters from one place.

None of this is Kubernetes itself. kube-apiserver, etcd and the scheduler are already running.

These tools sit on top: a GUI, a packaging format, a GitOps loop, a multi-cluster console.

## Why it matters

A 20-person company can run this whole list at $0. The same list at a company over $10 million in funding, with 80 engineers on Lens Pro and a managed Argo CD, is $24,000 a year in Lens seats plus $5,940 a year at Akuity's $495 floor, before anyone talks to SUSE.

That is the comparison finance will understand.

The other reason is on-call. Self-hosting Argo CD or Rancher is free until the control plane is the thing that pages you. Buying the managed or Prime SKU is a decision about who gets that page, not about features you cannot get from the open-source build.

## Key features to look for

License shape
Apache 2.0 with no paid edition, free until a revenue threshold, per-seat subscription, per-application managed control plane, or quote-only support. This decides the three-year cost more than any feature row.

The job it actually does
Cluster UI, package manager, GitOps controller, or fleet manager. These five do not compete. Buying two UIs and no GitOps is a common way to spend money and still kubectl-apply on Friday.

Where it runs
A binary on a laptop (Lens, k9s, helm), a controller in the cluster (Argo CD), or a multi-cluster control plane (Rancher). The operational cost follows that, not the download page.

What the free edition cuts
Lens Personal cuts EKS/AKS discovery, GitOps views, Ask AI and commercial support, and is off-limits above $10 million revenue. Rancher community cuts the SUSE SLA and certified images. Argo CD and Helm cut nothing; the paid products are someone else's hosting.

Who you call at 3am
Community Slack, a Lens seat with commercial support, Akuity on a managed control plane, or a SUSE Prime contract. Free software is free until the person who understands the upgrade is on holiday.

## Pricing

k9s, Helm, Argo CD and Rancher community are free if you run them yourself, so the cheapest credible entry is $0. Lens and Akuity publish list prices, while SUSE Rancher Prime and Akuity Enterprise are quote-only, and the only public SUSE price is Rancher for AWS at $25 per vCPU per month.

The first paid Lens seat is Plus or Pro at $25 per user per month billed yearly ($300 up front, or $30 monthly), and it applies once revenue or funding is over $10 million.

Costs jump on Lens Enterprise at $50 per user per month yearly ($600 up front, or $60 monthly), on Akuity Pro at $495 a month for 50 applications plus $99 a month per extra pack, and when free Rancher is replaced by a Prime quote.

Plan | Price | Best for |

Lens Personal | Free | Under $10 million annual revenue or funding |

Lens Plus | $25/user/mo annual ($300 up front) or $30/mo | Individual seat after the 30-day evaluation |

Lens Pro | $25/user/mo annual ($300 up front) or $30/mo | Team seat after the 30-day evaluation |

Lens Enterprise | $50/user/mo annual ($600 up front) or $60/mo | Air-gapped, SAML/OIDC SSO, SCIM, audit logs and a TAM |

k9s | Free | Apache 2.0, with no paid edition or support SKU |

Argo CD | Free | Apache 2.0 and CNCF graduated; you host it |

Akuity Pro | $495/mo includes 25 million AI tokens | Argo CD and Kargo: 1 control plane each, 50 apps, 50 stages |

Akuity Pro extra pack | $99/month per pack | 10 apps, 10 stages, 5 million tokens, up to 1,500 apps |

Akuity (qualifying startups) | 6 months free | Qualifying startups |

Akuity Enterprise | Custom quote | Self-host, custom control planes and premium support |

Helm | Free | Apache 2.0; Harbor, ChartMuseum or OCI hosting is extra |

Rancher community | Free | Open source multi-cluster management you operate |

SUSE Rancher Prime | Custom quote | Enterprise support and certified images |

SUSE Rancher for AWS | $25 per vCPU per month | 30-day trial; Rancher Manager, SUSE Observability, AI SRE |

Mistakes to avoid

×Buying Lens Enterprise for a team that needed k9s and a wiki page. $50 a seat is the right price for air-gap, SAML and a TAM. It is the wrong price for someone who restarts a deployment twice a week.

×Calling Helm 'our GitOps'. helm upgrade from a laptop is a package manager. Argo CD is the loop that notices when production drifted. You usually want both, and they do not cost extra.

×Signing Akuity at $495 a month for a dozen applications. The floor includes 50. If you are at 12, self-host Argo CD until the control plane is the thing that pages you.

×Treating Rancher community as unfinished and buying Prime on day one. Community is the product. Prime is support, certified images and a longer Kubernetes window. Buy it when you need the SLA, not because the download page has two columns.

Expert tips

→Pick one cluster UI as the standard. Lens or k9s, not both as policy. People will install the other anyway. That is fine. Do not buy two seat SKUs.

→If the company is already over $10 million, budget Lens at $300 per user per year before someone quietly stays on Personal past the 30-day evaluation.

→Keep production kubeconfigs out of any vendor cloud workspace you have not read the policy for. Lens Personal talks to local kubeconfigs. That is the safer default.

## The bottom line

Start at $0. Helm is not optional if you install other people's software. k9s is the cluster UI that never generates a purchase order.

Add Argo CD when humans applying YAML is how production drifted last quarter; host it yourself until the control plane is the incident.

Lens is the right desktop IDE, and it is free until the $10 million line.

Past that, $25 a seat on annual is the real price, $50 if compliance needs air-gap and SAML. Rancher is the right call when the problem is many clusters, not one stubborn Deployment. Community is free. Prime is a quote.

The $25 per vCPU AWS SKU is the only SUSE number you can put in a spreadsheet without a sales call.

## Frequently asked questions

What is the best free Kubernetes tool?

For looking at a cluster: k9s, no caveats. For installing software: Helm. For GitOps: Argo CD, self-hosted. Rancher community is free if the job is many clusters. Lens Personal is free under $10 million revenue or funding and is a 30-day evaluation above that.

How much does Lens cost?

On k8slens.dev/pricing, checked 1 September 2026: Personal is free below $10 million annual revenue or funding. Plus and Pro are $25 per user per month on annual ($300 billed up front) or $30 monthly. Enterprise is $50 per user per month on annual ($600 billed up front) or $60 monthly. Organisations over the $10 million line must buy Plus, Pro or Enterprise after 30 days.

Is Rancher free?

The community edition is free and open source. SUSE Rancher Prime is a paid support and certified-image subscription with no public list price (rancher.com/pricing is a quote form). SUSE Rancher for AWS is $25 per vCPU per month on AWS Marketplace, with a 30-day trial, checked 1 September 2026.

Do I need Argo CD if I already use Helm?

Usually yes, if more than one person deploys. Helm packages the app. Argo CD keeps the cluster matched to Git. Using only Helm from laptops is how you get three versions of the same chart and no record of who installed them. They are not alternatives.

Lens vs k9s: which should we standardise on?

k9s if the team already works in a terminal and you want $0 with no revenue clause. Lens if you need a mouse, onboarding, or you are already over $10 million and will pay $25 a seat. Do not buy Lens Enterprise for people who will keep using k9s.

Related guides

Ci Cd ToolsServerless PlatformsMonitoring ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Load Testing Tools in 2026

URL: https://devshot.co/reviews/best-load-testing-tools
Type: review
Published: 2026-08-26
Updated: 2026-08-26
Summary: The load testing tools worth running in 2026, compared on whether you pay for virtual-user hours, test minutes or reports…

Expert Guide

## The Best Load Testing Tools in 2026

Three of these are free to run on your own machines. The hosted ones bill virtual users, test hours or reports, and those units are not interchangeable.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Aug 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

k6 is free as open source. Grafana Cloud k6 is $19 a month with 500 virtual-user hours included, then $0.15 per VUh. JMeter and Locust are free to run yourself. Gatling Community is free; Enterprise Basic is €89 a month billed annually. Artillery Cloud is free for 30 reports a month, then $199 on Team.

Key facts5 tools compared: k6, Apache JMeter, Gatling, Locust, Artillery

- Updated: August 26, 2026
- Top pick: k6 (best for: Teams that want tests as JavaScript in Git, and a hosted runner when the laptop is not enough)
- Top pick price as of August 26, 2026: k6: From $19/mo + $0.15 per virtual-user hour (Grafana Cloud k6 Pro); open source and cloud Free

- 5 tools compared: k6, Apache JMeter, Gatling, Locust, Artillery
- Apache JMeter (best for: Teams that already have JMeter scripts, or that need protocols beyond HTTP without a vendor): Free (Apache License 2.0); you pay for the machines
- Gatling (best for: Teams that want hosted Gatling with a published limit on hours and generators): From €89/mo (Enterprise Basic, billed yearly); Community free
- Locust (best for: Python teams that want scenarios as code and will run their own workers): Free, open source; you pay for the worker machines

Load testing tools all generate traffic. They do not all sell the same unit of traffic. k6 Cloud bills virtual-user hours. Gatling Enterprise bills test hours on a load generator. Artillery Cloud bills reports.

JMeter and Locust bill nothing, because you bring the machines.

A comparison that lines up their headline prices is therefore meaningless. $19, €89 and $199 buy different things, and the free tools are free until you count the box that runs them.

## Top Picks

Based on features, real-world fit, and value for money.

Best Load Testing Tools in 2026: 5 tools compared, updated Aug 2026

Tool | Pricing | Best for |

k6 | From $19/mo + $0.15 per virtual-user hour (Grafana Cloud k6 Pro); open source and cloud Free | Teams that want tests as JavaScript in Git, and a hosted runner when the laptop is not enough |

Apache JMeter | Free (Apache License 2.0); you pay for the machines | Teams that already have JMeter scripts, or that need protocols beyond HTTP without a vendor |

Gatling | From €89/mo (Enterprise Basic, billed yearly); Community free | Teams that want hosted Gatling with a published limit on hours and generators |

Locust | Free, open source; you pay for the worker machines | Python teams that want scenarios as code and will run their own workers |

Artillery | From $199/mo (Team); Cloud Free ($0); annual saves 20% | Teams that run a limited number of YAML or JavaScript suites each month and want hosted reports |

Pricing read from each vendor's own published pricing page, checked Aug 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Aug 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### k6

Top Pick

Best for: Teams that want tests as JavaScript in Git, and a hosted runner when the laptop is not enough

PricingFrom $19/mo + $0.15 per virtual-user hour (Grafana Cloud k6 Pro); open source and cloud Free

+The OSS runner is complete, so you can keep tests in CI without paying

+VUh pricing is published, including the formula: (max VUs × minutes) / 60

+500 VUh a month is enough for regular API checks before you hit overage

−Browser tests multiply the VUh cost, and a quote that assumed protocol VUs will be wrong

−The $19 platform fee is a floor even in a quiet month

Visit k6 →

2

### Apache JMeter

Best for: Teams that already have JMeter scripts, or that need protocols beyond HTTP without a vendor

PricingFree (Apache License 2.0); you pay for the machines

+Genuinely free, with no hosted upsell required to run a test

+Protocol coverage and plugins still beat most newer tools

+Twenty years of Stack Overflow answers for the ugly parts

−The GUI produces plans that are painful to review in Git

−Distributed mode is yours to operate, including the generators

Visit Apache JMeter →

3

### Gatling

Best for: Teams that want hosted Gatling with a published limit on hours and generators

PricingFrom €89/mo (Enterprise Basic, billed yearly); Community free

+Community Edition is free and is the same scripting model as Enterprise

+Basic publishes a hard cap: 1 hour, 1 generator, 2 seats, so the bill is predictable

+The recorder and the as-code SDK cover both first-pass and CI

−€89 for 60 minutes is expensive if you run tests daily rather than before a launch

−Prices are in euros

Visit Gatling →

4

### Locust

Best for: Python teams that want scenarios as code and will run their own workers

PricingFree, open source; you pay for the worker machines

+Tests are ordinary Python, so the same people who write the service can write the load

+Distributed workers are a first-class idea, not a paid add-on

+No vendor meter, so a long soak test costs only electricity

−There is no official hosted dashboard in the open-source project

−You own scaling, result storage and access control

Visit Locust →

5

### Artillery

Best for: Teams that run a limited number of YAML or JavaScript suites each month and want hosted reports

PricingFrom $199/mo (Team); Cloud Free ($0); annual saves 20%

+Free includes the CLI and 30 reports, which is enough to prove the workflow

+The Team cap of 1,000 reports is a lot if a report is a CI run, not a 4-hour soak

+Load testing and E2E sit in one subscription

−A report-based meter punishes chatty CI more than a short heavy test

−$199 is a jump from $0, with no published step between them

Visit Artillery →

## What it is

A load testing tool drives concurrent virtual users against an HTTP API, a website or a protocol, then reports latency, errors and saturation. The script lives in code (JavaScript, Java, Python, Scala) or in a recorder. The runners live either on your CI agents or on a vendor's generators.

The category split is simple.

Open-source runners are free and local. Hosted platforms rent you generators, dashboards and a place to keep results. Most vendors now sell both: a free engine, and a cloud that meters the run.

## Why it matters

The bill is set by how you count a test, not by how good the charts look. A 10-minute test at 100 virtual users is 16.67 VUh on Grafana Cloud k6. The same shape on Gatling Basic consumes part of a 1-hour, 1-generator allowance.

On Artillery Free it consumes one of 30 monthly reports. Those are not comparable line items.

The other cost is engineering time writing the script twice. Teams that pick a hosted recorder and later need the test in CI discover they have been using a product, not a file they can commit.

## Key features to look for

Billing unit
Virtual-user hours, test minutes on a generator, or reports per month. This is the number that decides the invoice, and no two hosted tools here use the same one.

Script as code
Whether the test is a file in Git. Recorders are fine for a first pass. CI only works if the suite is text you can review.

Distributed generators
One machine saturates before your app does. Hosted plans sell extra workers. Self-hosted tools need you to bring them.

Protocol versus browser
Protocol VUs hit HTTP. Browser VUs drive a real browser and cost more, often by a factor of ten on hosted k6. Check which one the quote assumes.

Where results live
A local HTML report is enough for a spike test. A retained dashboard matters when you compare this week's run to last month's. Retention is often the first paid gate.

## Pricing

k6, Gatling and Artillery publish list prices, checked 26 August 2026. JMeter, Locust, k6 open source and Gatling Community are free on your own machines, Artillery Cloud Free is $0, and Gatling Enterprise is quote-only (on request).

The cheapest paid entry is Grafana Cloud k6 Pro at $19 a month, with 500 virtual-user hours included and $0.15 per virtual-user hour after that.

Costs jump from Artillery's free plan to Team at $199 a month, then Business at $499 a month and Enterprise add-ons from $1,199 a month. Gatling Basic at €89 a month billed annually jumps to Team at €356 a month billed annually. Hosted prices buy different units (virtual-user hours, minutes on a generator, or reports), and Artillery annual billing saves 20%.

Plan | Price | Best for |

k6 open source | Free | Full runner on your own machines |

Grafana Cloud k6 Free | Free | 500 virtual-user hours per month |

Grafana Cloud k6 Pro | $19/mo, then $0.15/VUh | 500 VUh/mo included; automatic volume discounts; higher price for browser than protocol |

Apache JMeter | Free | Apache License 2.0; you pay for the machines |

Gatling Community | Free | Free for local use |

Gatling Enterprise Basic | €89/mo billed yearly (€1,068/yr) or €99/mo | 60 minutes, 1 load generator, 2 seats, 60,000 VUs cap |

Gatling Team | €356/mo billed yearly (€4,272/yr) or €396/mo | 300 minutes, 3 generators, 10 seats |

Gatling credit | 1 credit = 1 minute on 1 generator | Billing unit for the minute-based plans |

Gatling Enterprise | Custom quote | Enterprise plan, price on request |

Locust | Free | Open source; you pay for the worker machines |

Artillery Cloud Free | $0 | 30 reports/month, 1-month retention, 2 members, 5 workers/test, 30-minute tests, CLI usable |

Artillery Team | $199/mo; annual saves 20% | 1,000 reports, 6-month retention, 8 members, 25 workers, 2-hour tests |

Artillery Business | $499/mo; annual saves 20% | 2,500 reports, 18-month retention, unlimited members and workers |

Artillery Enterprise | From $1,199/mo; annual saves 20% | Enterprise add-ons |

Mistakes to avoid

×Comparing $19, €89 and $199 as if they were the same product. One is a platform fee plus VUh, one is an hour on a generator, one is a report quota. Convert your actual test shape into each unit before you rank them.

×Paying for a hosted recorder and keeping the test only there. When you need the suite in CI you will rewrite it. Start from a file in Git even if a vendor runs the generators.

×Sizing a quote on protocol VUs and then running browser tests. On Grafana Cloud k6 the browser multiplier will blow the number you showed finance.

Expert tips

→Write the test as code first on the free runner (k6, JMeter, Gatling Community, Locust or Artillery CLI). Buy generators only when one machine cannot reach the concurrency you need.

→Estimate one realistic run in each vendor's unit: VUh for k6, generator-minutes for Gatling, reports for Artillery. The cheapest headline is often the most expensive run.

→Keep a soak test and a spike test as separate scripts. Hosted plans with an hour cap are built for spikes. Soaks belong on your own hardware.

## The bottom line

For most teams the runner should be free. k6 if the team writes JavaScript, Locust if they write Python, JMeter if the scripts already exist or you need odd protocols. Pay for generators when a laptop cannot produce the load, not before.

When you do pay, pick the unit that matches how you test.

Regular API checks: Grafana Cloud k6 at $19 plus $0.15/VUh. Occasional launch tests with a published hour cap: Gatling Basic at €89. Many CI reports and a hosted archive: Artillery Team at $199. Do not buy all three.

## Frequently asked questions

What is the best free load testing tool?

k6, JMeter and Locust are all free to run yourself. Pick k6 for JavaScript, Locust for Python, JMeter if you already have JMeter plans or need protocols the others treat as plugins. Gatling Community and the Artillery CLI are also free for local runs. None of the hosted tiers is free in a way that replaces a serious suite.

How much does Grafana Cloud k6 cost?

Checked on 26 August 2026: a $19 per month platform fee on Pro includes 500 virtual-user hours, then $0.15 per VUh, with volume discounts. A test at 100 VUs for 10 minutes uses 16.67 VUh. Browser VUs cost more than protocol VUs. The open-source runner is free.

Is JMeter still worth using in 2026?

Yes, if you have existing plans or need its protocol and plugin range. It is a poor first choice for a new JavaScript or Python team, because the GUI scripts are hard to review and you will operate your own distributed generators. The price remains zero.

k6 vs Gatling vs Artillery: which should I pay for?

Pay for the unit you will consume. k6 if you think in virtual-user hours and already like Grafana. Gatling if you want a fixed hour-and-generator cap in euros. Artillery if a report quota matches CI. Convert one real test into all three meters before you pick. The open-source runners stay free in every case.

Related guides

Api Testing ToolsCi Cd ToolsError Tracking ToolsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Docker Desktop Alternatives in 2026

URL: https://devshot.co/reviews/best-docker-desktop-alternatives
Type: review
Published: 2026-07-09
Updated: 2026-07-09
Summary: The best Docker Desktop alternatives in 2026, from OrbStack to Podman to Colima, ranked on speed, licensing cost, and Docker compatibility.

Expert Guide

## The Best Docker Desktop Alternatives in 2026

Run containers on your Mac or Windows machine faster, lighter, and without the licensing bill. The runtimes worth switching to, ranked.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Jul 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

The best Docker Desktop alternatives in 2026 are OrbStack for the fastest, lightest experience on Mac, Podman Desktop for a daemonless open-source runtime, Rancher Desktop if you want built-in Kubernetes, Colima for a free CLI-only option, and Lima as the Linux-VM foundation many of the others build on. Most people leave Docker Desktop for licensing cost or performance, and one of these fixes both.

Key facts5 tools compared: OrbStack, Podman Desktop, Rancher Desktop, Colima, Lima

- Updated: July 9, 2026
- Top pick: OrbStack (best for: The fastest, lightest container experience on Mac)
- Top pick price as of July 9, 2026: OrbStack: Free for personal use; Pro $8/user/mo for commercial

- 5 tools compared: OrbStack, Podman Desktop, Rancher Desktop, Colima, Lima
- Podman Desktop (best for: A daemonless, rootless open-source runtime): Free (open source)
- Rancher Desktop (best for: Developers who want built-in Kubernetes): Free (open source)
- Colima (best for: A free, CLI-only runtime on Mac and Linux): Free (open source)

Docker Desktop stopped being free for larger companies after its 2021 licensing change, and even where it is free it eats CPU and battery. That pushed a wave of alternatives that run the same containers with less overhead or no license fee.

The good news: the Docker CLI and Compose files you already have keep working with almost all of them. We ranked on speed, resource use, licensing, and how close a drop-in each one is. Here are five worth switching to.

## Top Picks

Based on features, real-world fit, and value for money.

Best Docker Desktop Alternatives in 2026: 5 tools compared, updated Jul 2026

Tool | Pricing | Best for |

[OrbStack](https://toolradar.com/tools/orbstack) | Free for personal use; Pro $8/user/mo for commercial | The fastest, lightest container experience on Mac |

[Podman Desktop](https://toolradar.com/tools/podman-desktop) | Free (open source) | A daemonless, rootless open-source runtime |

[Rancher Desktop](https://toolradar.com/tools/rancher-desktop) | Free (open source) | Developers who want built-in Kubernetes |

Colima | Free (open source) | A free, CLI-only runtime on Mac and Linux |

[Lima](https://toolradar.com/tools/lima) | Free (open source) | The Linux-VM layer underneath other tools |

Pricing read from each vendor's own published pricing page, checked Jul 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Jul 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. Every tool here publishes a monthly price.

1

### OrbStack

Top Pick

Best for: The fastest, lightest container experience on Mac

PricingFree for personal use; Pro $8/user/mo for commercial

+Fast start times and very low CPU and battery use

+Drop-in for the Docker CLI and Compose

+Also runs full Linux machines alongside containers

−macOS only, no Windows or Linux

−Commercial use requires a paid Pro license

Visit OrbStack →

2

### Podman Desktop

Best for: A daemonless, rootless open-source runtime

PricingFree (open source)

+Free and open source, no license thresholds

+Daemonless, rootless architecture is more secure

+Runs on macOS, Windows, and Linux

−A few command and Compose differences to adapt to

−Rougher edges than a polished commercial app

Visit Podman Desktop →

3

### Rancher Desktop

Best for: Developers who want built-in Kubernetes

PricingFree (open source)

+Kubernetes built in and easy to toggle

+Free and open source on all three platforms

+Choice of Docker or containerd runtime

−Heavier than a plain container runtime

−Can be slower to start than OrbStack or Colima

Visit Rancher Desktop →

4

### Colima

Best for: A free, CLI-only runtime on Mac and Linux

PricingFree (open source)

+Completely free and open source

+Lightweight, scriptable, no GUI overhead

+Works with the standard Docker CLI and Compose

−No graphical interface at all

−macOS and Linux only, no Windows

Visit Colima →

5

### Lima

Best for: The Linux-VM layer underneath other tools

PricingFree (open source)

+Free and open source

+Highly configurable Linux VMs on macOS

+The foundation that powers several other tools

−Lower level and more manual to set up

−Not a polished, ready-made Docker replacement on its own

Visit Lima →

## What it is

Containers are Linux technology, so running them on macOS or Windows needs a lightweight Linux virtual machine underneath. Docker Desktop bundles that VM, the container runtime, and a GUI into one paid app.

These alternatives provide the same core, a fast VM plus a Docker-compatible runtime, so your existing docker and docker compose commands keep working. Some add a GUI, some are CLI-only, and some swap Docker's daemon for a daemonless, rootless design.

## Why it matters

Two things drive teams off Docker Desktop. The first is money: the license is required once a company crosses Docker's size and revenue thresholds, and that adds up across a large engineering org.

The second is performance, because Docker Desktop's VM is heavier than it needs to be and shows up as fan noise and drained batteries on laptops. A lighter runtime gives you faster container start times, lower idle CPU, and no license to track, usually without changing a single command in your workflow.

## Key features to look for

Docker CLI and Compose compatibilityEssential
Your existing docker and docker compose commands and files keep working, so switching is a swap of the engine underneath, not a rewrite.

A fast, low-overhead VMEssential
Containers on Mac and Windows run inside a Linux VM, and how efficient that VM is decides your start times, CPU use, and battery life.

Free for commercial useEssential
The whole reason most people leave Docker Desktop, so a permissive license or free tier for company use matters most.

Kubernetes support
A built-in local cluster for developers who test against Kubernetes rather than plain Docker Compose.

Rootless and daemonless security
Running containers without a privileged background daemon shrinks the attack surface, which is Podman's headline advantage.

Cross-platform coverage
Whether the tool runs on macOS, Windows, and Linux decides if a mixed team can standardize on one runtime.

Mistakes to avoid

×Assuming Docker Desktop is free for your company. The license kicks in above Docker's size and revenue thresholds, and running it unlicensed at a large employer is a compliance problem.

×Expecting a Mac-only tool to help Windows users. OrbStack and Colima do not run on Windows, so a mixed team should look at Podman Desktop or Rancher Desktop to standardize.

×Forgetting to point the Docker CLI at the new runtime. If DOCKER_HOST or the context still targets Docker Desktop, your commands hit the old engine and nothing seems to change.

Expert tips

→On a Mac, OrbStack is the easiest drop-in if you can pay for commercial use, and Colima is the go-to when it has to be free.

→Choose Podman if you want a daemonless, rootless setup, and budget a little time to adjust to its small command differences.

→Test your Docker Compose stack on the new runtime before migrating the whole team, since complex multi-service setups occasionally need small tweaks.

## The bottom line

On a Mac, OrbStack is the clear winner for speed and battery life if a commercial license fits your budget, and Colima is the free CLI-only pick when it does not.

For a cross-platform, fully open runtime, Podman Desktop is the strongest choice, and Rancher Desktop wins if you want a local Kubernetes cluster in the box. All of them keep your existing Docker commands working, so the switch is usually painless.

## Frequently asked questions

Why replace Docker Desktop at all?

Two reasons. Larger companies must pay for a Docker Desktop license once they cross Docker's size and revenue thresholds, and even where it is free, its VM is heavier than the alternatives and drains laptop battery. Tools like OrbStack fix both at once.

Is Docker Desktop still free?

It is free for personal use, education, open-source projects, and small businesses. It requires a paid subscription for larger companies above Docker's employee and revenue thresholds, which is what drives most enterprise teams to alternatives.

What is the best free Docker Desktop alternative for Mac?

Colima if you are comfortable with a CLI-only tool, or OrbStack's free tier for personal, non-commercial use. Both are Docker-compatible and far lighter than Docker Desktop. For commercial use that must stay free, Podman Desktop is the cross-platform answer.

Do these run Docker Compose?

Yes, nearly all of them support Docker Compose, since they implement the same Docker API. OrbStack, Colima, and Rancher Desktop are close drop-ins, and Podman supports Compose with a few small differences worth testing on a complex stack.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [OrbStack pricing](https://orbstack.dev/pricing)checked Sep 2026
- [Podman Desktop pricing](https://podman-desktop.io/#pricing)checked Sep 2026
- [Rancher Desktop pricing](https://rancherdesktop.io)checked Sep 2026
- [Lima pricing](https://lima-vm.io)checked Sep 2026

Related guides

Self Hosting ToolsGit ClientsAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# The Best Self-Hosting Tools in 2026

URL: https://devshot.co/reviews/best-self-hosting-tools
Type: review
Published: 2026-07-09
Updated: 2026-07-09
Summary: The best self-hosting PaaS tools in 2026, from Coolify to Dokploy to CapRover, ranked on features, ease of setup, and cost.

Expert Guide

## The Best Self-Hosting Tools in 2026

Get a Heroku-style deploy experience on your own VPS, with git push, automatic HTTPS, and one-click databases. The self-hosted PaaS tools worth running, ranked.

LC

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · 930,000+ readers · Updated Jul 2026

Product links may be affiliate links. How we rate 5 tools compared

TL;DR

The best self-hosting tools in 2026 are Coolify as the most popular and polished open-source PaaS, Dokploy for a cleaner modern take, CapRover as the mature Docker Swarm option, Dokku for a minimal push-to-deploy setup, and Portainer when you want a management UI over raw containers. All of them turn a cheap VPS into a Vercel or Heroku replacement you fully control.

Key facts5 tools compared: Coolify, Dokploy, CapRover, Dokku, Portainer

- Updated: July 9, 2026
- Top pick: Coolify (best for: The most complete open-source PaaS)
- Top pick price as of July 9, 2026: Coolify: Free (self-hosted, open source); Cloud from $5/mo

- 5 tools compared: Coolify, Dokploy, CapRover, Dokku, Portainer
- Dokploy (best for: A cleaner, more modern self-hosted PaaS): Free (open source); managed plan available
- CapRover (best for: A mature, configurable Docker Swarm PaaS): Free (open source)
- Dokku (best for: A minimal, Heroku-style push-to-deploy): Free (open source)

Managed platforms like Vercel, Heroku, and Railway are convenient right up until the usage bill arrives or you want your data on your own hardware. Self-hosted PaaS tools give you the same git-push-to-deploy experience on a VPS you rent for a few dollars a month, with automatic HTTPS, one-click databases, and a dashboard to manage it all.

You trade convenience for control and cost savings. We ranked on features, how smooth setup is, and price. Here are five worth running.

## Top Picks

Based on features, real-world fit, and value for money.

Best Self-Hosting Tools in 2026: 5 tools compared, updated Jul 2026

Tool | Pricing | Best for |

[Coolify](https://toolradar.com/tools/coolify) | Free (self-hosted, open source); Cloud from $5/mo | The most complete open-source PaaS |

[Dokploy](https://toolradar.com/tools/dokploy) | Free (open source); managed plan available | A cleaner, more modern self-hosted PaaS |

[CapRover](https://toolradar.com/tools/caprover) | Free (open source) | A mature, configurable Docker Swarm PaaS |

[Dokku](https://toolradar.com/tools/dokku) | Free (open source) | A minimal, Heroku-style push-to-deploy |

[Portainer](https://toolradar.com/tools/portainer) | Free up to 3 nodes; Business paid above | A management UI over Docker and Kubernetes |

Pricing read from each vendor's own published pricing page, checked Jul 2026. Every vendor here publishes a price.

Lowest monthly figure each vendor publishes, checked Jul 2026. A tilde marks a figure the vendor states approximately. Per-seat and usage charges can sit on top of it. 1 of 5 does not publish a comparable monthly price and is left out rather than estimated.

1

### Coolify

Top Pick

Best for: The most complete open-source PaaS

PricingFree (self-hosted, open source); Cloud from $5/mo

+Free and open source under Apache 2.0

+Hundreds of one-click services and templates

+Large, active community and fast development

−No built-in analytics or error tracking

−Broad feature set means more to learn upfront

Visit Coolify →

2

### Dokploy

Best for: A cleaner, more modern self-hosted PaaS

PricingFree (open source); managed plan available

+Free and open source

+Clean, approachable interface

+Fast-growing with a focused feature set

−Younger project with a smaller community

−Fewer templates and integrations than Coolify

Visit Dokploy →

3

### CapRover

Best for: A mature, configurable Docker Swarm PaaS

PricingFree (open source)

+Completely free, no commercial tier

+Mature and battle-tested for years

+One-click apps and simple scaling on Swarm

−Interface feels dated next to newer tools

−Docker Swarm is less current than the alternatives

Visit CapRover →

4

### Dokku

Best for: A minimal, Heroku-style push-to-deploy

PricingFree (open source)

+Very lightweight, runs on a small VPS

+Classic git-push-to-deploy workflow

+Extensible through a large plugin ecosystem

−Command-line first, no polished dashboard

−More manual setup than the newer GUI tools

Visit Dokku →

5

### Portainer

Best for: A management UI over Docker and Kubernetes

PricingFree up to 3 nodes; Business paid above

+Excellent visual container and stack management

+Handles Docker, Swarm, and Kubernetes

+Free for up to three nodes

−Manages containers, does not deploy from Git like a PaaS

−Business features are paid beyond three nodes

Visit Portainer →

## What it is

A self-hosted PaaS is software you install on your own server that recreates the developer experience of a managed platform. You connect a Git repository, the tool builds your app with a Dockerfile or a buildpack like Nixpacks, provisions any databases you need, sets up a reverse proxy, and issues TLS certificates automatically.

From then on a git push redeploys. You bring the VPS; the tool handles Docker, routing, certificates, and the app lifecycle so you do not wire it together by hand.

## Why it matters

Managed platforms bill by usage, and a project that costs a few dollars on a VPS can run into hundreds a month once it grows on a hosted platform. Self-hosting caps that at the price of your server, keeps your data and traffic under your control, and avoids vendor lock-in when a provider changes pricing or policy.

The trade is real: you now own uptime, backups, and security. These tools shrink that burden to a manageable size, but they do not erase it, so a backup and monitoring plan is not optional.

## Key features to look for

Git-based push-to-deployEssential
Connect a repository and every push builds and ships automatically, recreating the core convenience of a managed platform.

Automatic HTTPS and reverse proxyEssential
Free TLS certificates and routing set up for you, so each app gets a working domain without hand-editing Nginx or chasing certs.

One-click databases and servicesEssential
Provision Postgres, Redis, and other backing services from the dashboard instead of installing and securing each one manually.

Buildpack or Dockerfile builds
Build apps from a Dockerfile or automatically with a buildpack like Nixpacks, so you can deploy without writing container config.

Multi-server management
Manage more than one node from a single control plane once one VPS is no longer enough for your workloads.

Web dashboard and logs
A UI to deploy, roll back, read logs, and watch resource use, which is what separates a PaaS from a pile of shell scripts.

Mistakes to avoid

×Running production without a tested backup and restore plan. Self-hosting means you own recovery, and a database with no verified backup is a disaster waiting for a bad deploy.

×Putting everything on one tiny VPS with no monitoring. When the box runs out of memory or disk at 2am and nothing is watching, you find out from your users.

×Underestimating the security and maintenance burden. A managed platform patches the host for you; self-hosting hands you the firewall, OS updates, and exposed ports to manage.

Expert tips

→Start with Coolify or Dokploy for the smoothest modern experience, then graduate to something leaner only if you want more control.

→Put a firewall and automated, offsite backups in place before you point real traffic at the server, not after the first incident.

→Test the PaaS itself on a throwaway staging server first, so you learn its quirks before it is responsible for anything that matters.

## The bottom line

For most developers, Coolify is the place to start, since it is free, open source, and the most complete option with the largest community. Want something cleaner and more minimal, Dokploy is the modern challenger, and CapRover is the proven, no-cost veteran if you are comfortable on Docker Swarm.

Reach for Dokku when you want the leanest possible push-to-deploy, and add Portainer as a management dashboard over whatever you run.

## Frequently asked questions

Why self-host instead of using Vercel or Heroku?

Cost, control, and lock-in. A project that runs on a five-dollar VPS can cost far more on a usage-billed managed platform as it grows, and self-hosting keeps your data on your own infrastructure. The trade is that you take on uptime, backups, and security yourself.

Is Coolify really free?

Yes. The self-hosted version is open source under Apache 2.0 and free forever with every feature included, so you only pay for the VPS it runs on. Coolify Cloud is a paid managed option starting around five dollars a month if you would rather not host the dashboard yourself.

Coolify, Dokploy, or CapRover?

Coolify for the most features and the biggest community, Dokploy for a cleaner and more minimal interface, and CapRover if you want a mature, completely free tool and are happy on Docker Swarm. All three give you the same core git-push-to-deploy experience.

Is self-hosting production-ready?

Yes, with discipline. These tools run real production workloads every day, but you have to add automated backups, monitoring, and a firewall yourself. Skip those and self-hosting is fragile; put them in place and a single VPS can reliably replace a much larger managed bill.

## Sources

Prices and plan details come from each vendor's own pricing page, re-checked by the Toolradar pricing tracker.

- [Coolify pricing](https://coolify.io/pricing)checked Sep 2026
- [Dokploy pricing](https://dokploy.com)checked Sep 2026
- [CapRover pricing](https://caprover.com)checked Sep 2026
- [Dokku pricing](https://dokku.com)checked Sep 2026
- [Portainer pricing](https://portainer.io/pricing)checked Sep 2026

Related guides

Docker Desktop AlternativesTerraform Iac ToolsN8n ReviewFree Uptime Monitoring 2026: Free Plan Limits ComparedAI Coding Statistics 2026

Some offers on this page may be paid placements or contain affiliate links.

---

# Algolia Alternatives 2026: Search API Pricing Compared

URL: https://devshot.co/blog/algolia-alternatives
Type: blog
Published: 2026-10-14
Updated: 2026-10-14
Summary: Meilisearch Cloud is the cheapest hosted search API we sized at 100K records and 500K searches a month: $22.53. Algolia Grow costs $245.

Guide

## Algolia Alternatives 2026: Search API Pricing Compared

Meilisearch Cloud is the cheapest hosted search API we sized at 100K records and 500K searches a month: $22.53. Algolia Grow costs $245. Checked October 10, 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 14, 2026 · 8 min read

For 100,000 records and 500,000 searches a month, Meilisearch Cloud is the cheapest hosted search API we sized, at $22.53 a month on its resource-based XS instance. Typesense Cloud's single 1 GB node costs $28.80 before bandwidth, and Algolia's Grow plan costs $245 at the same volume. Every price was read on the vendor's own pricing page on October 10, 2026.

The Meilisearch figure holds only while records average under 3.5 KB. Self-hosting Meilisearch costs only the server you run it on, and this page does not price servers.

Devshot data: our list has 28,000 developers, and the average open rate is 30% (Beehiiv, updated September 20, 2026). Our readers build products, so every figure below is one production app's monthly bill.

## The summary table

Option | Billing basis | Cost at 100K records, 500K searches a month | Free allowance | Watch for |

Meilisearch Cloud, resource-based XS | Instance plus disk | $22.53 | 14-day trial | Steps up above about 3.5 KB per record |

Typesense Cloud, one 1 GB node | Per node hour, plus bandwidth | $28.80, or $29.25 with bandwidth | Free hours apply only to the 0.5 GB size | One node, no failover |

Typesense Cloud, 1 GB, three nodes | Per node hour, plus bandwidth | $108.00, or $108.45 with bandwidth | Same as above | The setup the vendor recommends for production |

Meilisearch Cloud, usage-based | Base fee plus overage | $210 | 14-day trial | Searches above 50,000 cost $0.40 per 1,000 |

Algolia Grow | Pay as you go, billed monthly | $245 | 10,000 searches included; the free plan has the same limit | Autocomplete counts a request per keystroke |

Algolia Grow Plus | Pay as you go, billed monthly | $857.50 | 10,000 searches included | Adds AI ranking; overage is $1.75 per 1,000 searches, 3.5 times Grow's rate |

Rows are sorted by cost at the scenario. Elastic is missing, and the reason is in its section below.

## How each bill is worked out

We priced the same scenario everywhere: 100,000 records and 500,000 search requests a month. The arithmetic is shown once here.

- Algolia Grow: 10,000 searches are included, so 490,000 are billed. That is 490 blocks of 1,000 at $0.50, or $245. The 100,000 records sit exactly at the included limit, so they add nothing.
- Algolia Grow Plus: the same 490 blocks at $1.75 come to $857.50.
- Meilisearch usage-based: the $30 base covers 100,000 documents and 50,000 searches. The other 450,000 searches are 450 blocks at $0.40, which adds $180, for $210.
- Meilisearch resource-based: the calculator puts this load on the XS instance (0.5 vCPU, 1 GB RAM). The instance costs $17.73 and the disk minimum of 32 GiB costs $4.80, so $22.53. The calculator shows the two lines rounded to $18 and $5, which is why its total reads $23.
- Typesense: one 1 GB node costs $0.04 an hour, and 720 hours is $28.80. The three-node setup costs $0.15 an hour in total.
- Bandwidth: we assume 10 KB per search response, so 500,000 responses are 5 GB. Typesense bills $0.09 per GB out, which is $0.45. Neither Algolia's nor Meilisearch's pricing page lists a bandwidth charge.

## Meilisearch: cheapest, with a size condition

Meilisearch's Cloud pricing has two models. The usage-based plan starts at $30 a month for 100,000 documents and 50,000 searches, and the calculator adds $0.30 per 1,000 extra documents and $0.40 per 1,000 extra searches. That is predictable at small volumes, but 500,000 searches push it to $210.

The resource-based model is the cheap one, because it bills the instance the calculator picks for your records and searches. The Cloud card says it starts at $20 a month, and the calculator's default estimate of 100,000 records and 10,000 searches shows $23 for the same XS instance.

The condition is memory. The calculator estimates about 0.86 GiB for 100,000 records at 3 KB each, which fits the XS instance's 1 GB. Once records average more than about 3.5 KB, the calculator moves to the S instance (1 vCPU, 2 GB), which costs about $40 a month with disk. Check your average record size before you choose this option.

Self-hosting is also an option. Meilisearch's pricing FAQ says it is MIT-licensed and has no usage limits when you run it on your own infrastructure.

## Typesense: priced by the hour, with no per-search meter

Typesense Cloud gives each cluster dedicated nodes, and its pricing page says there are no limits on records or operations. You pay per node per hour and per GB of bandwidth out. The smallest size, 0.5 GB, costs $21.60 a month, below every option priced here, and its first 720 hours are free once per account. We did not size it for this scenario. The vendor lists 0.5 GB as its minimum, and 100,000 records at 3 KB are about 300 MB of raw data before the index adds its overhead, which leaves little headroom. We priced 1 GB instead.

The page recommends high availability for production. Three 1 GB nodes come to $108.00 a month, which is the production figure in the table. Each node bills every hour it runs, busy or idle.

## Algolia: the leader's bill has more meters

Algolia Grow is pay as you go with no contract. It includes 10,000 search requests and 100,000 records a month, then bills $0.50 per extra 1,000 searches and $0.40 per extra 1,000 records. At this volume Grow costs $222.47 a month more than Meilisearch's resource-based instance.

Two rules change the bill. Autocomplete sends a new search request on every keystroke, so a search box can count far more requests than the searches a user finishes. And record overage is charged on the highest record count reached in the month, after ignoring the three busiest days.

Grow Plus adds AI synonyms, AI ranking and advanced personalization, and its search overage is $1.75 per extra 1,000. Algolia's enterprise tier has no public price, so it is not in the table.

## What changes at a million searches

Volume moves the ranking more than the entry price does. Each vendor scales differently, so the same four bills look very different at one million searches a month, with records held at 100,000.

- Meilisearch resource-based: the calculator sizes one vCPU per million searches a month, so the instance moves from XS to S (1 vCPU, 2 GB). That is $34.98 plus $4.80 of disk, about $40.
- Meilisearch usage-based: 950,000 searches above the included 50,000 add $380 at $0.40 per 1,000, for $410 a month.
- Algolia Grow: 990,000 billed searches at $0.50 per 1,000 come to $495 a month.
- Typesense: the node price does not move with search volume. The vendor advises at least 4 vCPUs above 20 requests a second, and a million searches a month averages about 0.39 a second. Check your peak traffic before you settle on the 1 GB size.

The ranking flips at a million searches. The 1 GB Typesense node, at $29.70 once bandwidth for 10 GB is added, undercuts Meilisearch's resource-based S instance at about $40, because Typesense does not add a charge for each search.

## Elastic: left out

Elastic's pricing page describes resource-based and usage-based billing for Elastic Cloud but lists no dollar rate. This table leaves Elastic out rather than quote a number from elsewhere.

## Which to pick at this volume

- Lowest bill we sized, records under 3.5 KB: Meilisearch resource-based. Its price moves with record size.
- Production failover with a per-node price: Typesense, three nodes, plus bandwidth. Idle nodes still cost money.
- Managed keyword search, if per-search billing is acceptable: Algolia Grow. Autocomplete raises the request count.
- A prototype: Algolia's free plan, or Typesense's first 720 hours on the 0.5 GB size.

The rest of the stack has its own bills. Vercel pricing, Supabase pricing and Supabase vs Firebase cover hosting and database costs, and cloud free tiers compared lists what the big clouds give away. For the backend around a search box, read best backend-as-a-service platforms, the Supabase review and the Firebase review. For cheaper hosting alternatives in general, see cheaper Vercel alternatives and object storage pricing compared.

## Method

We read each vendor's pricing page on October 10, 2026, and used only the rates those pages publish, including their calculators. The scenario is one production app with 100,000 records averaging 3 KB and 500,000 search requests a month, with bandwidth at an assumed 10 KB per response. We did not measure query speed or relevance. Those depend on your data and should be checked on it.

Prices were read on [Algolia pricing](https://www.algolia.com/pricing), [Meilisearch pricing](https://www.meilisearch.com/pricing), [Typesense Cloud pricing](https://cloud.typesense.org/pricing/calculator) and [Typesense's free tier article](https://cloud-help.typesense.org/article/how-does-the-free-tier-work). Vendor records are on Toolradar: [Algolia](https://toolradar.com/tools/algolia) and [Typesense](https://toolradar.com/tools/typesense).

## FAQ

### Is Meilisearch a free Algolia alternative?

Self-hosting Meilisearch has no software fee: it is MIT-licensed, and its pricing FAQ says it has no usage limits, so you pay only for the server. Hosted, the cheapest option we sized at this volume is Meilisearch Cloud's resource-based XS instance, at $22.53 a month for records averaging under 3.5 KB.

### How much does Algolia cost at 500,000 searches a month?

Grow costs $245 a month, because 10,000 searches are included and the other 490,000 cost $0.50 per 1,000. Grow Plus costs $857.50 at the same volume. The enterprise tier has no public price.

### Does Algolia have a free plan?

Yes. It covers 10,000 search requests and 50,000 records a month, with no credit card required. It suits a prototype or a small project, and this scenario is well past its limit.

### Is Typesense cheaper than Algolia?

At this scenario, yes. One 1 GB node costs $28.80 a month plus $0.45 bandwidth, and three nodes for high availability cost $108.00 plus bandwidth.

### Does Typesense charge per search?

No. Its pricing page describes a dedicated cluster with no limits on records or operations, and it bills nodes by the hour and bandwidth out per GB.

### Does autocomplete make Algolia more expensive?

Yes. Algolia's pricing FAQ says autocomplete sends a new search request on every keystroke, so the request count can run well above the searches a user completes.

Cite this: Devshot, "Algolia Alternatives 2026: Search API Pricing Compared", October 2026, with prices read on October 10, 2026.

Subscribe to the Devshot brief at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Authentication Pricing Compared (2026): Clerk vs Auth0 vs Supabase Auth vs Firebase at 10K, 100K and 1M Users

URL: https://devshot.co/blog/authentication-pricing-compared
Type: blog
Published: 2026-10-14
Updated: 2026-10-14
Summary: At 100,000 monthly active users, Supabase Auth costs $25 a month, Firebase $275 and Clerk $1,025, while Auth0 quotes no price. Checked October 10, 2026.

Guide

## Authentication Pricing Compared (2026): Clerk vs Auth0 vs Supabase Auth vs Firebase at 10K, 100K and 1M Users

At 100,000 monthly active users, Supabase Auth costs $25 a month, Firebase $275 and Clerk $1,025, while Auth0 quotes no price. Checked October 10, 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 14, 2026 · 8 min read

At 10,000 monthly active users, all four options cost $0 on an entry plan, so the real choice starts at 100,000. There, Supabase Auth is cheapest at $25 a month because its Pro plan includes 100,000 users, Firebase Identity Platform costs $275, and Clerk costs $1,025. Auth0 quotes no price at that volume.

Every figure below is in US dollars and was read on each vendor's own pricing page on October 10, 2026. The table covers the login line only. Databases, compute and enterprise contracts are priced separately, and the sections below show where each meter starts.

Devshot data: the audience list updated September 20, 2026 counts 28,000 developers, and our data puts the open rate at 30%. This page uses 10,000 and 100,000 users because three of the four free tiers end at 50,000 users and Auth0's ends at 25,000. Ten thousand sits inside all four, and 100,000 sits past all four.

## At a glance

Vendor | Entry plan | 10,000 users | 100,000 users |

Auth0 | Free to 25,000 MAU, then Essentials from $35 a month at 500 MAU | $0 on Free | Not quoted, about $7,000 at the plan rate |

Clerk | Hobby free to 50,000 retained users, then Pro | $0 on Hobby | $1,025 on Pro |

Supabase Auth | Free to 50,000 MAU, then Pro | $0 on Free | $25 on Pro |

Firebase Identity Platform | No cost up to 50,000 MAU, then Google Cloud rates | Free | $275 |

## 10,000 users: every entry plan is free, and the limits differ

At 10,000 users the bill is zero on all four, so the question is what each free plan refuses. Auth0's Free plan covers 25,000 monthly active users with no card, but multi-factor authentication factors are not on it. They start on Essentials, which is $35 a month at 500 users and, at the same rate per user, about $700 at 10,000.

Clerk's Hobby plan covers 50,000 retained users per app with three dashboard seats. MFA sits on Pro at a flat $25 a month. Firebase bills MFA and phone sign-in per text message on Google's table, outside the user count.

Supabase Free covers 50,000 users, but a free project pauses after a week without activity, and only two projects can be active. A live app therefore moves to Pro. Firebase's Identity Platform is no cost to 50,000 users, so at this volume the choice comes down to features and limits, not price. Firebase review and Supabase review cover the products themselves.

## 100,000 users: where the bill starts

Supabase Auth is the cheapest line at this volume. Pro includes 100,000 monthly active users for $25 a month, and each user above that costs $0.00325. The Pro fee covers the whole organization, and its $10 compute credit covers one Micro instance. The full Supabase invoice, with compute and add-ons, is in our Supabase pricing breakdown.

Firebase Identity Platform costs $275 a month at 100,000 users. Google's table gives the first 50,000 users no cost and charges $0.0055 for each user from 50,001 to 100,000. We read the table band by band, but it does not say whether a band's rate covers every user or only the users inside that band. If the rate applied to all 100,000, the bill would be $550, so check the Google Cloud calculator before you commit to a number.

Clerk costs $1,025 a month at 100,000 users on monthly billing. That is the Pro fee plus 50,000 users above the included 50,000, at $0.02 each. On annual billing the base fee drops to $20, which makes $1,020. Clerk's band table gives the same answer here, because every one of those 50,000 extra users falls in the $0.02 band.

Auth0 does not quote 100,000 users. Its calculator stops at a 30,000+ step, and Enterprise is a "Contact us" button. Auth0's plan data prices Essentials at $0.07 per user from 500 users up, so 100,000 users would be about $7,000 a month. That is our arithmetic from that rate, not a quote. Professional costs more: it starts at $240 a month for 500 users, where Essentials starts at $35.

## Clerk bills retained users, so count yours first

Clerk's base unit is a monthly retained user, not a monthly active user. A user counts as retained when they come back 24 hours or more after signing up. An app with 100,000 monthly actives may therefore bill a different number. Before you compare that line with anyone else's bill, count your retained users, because that is the number Clerk charges on.

Clerk's Pro plan also includes one SSO connection, and each extra connection costs $75 a month. Organizations work differently: Clerk includes 100 monthly retained organizations, and its Enhanced add-on costs $100 a month. Those lines are not in the monthly figure, so add them if you sell to companies.

## At a million users, the order holds and the gaps widen

Supabase Pro comes to $2,950 a month at a million users: the Pro fee plus 900,000 users above the included 100,000 at $0.00325 each. Firebase comes to $4,415: $275 for users 50,001 to 100,000 and $4,140 for the next 900,000 at $0.0046 each. Clerk comes to $17,225 read band by band, which is the Pro fee plus $1,000 for the 50,000 users in the $0.02 band and $16,200 for the 900,000 in the $0.018 band. Clerk's table does not say whether a band's rate covers every user or only the users inside it. Under the other reading the figure is $17,125. Auth0 Enterprise is a sales quote, so there is no published figure at a million users.

## The features that move the bill

Yearly billing on Auth0 charges 11 months. Essentials at 500 users comes to $385 a year on the annual toggle. Auth0 also keeps MFA factors off its Free plan, so an app that needs a second factor is on Essentials from the start.

Firebase charges extra for enterprise sign-in. SAML and OIDC are free up to 50 users per project, then $0.015 per user on Google's table. Phone sign-in and MFA are billed per SMS, separately from the user count, so a volume of text messages can outrun the user line.

Supabase's $10 credit covers compute only. It pays for one Micro instance, so every extra project is billed on top of the Pro fee, from $10 a month each.

## Which to pick, by user count

With no MFA, a free tier works up to 50,000 users on three of the four vendors, and Auth0's free plan ends at 25,000. Choose by the limits you can live with: Supabase Free pauses after a week, and Clerk Hobby stops at three seats. The Vercel review covers the hosting side of the same app. For a team that needs MFA, Clerk Pro includes it in a flat monthly fee, while Firebase bills it per text message.

Already paying for Supabase at 100,000 users, Pro is the cheapest login line, because the database is on the same invoice. Among standalone services at 100,000 users, Firebase Identity Platform is the lowest. The Firebase versus Supabase comparison covers when the database choice matters more than the login price. Teams that want Postgres without Supabase can compare hosts in the Postgres hosting list and read the Neon review.

Auth0's list rate implies about $7,000 a month at 100,000 users on Essentials, so ask for a written quote before you assume that figure. For the wider set of backend platforms, the backend-as-a-service roundup and the serverless platforms list show the alternatives.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this comparison. The prices come from the pricing pages of [Clerk](https://clerk.com/pricing), [Auth0](https://auth0.com/pricing), [Supabase](https://supabase.com/pricing) and [Firebase](https://firebase.google.com/pricing), plus the [Identity Platform pricing table](https://cloud.google.com/identity-platform/pricing), all read on October 10, 2026. Figures built from a vendor's per-user rate multiplied by the volume are our arithmetic, not a published total. Each vendor's product record is on Toolradar: [Clerk](https://toolradar.com/tools/clerk), [Auth0](https://toolradar.com/tools/auth0), [Supabase](https://toolradar.com/tools/supabase) and [Firebase](https://toolradar.com/tools/firebase).

## FAQ

### How much does Auth0 cost in 2026?

Auth0's Free plan is free for up to 25,000 monthly active users and needs no card. Essentials starts at $35 a month for up to 500 users, and Professional starts at $240 a month for up to 500. Above the calculator's 30,000+ step, Auth0 routes you to sales, and yearly billing charges 11 months.

### Is Auth0 free for 10,000 users?

Yes. The Free plan covers up to 25,000 monthly active users at $0. It does not include MFA factors, which start on Essentials.

### Is Clerk cheaper than Auth0?

At 10,000 users both are free on their entry plans, and Clerk Pro is the cheaper MFA option. At 100,000 users, Clerk costs far less than the roughly $7,000 a month that Auth0's Essentials implies. That figure is our arithmetic, not a quote.

### What is the cheapest authentication service for 100,000 users?

If you already run Supabase, Pro is the cheapest login line, because the plan covers 100,000 users. Among standalone services, Firebase Identity Platform at $275 a month is the lowest. Clerk costs more.

### Is Firebase authentication free?

It is free up to 50,000 monthly active users with Identity Platform. Above that, Google's table charges $0.0055 per user from 50,001 to 100,000, and phone sign-in and MFA are billed per SMS. SAML and OIDC sign-in is free for the first 50 users per project.

### Does Supabase charge per user for auth?

Yes, by monthly active user. Free includes 50,000 users, and Pro includes 100,000 in its monthly fee, then $0.00325 for each user above that. A free project pauses after a week without activity, and only two projects can be active.

### How much does Clerk cost for 100,000 users?

It costs $1,025 a month on monthly billing: the Pro fee plus $1,000 for the 50,000 users above the included 50,000 at $0.02 each. Annual billing lowers the base fee to $20, for $1,020. Clerk bills retained users, which can differ from monthly active users, so count yours before you compare.

Cite this: Devshot, "Authentication Pricing Compared (2026): Clerk vs Auth0 vs Supabase Auth vs Firebase", October 2026, with prices read on October 10, 2026.

Subscribe to the Devshot brief at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Free Uptime Monitoring 2026: Free Plan Limits Compared

URL: https://devshot.co/blog/free-uptime-monitoring
Type: blog
Published: 2026-10-14
Updated: 2026-10-14
Summary: UptimeRobot's free plan covers 50 monitors at 5-minute checks. StatusCake and Better Stack stop at 10. Free limits compared on vendor pricing pages.

Guide

## Free Uptime Monitoring 2026: Free Plan Limits Compared

UptimeRobot's free plan covers 50 monitors at 5-minute checks. StatusCake and Better Stack stop at 10. Free limits compared on vendor pricing pages.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 14, 2026 · 8 min read

UptimeRobot has the largest hosted free plan here: 50 monitors checked every 5 minutes at $0, according to its pricing page as read on October 10, 2026. StatusCake and Better Stack stop at 10 monitors, and Better Stack's free plan is for personal projects only. If you can run a server, Uptime Kuma is open source under the MIT license, and its README lists no monitor limit.

Toolradar data: of the 20 status page and uptime tools we rank, 15 have a free plan, as of October 2026. Free is common, so the limits decide which one fits.

Every limit in this guide comes from the vendor's own site, read on October 10, 2026. Where a vendor shows prices only in euros, we leave the price out rather than convert it. Free tiers change without notice, so check the vendor page before you open an account.

## What free uptime monitoring covers

Free uptime monitoring checks that a site or endpoint answers at a fixed interval, and alerts you when it stops. Free plans trim something: the number of monitors, how often each one runs, which alert channels you get, how long logs are kept, and whether you get a status page. The same trade-off runs through other free tiers, such as the AI API rate limits in Free LLM APIs in 2026.

## Free uptime plans at a glance

Tool | Monitors on free | Shortest free check | Alerts on free | Status page on free | Log retention on free | What forces an upgrade |

UptimeRobot | 50 | 5 minutes | Email and five integrations | 1 basic status page | 3 months | 60-second checks, 3 status pages and 12 months of logs (Solo); 3 login seats, webhooks, Zapier and PagerDuty (Team) |

StatusCake | 10 uptime monitors | 5 minutes | Alerts through integrations | Not listed on the free plan | Not stated on the pricing page | 100 uptime monitors and 1-minute checks (Superior) |

Better Stack | 10 monitors and heartbeats | Not stated for free | Slack and email | 1 status page | 3 days for logs, traces and web events | Phone and SMS alerts (Responder, $29 to $34 per license a month) |

Uptime Kuma (self-hosted) | No limit listed in its README | Set by you (README: 20-second intervals) | 90+ notification services | Multiple status pages | Set by you | Nothing from the vendor: you pay for the server |

Read across the table and two things stand out. Among hosted plans, UptimeRobot holds the most monitors and keeps three months of history, while Better Stack keeps logs for three days. Then check who the plan is for, because two of the free plans describe their intended user in their own wording.

## UptimeRobot: 50 monitors and no login seats

UptimeRobot gives the largest free allowance of the hosted plans: 50 monitors, each checked every 5 minutes. Its free plan lists HTTP, port and ping monitors, keyword, API, UDP and DNS monitors, and SSL and domain expiry alerts. Alerts go to email and five integrations, and the pricing page says "Only 5 integrations."

The free plan has no login seats, so a teammate cannot sign in to the dashboard. Logs stay for 3 months. SMS alerts are not included, but you can buy credits: 10 credits cost $3.00 and 100 cost $15.00, both one-time, per the pricing page. The free status page is basic, and an own domain for a status page starts on Solo.

Solo adds 60-second checks, three status pages on your own domain, and 12 months of logs. Team adds three login seats and webhook, Zapier and PagerDuty alerts. UptimeRobot lists its plan prices in euros only, so this guide gives no dollar price for the paid tiers. It also describes the free plan as good for hobby and non-profit projects, which is worth checking before a client site goes on it.

## StatusCake: 10 uptime monitors and three other test types

StatusCake's free plan gives 10 uptime monitors at 5-minute intervals. It adds one page-speed monitor tested every 24 hours, one domain monitor checked every 7 days, one SSL monitor checked every 24 hours, and alerts through integrations. That covers one marketing site well. It covers a portfolio poorly, because 10 uptime monitors is the whole uptime allowance.

Superior, the next plan, lifts uptime monitors to 100, drops the check interval to 1 minute, and adds three server monitors. StatusCake shows its plan prices in euros only. The page mentions USD invoices but no US dollar plan price, so this guide gives none.

## Better Stack: free for personal projects

Better Stack's free plan costs $0 and is labelled free for personal projects. It includes 10 monitors and heartbeats, one status page, Slack and email alerts, 100,000 exceptions a month and 5,000 session replays. Logs, traces and web events are kept for 3 days, with 3 GB of each.

The exceptions and session replays belong to error tracking and user recording, not uptime checks. For uptime alone, the 10-monitor allowance is the number that matters. The pricing page lists checks as frequent as 30 seconds for uptime monitoring, but it does not say which plan includes them, and it states no free check interval. Do not assume the free plan checks at any particular speed.

Phone and SMS alerts need the Responder license, which the page lists at $29 to $34 per license a month. Extra status pages cost $12 to $15 a month each, and packs of 50 extra monitors cost $21 to $25 a month. The page describes the free plan for personal projects, so a company site needs a paid plan under those terms.

## Uptime Kuma: free software, your own server

Uptime Kuma is free to download. Its GitHub repository carries an MIT license, and the README describes a self-hosted tool you can install with Docker. It checks HTTP(s), TCP, keyword, JSON query, WebSocket, ping, DNS, push, Steam game servers and Docker containers, with 20-second intervals listed as a feature. It sends alerts through 90+ notification services and runs multiple status pages.

The README lists no monitor limit. What you pay for is the server it runs on, and this guide does not price one. A cheap VPS for the job is covered in Best Cheap VPS Hosting in 2026, and other self-hosted options are in Best Self-Hosting Tools in 2026.

One trap decides whether self-hosting works. If Uptime Kuma runs on the same server as the site it watches, an outage of that server also silences the alert. Put the monitor on a separate host.

## Which free plan fits 25 client sites

Take a freelancer with 25 client sites, each checked every 5 minutes. That is 300 checks an hour across the set (25 monitors, 12 checks each per hour). Only two of the four options hold 25 monitors without a paid plan.

- UptimeRobot: the free plan holds 50, so 25 sites fit with 25 spare. UptimeRobot's own description of the free plan names hobby and non-profit projects, so client work needs a terms check.
- StatusCake: the free plan holds 10, so 15 sites need Superior, which the page prices only in euros.
- Better Stack: the free plan holds 10 and is labelled for personal projects. Packs of 50 extra monitors cost $21 to $25 a month on a paid plan.
- Uptime Kuma: it holds 25 if the server does, since the README lists no monitor limit.

At 5-minute checks, an outage can go unseen for up to five minutes.

## Gotchas before you pick

- Check speed: UptimeRobot and StatusCake check their free monitors every 5 minutes. Better Stack's free check interval is not on its page.
- History: UptimeRobot keeps logs for 3 months on free. Better Stack keeps logs, traces and web events for 3 days.
- Team access: UptimeRobot's free plan has no login seats, so a teammate cannot see the dashboard.
- SMS: UptimeRobot's free plan does not include SMS. Credits cost $3.00 for 10 or $15.00 for 100, one-time.
- Terms: Better Stack says personal projects, and UptimeRobot says hobby and non-profit projects. Read the terms against your use.
- Hosting: keep a self-hosted monitor off the host it watches.

## Frequently asked questions

### Is there a free uptime monitor with no monitor limit?

Uptime Kuma. Its README lists no monitor limit, its license is MIT, and the cost is the server you run it on. Among hosted free plans, UptimeRobot's 50 monitors is the largest allowance on this page.

### Which free uptime plan has the most monitors?

Among hosted plans, UptimeRobot has the most: 50 monitors on its free plan, each checked every 5 minutes. Uptime Kuma lists no limit, but you run it on your own server. StatusCake's free plan includes 10 uptime monitors, and Better Stack's includes 10 monitors and heartbeats.

### Does UptimeRobot's free plan include SMS alerts?

No. SMS and voice credits are not included on the free plan, but you can buy them: 10 credits cost $3.00 and 100 credits cost $15.00, one-time, per the pricing page read on October 10, 2026.

### How often does a free uptime check run?

Every 5 minutes on UptimeRobot's free plan and on StatusCake's free uptime monitors. Better Stack's pricing page does not state a free check interval.

### Who are the free plans for?

UptimeRobot describes its free plan as good for hobby and non-profit projects. Better Stack labels its free plan for personal projects. Check each vendor's terms before you put a client or company site on a free plan.

## How we checked

Plan limits and prices come from the vendors' own pricing pages, read on October 10, 2026: [UptimeRobot](https://uptimerobot.com/pricing/), [StatusCake](https://www.statuscake.com/pricing/) and [Better Stack](https://betterstack.com/pricing/). The Uptime Kuma license and feature list come from its [GitHub repository](https://github.com/louislam/uptime-kuma). The first-party count comes from Toolradar's [status page and uptime tools ranking](https://toolradar.com/best/status-pages), updated October 2026. Tool pages: [UptimeRobot](https://toolradar.com/tools/uptimerobot), [StatusCake](https://toolradar.com/tools/statuscake), [Better Stack](https://toolradar.com/tools/betterstack) and [Uptime Kuma](https://toolradar.com/tools/uptime-kuma).

Cite this: Devshot, "Free Uptime Monitoring 2026: Free Plan Limits Compared", October 2026.

## Keep reading on Devshot

- Best Monitoring Tools in 2026: observability platforms compared on billing model, for when uptime checks are not enough.
- Best Log Management Tools in 2026: log platforms compared, for the history that free plans trim.
- Best Error Tracking Tools in 2026: error tracking compared, the side of Better Stack's free plan that is not uptime.
- Sentry Review 2026: Sentry's pricing and limits.
- Cloud Free Tiers Compared (2026): the free tiers of AWS, Google Cloud, Azure and Oracle, for a server to host Uptime Kuma.
- Best Free Database Hosting in 2026: free database hosting tiers.
- Best Cheap VPS Hosting in 2026: cheap VPS plans for a self-hosted monitor.
- Best VPS Hosting for Developers in 2026: VPS hosting compared for developers.

Some offers on this page may be paid placements or contain affiliate links.

---

# Google Maps API Pricing 2026: Cheaper Alternatives per 1,000 Requests

URL: https://devshot.co/blog/google-maps-api-alternatives
Type: blog
Published: 2026-10-14
Updated: 2026-10-14
Summary: Stadia Maps, Geoapify, Mapbox and MapTiler priced against Google Maps for 100,000 map loads a month, on vendor rate cards checked October 10, 2026.

Guide

## Google Maps API Pricing 2026: Cheaper Alternatives per 1,000 Requests

Stadia Maps, Geoapify, Mapbox and MapTiler priced against Google Maps for 100,000 map loads a month, on vendor rate cards checked October 10, 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 14, 2026 · 10 min read

For 100,000 map loads and 20,000 geocoding requests a month, Stadia Maps' Standard plan is the lowest published price we found at $80, checked on the vendor's pricing page on October 10, 2026. Google Maps Platform bills the same usage at $680 on pay-as-you-go. The Stadia figure assumes an average load of 50 map tiles. Mapbox, at $250, is the cheapest option that needs no tile assumption.

A store locator shows the problem. Every store page loads a map, and every address typed into search or imported from a CSV gets geocoded. Once that traffic is steady, the map bill is a line item someone asks about. This page prices that usage on five map APIs, using only rates read on each vendor's own pages, and checks a sixth, HERE, which publishes no rates. It also shows where the cheapest option changes, because the vendors count usage in different units.

## How we priced it

Each vendor bills in its own unit. Google and Mapbox charge a map load when the map initializes. MapTiler charges a session for the same event. Stadia and Geoapify charge per tile and per credit, so their cost depends on how many tiles a load fetches. We keep each vendor's unit as published and do not convert between them.

Free allowances come off first. Each tier's rate then applies to the calls inside that band. All prices are in US dollars. Sources: [Google's per-SKU price list](https://developers.google.com/maps/billing-and-pricing/pricing), [Google's subscriptions](https://developers.google.com/maps/billing-and-pricing/subscriptions), [Mapbox pricing](https://www.mapbox.com/pricing), [MapTiler pricing](https://www.maptiler.com/cloud/pricing/), [Geoapify pricing](https://www.geoapify.com/pricing/) and its [pricing details](https://www.geoapify.com/pricing-details/), [Stadia Maps pricing](https://stadiamaps.com/pricing/), and [HERE's Base Plan page](https://www.here.com/get-started/pricing). All were read on October 10, 2026. Google's per-SKU list was last updated October 7, 2026.

Devshot data: our data model prices one usage profile on five published rate cards, read October 10, 2026. Dupple has no usage data for map APIs, so the scenario is set by us, and the arithmetic is shown.

For the same method applied to storage, see object storage pricing compared. AI calls are priced per million tokens in best LLM APIs.

## Cost at 100,000 loads and 20,000 geocodes

Sorted by monthly cost. Read the last column before you compare, because the units differ.

Provider | Plan or rate used | Free or included | Monthly cost | Billing unit |

[Stadia Maps](https://stadiamaps.com/pricing/) | Standard plan | 7,500,000 credits | $80 | 1 credit per vector tile, 20 per geocode |

[Geoapify](https://www.geoapify.com/pricing/) | API 50 plan | 50,000 credits a day | $179 | 0.25 credit per tile, 1 per geocode |

[Mapbox](https://www.mapbox.com/pricing) | Pay-as-you-go | 50,000 map loads, 100,000 geocodes | $250 | $5.00 per 1,000 loads above 50,000 |

[MapTiler](https://www.maptiler.com/cloud/pricing/) | Flex, $30 a month plus extras | 25,000 sessions, 3,000 search sessions | $260 | $2.50 per 1,000 extra sessions |

[Google Maps Platform](https://developers.google.com/maps/billing-and-pricing/pricing) | Pay-as-you-go, Essentials tier | 10,000 calls per SKU | $680 | $7.00 per 1,000 loads, $5.00 per 1,000 geocodes |

[HERE](https://www.here.com/get-started/pricing) | Base Plan | None published | No list price | Quote only |

### The arithmetic, once

- Google: 90,000 loads above the free 10,000 at $7.00 per 1,000 come to $630. The 10,000 geocodes above the free 10,000 add $50. Total: $680.
- Mapbox: 50,000 loads above the free 50,000 come to $250 at its rate for that band. The 20,000 geocodes sit inside the 100,000 free, so $0.
- MapTiler: the $30 Flex fee, plus 75,000 sessions above 25,000 at $2.50 per 1,000 ($187.50), plus 17,000 search sessions above 3,000 at $2.50 per 1,000 ($42.50). Total: $260. This assumes each geocode is one search session. MapTiler prices search sessions, not requests.
- Geoapify: 100,000 loads at 50 tiles and 0.25 credit per tile is 1.25 million credits. Adding 20,000 geocodes at one credit each gives 1.27 million a month, about 42,300 a day. The API 50 plan allows 50,000 credits a day, so it is $179.
- Stadia: 100,000 loads at 50 tiles and one credit per tile is 5 million credits. Adding 20,000 geocodes at 20 credits each gives 5.4 million, inside Standard's 7.5 million. Standard costs $80.

The 50-tile figure is Geoapify's own estimate of tiles per user session. Stadia publishes no tiles-per-load figure, so we apply the same count to it, and the result depends on that count. At 14 tiles per map view, Geoapify's other estimate, Stadia's Starter plan with 800,000 credits of overage comes to $44. Geoapify's API 25 plan at $109 covers the same load. Stadia stays the cheapest here, at $44.

Mapbox counts one map load for each initialization of its web map, and that load includes all tile requests. Its price does not move with tile counts.

## How the ranking moves with volume

Two other volumes, same method:

Monthly usage | Google Maps | Mapbox | MapTiler Flex | Geoapify | Stadia Maps |

10,000 loads, 2,000 geocodes | $0 | $0 | $30 | $59 | $20 |

200,000 loads, 40,000 geocodes | $1,340 | $650 | $560 | $299 | $146 |

At the smallest volume, Google and Mapbox both bill $0, because the free allowances cover the usage. Geoapify's free plan gives 3,000 credits a day, about 90,000 a month, and the 127,000 credits this load needs go past that, so its API 10 plan at $59 applies. Stadia's $20 uses the Starter plan. We assume its forward geocoding is included, since the page lists basic and standard geocoding APIs.

At 200,000 loads, Google costs about twice Mapbox. Stadia's cheapest plan is Standard with overage, 10.8 million credits in total, which comes to $146. In dollars, the gap between Google and Mapbox grows with scale: $430 a month at 100,000 loads and $690 at 200,000.

Over a year, the 100,000-load scenario costs $8,160 on Google, $3,000 on Mapbox and $960 on Stadia.

## Google: what the monthly bill contains

Google Maps Platform bills through the Google Cloud console, and [Toolradar profiles Google Cloud](https://toolradar.com/tools/google-cloud).

- The $200 credit is gone. Google's pricing page says that from March 1, 2025 it replaced the $200 monthly credit with free calls per SKU. Dynamic Maps and Geocoding each get 10,000 free a month. Any guide built on the old credit is out of date.
- The bands. Dynamic Maps costs $7.00 per 1,000 loads from 10,001 to 100,000, and $5.60 from 100,001 to 500,000. Geocoding steps down to $4.00 per 1,000 for requests from 100,001 to 500,000.
- Subscriptions. Starter is $100 a month for 50,000 combined calls, covering Dynamic Maps and Geocoding. Essentials is $275 for 100,000 calls, and Pro is $1,200 for 250,000. Unused calls do not roll over. Starter is the cheap Google route: 50,000 loads cost $280 pay-as-you-go, so Starter saves $180 a month.
- Over the limit. Calls above a plan's limit bill at pay-as-you-go rates. The scenario here, 100,000 loads plus 20,000 geocodes, is above Essentials' 100,000-call limit, so it is priced pay-as-you-go.

## The alternatives, one by one

### Mapbox: the price that needs no assumption

Mapbox's map loads are free to 50,000 a month, then $5.00 per 1,000 to 100,000, and $4.00 per 1,000 to 200,000. Its Temporary Geocoding API is free to 100,000 requests, then $0.75 per 1,000 to 500,000. The trap is storage. Results you keep, or batch-geocode, need the Permanent Geocoding API. That product has no free tier on the page, and it costs $5.00 per 1,000 for the first 500,000 monthly requests. Storing 20,000 geocodes costs $100 a month.

### Stadia Maps: cheapest paid tiers, priced in credits

Starter is $20 a month for 1,000,000 credits, with extra credits at 3 cents per 1,000. Standard gives 7,500,000 credits, with extra at 2 cents per 1,000. Professional gives 25,000,000 credits, with extra at 1.5 cents per 1,000. A vector tile costs one credit and a forward geocode costs 20, so geocoding is expensive per call. Stadia's free plan allows 200,000 credits a month, but it does not permit commercial use, so it fits prototypes only.

### Geoapify: a flat fee for a daily credit allowance

Geoapify sells daily credit allowances at a flat monthly fee: 10,000 credits a day for $59, 25,000 for $109, 50,000 for $179, and 100,000 for $299. A tile costs 0.25 credit and a geocode one. Its free plan gives 3,000 credits a day and allows commercial use with attribution. The limits are soft. Daily caps make peak traffic the risk, so check your busiest day, not only the monthly total.

### MapTiler: sessions plus API requests

MapTiler bills a session each time a map object initializes. Flex is $30 a month, with 25,000 sessions, 3,000 search sessions and 500,000 API requests. Above those, sessions and search sessions cost $2.50 per 1,000, and API requests cost $0.15 per 1,000. The Free plan, 5,000 sessions, is for testing and non-commercial use.

### HERE: no price to rank

HERE's Base Plan page offers a "Contact Us" button and no rates. A quote may suit a large account, but nothing on the page lets a buyer compare it, so HERE is left out of the ranking.

## Which one to pick

- Keep Google if you need the rest of its platform. This page prices map loads and geocoding only, and Google's cost at this volume is the highest here.
- Pick Mapbox when the bill must not depend on tile counts, and when you do not store geocodes, since storage is the costly tier.
- Pick Stadia when you can measure tiles per load and the traffic is large, on a paid tier. Its free plan does not allow commercial use.
- Pick Geoapify for a flat fee and steady traffic, and a free plan that allows commercial use with attribution.

The map is one line on a developer's bill, next to hosting, authentication and email. Cheaper Vercel alternatives prices hosting for a three-developer team, Vercel vs Netlify compares the two platforms, and authentication pricing compared covers the login layer. Best transactional email APIs prices email at 50,000 messages, and best serverless platforms lists where the app itself can run.

If you serve tiles yourself, the cost moves to a server bill. The best cheap VPS hosting guide prices that side. Free tiers across providers are compared in cloud free tiers compared, and free database hosting covers the same question for databases.

## Frequently asked questions

### What does the Google Maps API cost per 1,000 requests?

Dynamic Maps costs $7.00 per 1,000 loads from 10,001 to 100,000 a month, after 10,000 free loads. Geocoding costs $5.00 per 1,000 requests from 10,001 to 100,000, after 10,000 free. Both rates are from Google's per-SKU list, last updated October 7, 2026.

### Is there a cheaper Google Maps alternative?

Yes. For 100,000 map loads and 20,000 geocodes a month, Stadia Maps' Standard plan is $80, against $680 on Google pay-as-you-go. The Stadia figure assumes 50 tiles per load. Mapbox needs no such assumption, and it also undercuts Google at that volume.

### Is Mapbox cheaper than Google Maps?

At 100,000 loads and 20,000 geocodes, yes: $250 a month against Google's pay-as-you-go bill. Mapbox's 50,000 free loads beat Google's 10,000 free per SKU, and its map load includes tile requests. Stored geocodes cost more on Mapbox, and that tier has no free allowance.

### Is there a free Google Maps API?

Yes, inside per-SKU free calls: 10,000 Dynamic Maps loads and 10,000 geocodes a month. Google replaced its $200 monthly credit with those allowances on March 1, 2025. Guides that still quote the credit are out of date.

### Which map APIs have a free plan for commercial use?

Geoapify's free plan allows commercial use with attribution. Stadia's free plan does not, and MapTiler's Free plan covers testing and non-commercial work only. Google and Mapbox give free monthly allowances inside their pay-as-you-go pricing, rather than a separate free plan. Check their terms for your use case.

### Does Google Maps have a cheaper subscription?

Starter costs $100 a month for 50,000 combined calls, covering Dynamic Maps and Geocoding. That is $180 less than pay-as-you-go for 50,000 loads. Essentials costs $275 for 100,000 calls. Calls over a plan's limit bill at pay-as-you-go rates.

### Does HERE publish maps prices?

Not on its Base Plan page, which offers only "Contact Us". A quote is needed before any comparison, so HERE is not ranked here.

Cite this: Devshot, "Google Maps API Pricing 2026: Cheaper Alternatives per 1,000 Requests", October 2026.

Rates change without notice, so check each vendor's page before you commit a budget. Devshot's newsletter follows the developer tool and price changes behind pages like this one, at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Replit Pricing 2026: Every Plan, Agent Usage and Real Monthly Cost

URL: https://devshot.co/blog/replit-pricing
Type: blog
Published: 2026-10-14
Updated: 2026-10-14
Summary: Replit pricing in 2026: Core is $20/mo, Pro is $100/mo, and Agent work plus publishing draw on monthly credits. See the real monthly cost.

Guide

## Replit Pricing 2026: Every Plan, Agent Usage and Real Monthly Cost

Replit pricing in 2026: Core is $20/mo, Pro is $100/mo, and Agent work plus publishing draw on monthly credits. See the real monthly cost.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 14, 2026 · 11 min read

Key facts4 plans compared: Starter, Core, Pro, Enterprise

- Updated: October 14, 2026
- Prices as of: October 2026

- 4 plans compared: Starter, Core, Pro, Enterprise
- Starter (best for: Trying Agent on one small project): Free
- Core (best for: One builder shipping real apps): $20/mo, or $18/mo billed annually
- Pro (best for: A small team or heavy Agent use): $100/mo, or $90/mo billed annually
- Enterprise (best for: SSO and security review): Custom quote

Replit Core costs $20/mo and Replit Pro costs $100/mo, verified on Replit's pricing page in October 2026. Each plan comes with monthly credits worth its own price, and Agent work and published apps both spend those credits. Once they run out, usage bills on top, so the sticker is only the floor.

Replit is a browser-based platform where an AI Agent writes the app, then hosts it and stores its data in the same account. That makes it simpler to start than stitching together a code generator, a host and a database, and harder to budget, because three kinds of spend share one balance.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 155 tools, updated October 2026, and the directory record for this platform is [Replit on Toolradar](https://toolradar.com/tools/replit). The product verdict lives in our Replit review, so this page sticks to the bill.

Plan | Price | Best for | What changes the bill |

Starter | Free | Trying Agent on one small project | Daily Agent credits with a monthly cap, one published app that goes down after 30 days |

Core | $20/mo, or $18/mo billed annually | One builder shipping real apps | Monthly credits equal to the price, then usage-based billing |

Pro | $100/mo, or $90/mo billed annually | A small team or heavy Agent use | Credits equal to the price, 10 parallel agents, up to 15 collaborators |

Enterprise | Custom quote | SSO and security review | Sales, not checkout |

## How much does Replit cost per plan?

Core is the plan most people mean when they ask about Replit pricing. The plan includes Full build, Plan Mode, connectors, unlimited published apps and up to 5 builders in the workspace, plus monthly credits equal to the price. The annual rate of $18/mo saves $24 a year, which is worth taking only after a month or two proves you will keep building.

Pro suits a different buyer, since it adds 10 parallel agents, up to 15 collaborators, premium support and database rollback of up to 28 days, and the credit allowance rises to match its price. Pro credits roll over for two months. A solo developer who rarely spends the Core credits has no reason to pay five times the price, while a team that keeps hitting the ceiling gets a bigger pool and shared budget controls.

Enterprise has no public price, and Replit lists SSO and SAML, custom seat limits, single-tenant environments and static outbound IPs, which makes it the plan for a security review rather than a credit card.

Prices exclude tax, which depends on your location.

## Is Replit free?

Yes, Starter costs nothing. It gives you daily Agent credits that reset each day up to a monthly cap, Lite build for quick edits, and some monthly credit for production databases, storage and publishing. It also gives you one published app, which is taken down after 30 days and carries a "Made with Replit" badge.

That is enough to see whether the workflow suits you, and not enough to run anything for a customer. The 30-day expiry is the real limit: a free app you want to keep needs Core, where publishing is unlimited and the badge comes off.

## How does Replit Agent effort-based pricing work?

Agent bills by effort rather than by token, so each request ends in one checkpoint, and a bug fix costs less than a full feature or integration, which Replit bundles into a single larger checkpoint. Replit shows the cost of each checkpoint in the Agent tab, and asks you to confirm before a paid action starts.

Three modes set how fast the balance drains. Free Mode covers everyday chat and building within an allowance, and Core includes up to 30 hours of chat and up to 60 projects in it. Power Mode is the balanced paid option, and Max Mode uses the strongest models for large codebases and hard problems. Power and Max both draw on your monthly credits.

Replit publishes no dollar price per checkpoint, so you cannot compute an Agent bill from the docs alone. That matters if you came here after a runaway-bill thread: the safe habit is to start in Free Mode and use Plan Mode to scope work, because planning reasoning in the paid path can itself be billable, and to read the cost on each checkpoint before sending the next prompt. Agent features that call outside AI services, such as Claude, ChatGPT or Gemini image models, are billed at the provider's public API rate out of the same credits.

## How much does Replit hosting and publishing cost?

Publishing costs come out of the same monthly credits as Agent work, and you pay usage fees only after the credits are gone. That is why a modest app can look free on Core: the credits absorb it before the invoice moves.

Replit's [publishing costs page](https://docs.replit.com/billing/deployment-pricing) prices four deployment types:

Deployment type | Price | Suits |

Static | Free hosting, $0.05 per GiB of data transfer | Docs sites, portfolios, single-page apps |

Autoscale | $0.60 per million compute units and $0.40 per million requests, plus a small monthly base fee | Web apps and APIs with uneven traffic |

Scheduled | $0.60 per million compute units, scheduler free | Background jobs |

Reserved VM | $15/mo for 0.5 vCPU and 2 GB RAM, then $35, $50 and $130 for dedicated 1, 2 and 4 vCPU machines | Steady traffic and guaranteed resources |

Autoscale is the cheap one for a small app because it bills only while a request is being served and goes idle after 15 minutes. Replit's own examples put a small business site at about $2.39 a month and a 10,000-calls-a-day API at about $4.50. Reserved VMs are for apps that must stay warm, and a dedicated machine alone costs more than the Core plan.

Production databases bill on compute hours and peak storage, and development databases are free. App Storage is $0.015 per GiB a month with $0.05 per GiB of transfer.

Replit cut cloud prices on August 1, 2026, with Autoscale compute dropping from $3.20 to $0.60 per million units, outbound transfer from $0.10 to $0.05 per GiB and database storage from $1.50 to $0.35 per GiB. If you read an older Replit pricing post, its hosting numbers are now several times too high.

## What does Replit really cost a month?

These scenarios use only the plan prices and the publishing rates above. Agent spend is the unknown, so each row shows the bill before any Agent work beyond the included credits.

Scenario | Plan and setup | Monthly bill |

Weekend prototype | Starter, one 30-day published app | $0 |

Solo builder, one small app | Core with Autoscale hosting | The Core price, with hosting absorbed by the credits |

Always-on app, light Agent use | Core plus a $15 shared Reserved VM | The Core price, because the VM uses $15 of the credits |

Always-on app, bigger server | Core plus a $35 dedicated Reserved VM | $35 before any Agent work |

Three-person team | Pro, no per-seat fee | The Pro price, or the annual rate if you commit |

The second-to-last row is the one that surprises people. A $35 dedicated machine exceeds the monthly credit by $15, so the bill is $35 with zero Agent prompts, and every prompt after that is extra. Computed figure: a three-person team pays five times as much on Pro as on Core, though Core only fits if the group stays within 5 builders and one credit balance.

Credit packs are the top-up route when a heavy month runs past your balance. They come in $100, $300, $500 and $1,000 sizes at $100, $290, $480 and $950, they expire after six months, and they do not renew.

## How do I stop a surprise Replit bill?

Set a limit before the first paid prompt. In Settings you can cap usage beyond your monthly credits, and when the cap is reached usage-based services are blocked until the next cycle or until you raise it. Setting the limit to $0.01 keeps spending inside credit packs you have already bought, which suits anyone who wants a hard ceiling.

A usage page shows spend by project and resource, and usage can take up to 30 minutes to appear. On organizations, budgets are set in $500 increments. Pair a limit with Plan Mode and Free Mode, and the cost stops being a surprise.

## Replit vs Lovable, Bolt and v0: entry prices

The cheapest way to compare is the first paid plan. Prices below were read on each vendor's own pricing page in October 2026.

Tool | Free tier | First paid plan | What the plan buys |

Replit | Starter, daily Agent credits | Core, $20/mo | Credits equal to the price, hosting and storage in one bill |

[Lovable](https://lovable.dev/pricing) | 5 build credits a day, up to 30 a month | Pro, from $25/mo | 100 credits a month, no cap on workspace members |

[Bolt](https://bolt.new/pricing) | 1M tokens a month, 300K a day | Pro, $25/mo | 10M tokens to start, rollover, custom domains |

[v0](https://v0.app/pricing) | 7 messages a day | Plus, $30/user/mo | $30 of monthly credits per user |

Replit is the cheapest first paid plan on the table, and the only one that bundles hosting and a database into the same credits. The catch is on the team side: Bolt Teams and v0 Plus are $30 per member, so three people cost $90 there, close to Replit Pro's price but with separate allowances. Lovable bills by credits, not by seats, so a full workspace shares one balance. Our Lovable review covers that model, and vibe coding tools ranks the whole group.

If hosting is the part you would rather buy separately, price it against Vercel, the serverless platforms or a VPS for developers. If you want a code editor with agents and not an app builder, compare the Cursor pricing and Claude Code pricing pages instead. For a wider shortlist, see AI app development platforms, and the directory records for [Lovable](https://toolradar.com/tools/lovable) and [v0](https://toolradar.com/tools/v0) sit on Toolradar.

## Is Replit worth paying for?

Pay for Core when you want the app live this week and are comfortable with a balance that moves. Stay on Starter until a project is worth keeping online past 30 days. Move to Pro only when you have a team that needs shared credits, parallel agents or rollover, because that is what the extra money buys.

Skip Replit if a fixed monthly cost matters more than convenience. A credit balance that pays for prompts, servers and databases together cannot be forecast like a flat seat, and the lack of a published per-checkpoint price makes it harder still.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. Replit figures were read in October 2026 on its [pricing page](https://replit.com/pricing), the [AI billing doc](https://docs.replit.com/billing/ai-billing), the [Core](https://docs.replit.com/billing/plans/replit-core), [Pro](https://docs.replit.com/billing/plans/replit-pro) and [Starter](https://docs.replit.com/billing/plans/starter-plan) plan docs, [Managing your spend](https://docs.replit.com/billing/managing-spend), [Agent modes](https://docs.replit.com/chat/agent-modes), [publishing and database billing](https://docs.replit.com/billing/about-usage-based-billing), [App Storage billing](https://docs.replit.com/billing/object-storage-billing) and the [August 2026 cloud price update](https://docs.replit.com/billing/aug-cloud-billing-updates).

Competitor figures come from [Lovable's subscription plans doc](https://docs.lovable.dev/introduction/subscription-plans) and [pricing page](https://lovable.dev/pricing), [Bolt](https://bolt.new/pricing) and [v0](https://v0.app/pricing). The 155 tool count is from Toolradar's live AI coding ranking that month. Scenario totals are our arithmetic on those published prices, not a measured bill. More on how we publish is on the Devshot about page.

## FAQ

### How much does Replit cost in 2026?

Starter is free, Core is $20/mo or $18/mo billed annually, and Pro is $100/mo or $90/mo billed annually. Enterprise is a custom quote, while Core and Pro each include monthly credits worth the plan price, and Agent work, publishing and databases all draw on them. Prices are verified on Replit's pricing page in October 2026 and exclude tax.

### Is Replit free?

Yes, the Starter plan is free and includes daily Agent credits up to a monthly cap and Lite build. It allows one published app, which goes down after 30 days. Free Mode also exists inside Core and Pro as an allowance for everyday chat and building.

### How does Replit Agent pricing work?

Agent uses effort-based pricing: each request produces one checkpoint, and simple fixes cost less than full features. Free Mode is no-cost within an allowance, and Power and Max modes use paid credits. Replit asks you to confirm before a paid action starts and publishes no dollar price per checkpoint.

### How much does Replit hosting cost?

Static hosting is free with a $0.05 per GiB transfer fee, Autoscale bills compute and requests as used, and a shared Reserved VM is $15/mo. Hosting spends your monthly credits first, so a small app often adds nothing to a Core bill. Cloud prices dropped on August 1, 2026.

### What is the difference between Replit Core and Pro?

Core is the entry paid plan for one builder's workflow with up to 5 collaborators. Pro has 10 parallel agents, up to 15 collaborators, premium support, 28-day database rollback and two-month credit rollover. Both include the paid Agent modes, and Pro raises the credit pool and team limits.

### Is Replit cheaper than Lovable or Bolt?

At the first paid plan, yes: Core undercuts the $25 that Lovable Pro and Bolt Pro both charge. For teams it depends, since Bolt Teams is $30 per member, so three seats are $90, and Replit Pro is $100 with no seat fees. Replit also bundles hosting and a database into its credits.

### Can I cap my Replit spending?

Yes. You can set a usage limit that blocks usage-based services once it is reached, and a limit of $0.01 restricts you to credit packs you have already bought. Usage alerts and a spending dashboard help you watch the balance.

Cite this: Devshot, "Replit Pricing 2026: Every Plan, Agent Usage and Real Monthly Cost", October 2026.

Devshot covers the tools and bill changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Codex vs Claude Code 2026: Plans, Limits and Real Costs

URL: https://devshot.co/blog/claude-code-vs-codex
Type: blog
Published: 2026-10-13
Updated: 2026-10-13
Summary: Codex vs Claude Code in 2026: plan prices, usage meters, a five-seat team bill and API cost per million tokens, from vendor pages read October 10, 2026.

Guide

## Codex vs Claude Code 2026: Plans, Limits and Real Costs

Codex vs Claude Code in 2026: plan prices, usage meters, a five-seat team bill and API cost per million tokens, from vendor pages read October 10, 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 13, 2026 · 10 min read

Short answer: Claude Pro is the cheaper of the two standard plans, at $17 a month billed annually against $20 a month for ChatGPT Plus. Codex is the cheaper way to try an agent, because OpenAI's Free plan includes it at no charge and its Go plan costs $8 a month. Those lower tiers cover lighter work only. The heavy plans start at the same monthly price, so the choice there comes down to usage limits and surfaces.

The price gap is small. Where you work matters more. Claude Code runs in your terminal and opens pull requests. Codex runs in the terminal too, but it also lives in the browser, on iOS, and in cloud tools such as automatic code review. We did not run a benchmark, so this page does not rank code quality.

Toolradar data: the AI coding ranking evaluates 156 tools as of October 2026. Claude Code places third there, OpenAI Codex sits between 11 and 30, and Claude Code is one of two paid-only tools in the top 10. Both products are in that set, with a [Claude Code page](https://toolradar.com/tools/claude-code) and an [OpenAI Codex page](https://toolradar.com/tools/openai-codex).

## Quick verdict

Dimension | Claude Code | Codex |

Free access to the agent | No. Claude's Free plan excludes Claude Code | Yes, in the desktop app on GPT-6 Luna, subject to rollout |

Entry paid plan | Pro, billed annually at $17 a month, or monthly | Go, $8 a month for lightweight tasks; Plus includes all four Codex surfaces |

Heavier individual plan | Max, at 5x or 20x Pro usage per five-hour session | Pro, same starting price, with no five-hour limit at the moment |

Team seat | Standard, billed annually, for 2 to 150 seats | Business, billed annually, for 2 or more users |

Usage meter | Rolling five-hour window plus weekly limits; no message counts published | Ranges of local messages per five-hour period on Plus, described as estimates |

Where it runs | Terminal, VS Code, JetBrains, desktop app, web, mobile, Slack | Web, CLI, IDE extension, iOS, and cloud integrations such as automatic code review |

Pay per token | Claude Console account, at standard API rates | API key, at API rates, in the CLI, SDK and IDE extension only |

## What one developer pays

Two plans at the same monthly price look alike until you read the billing terms. Claude Pro billed annually comes to $200 up front, which is $40 less than ChatGPT Plus at its monthly rate for the year. OpenAI's pricing page prints no annual price for Plus, so the comparison uses the monthly rate. Billed monthly, Claude Pro and Plus are level. Our Claude Code pricing explainer lists every Claude plan, and our OpenAI Codex review covers Codex on its own.

Plan | Price | Per year | Agent included |

Claude Free | Free | Free | None. Claude Code is not included |

Codex on the Free plan | Free | Free | GPT-6 Luna at Standard speed in the desktop app, subject to rollout |

ChatGPT Go | $8 a month | $96 | Codex for lightweight coding tasks |

Claude Pro, billed annually | $17 a month ($200 up front) | $200 | Claude Code |

Claude Pro, billed monthly | $20 a month | $240 | Claude Code |

ChatGPT Plus | $20 a month | $240 at the monthly rate | Codex on the web, CLI, IDE extension and iOS |

Claude Max | From $100 a month, billed monthly | From $1,200 | Claude Code at 5x or 20x Pro usage |

ChatGPT Pro | From $100 a month | From $1,200 | Codex, with no five-hour limit at the moment |

The heavy tiers meet on price. Claude Max is priced against Claude Pro: its 5x and 20x tiers multiply Pro's usage in each five-hour session. Anthropic prints no dollar price for the 20x tier on its pricing page, so this page gives none. ChatGPT Pro has no five-hour limit at the moment, so the two heavy plans meter usage differently.

## Two ways to count usage

Claude counts usage in a rolling five-hour session and adds weekly limits on paid plans. The web app, the desktop app, mobile and Claude Code draw from one pool. A long agent session in the terminal therefore uses up the allowance you planned for chat. Anthropic publishes no message or token counts, so you cannot predict a session's cost in advance.

OpenAI prints ranges instead. On Plus, a five-hour period holds 350 to 3,000 local messages on GPT-6 Luna, 15 to 160 on GPT-6.1 Sol, 15 to 150 on GPT-6 Sol and 5 to 45 on GPT-6 Astra. The page calls these estimates, not fixed limits, and adds that weekly limits may also apply. The spread is wide. A developer who switches between Luna and Sol should plan around the low end of the Sol range.

When Plus or Pro users reach the limit, they can buy credits. OpenAI prints no dollar price per credit and says credit prices depend on your plan or agreement. The overage cost of Codex therefore cannot be stated from the page, while Claude Code can also bill per token through a Claude Console account, covered below.

## A five-seat team

Team pricing is closer than the individual one. Claude Team Standard and ChatGPT Business both charge $20 a seat billed annually and $25 billed monthly. At five seats, the two bills match exactly.

Five seats | Claude Team Standard | ChatGPT Business |

Billed annually, per month | $100 | $100 |

Billed annually, per year | $1,200 | $1,200 |

Billed monthly, per year | $1,500 | $1,500 |

Higher-usage seat, billed annually, per year | $6,000 for Premium seats with five times Standard usage | No higher-usage Business seat is printed |

The matching price moves the decision to features. Business adds SAML single sign-on, MFA, a dedicated workspace and larger virtual machines for cloud chats, and OpenAI says business data is not used for training by default. Claude Team adds Premium seats with five times the Standard usage, and it runs from 2 to 150 seats. If your team needs the Premium usage level, Claude is the only one of the two that prices it. Our best AI coding assistants ranking covers the wider field if neither tool fits.

## Pay per token through the API

Developers who skip subscriptions pay per token. Claude Code through a Claude Console account uses standard API pricing. Codex through an API key bills at API rates in the CLI, SDK and IDE extension, with no cloud features. On the list rates, Sonnet 5.5 and GPT-6.1 Sol are identical: $2 per million input tokens and $10 per million output tokens, with cached input at $0.10. Opus 5.5 costs double on each side.

Model, standard tier, per 1M tokens | Input | Cached input | Output |

Claude Sonnet 5.5 | $2.00 | $0.10 | $10.00 |

Claude Opus 5.5 | $4.00 | $0.20 | $20.00 |

GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |

gpt-5.3-codex, the Codex-labelled model | $1.75 | $0.175 | $14.00 |

Take a developer who sends 30 million input tokens and 3 million output tokens a month. With no caching, that is 30 x $2 + 3 x $10 = $90 on Sonnet 5.5 or GPT-6.1 Sol, $180 on Opus 5.5, whose rates are double Sonnet's on each side, and $94.50 on gpt-5.3-codex (30 x $1.75 + 3 x $14). If 80% of input is cached, Sonnet 5.5 and GPT-6.1 Sol both fall to $44.40 (24 x $0.10 + 6 x $2 + $30), Opus 5.5 falls to $88.80, and gpt-5.3-codex falls to $56.70.

The 80% cache rate is an assumption, not a measurement. Anthropic also bills cache writes at $2.50 per million tokens on Sonnet 5.5, and the Sonnet figures above leave that out. Anthropic's Claude Code page says Max includes $100 a month of API credits on 5x and $200 a month on 20x, which changes the math for a Max subscriber who would otherwise pay per token.

## Which fits how you code

The vendor pages we read publish no benchmark we can check, so this section compares surfaces and workflows, not code quality.

Claude Code runs in the terminal, as an extension in VS Code and JetBrains, in the desktop app on paid plans, on the web and on mobile. It can also start a coding task from Slack and works with GitHub and GitLab. It can open pull requests, so a task that ends in a PR fits the tool. Our how to use Claude Code guide covers the setup, and our Claude Code review covers the product in more depth.

Codex runs on the web, in the CLI, in an IDE extension and on iOS, and Plus includes all four. Its cloud integrations include automatic code review and Slack. Reviews run through GitHub count toward code review usage, and reviews run locally or outside GitHub count toward general usage. A team that reviews every pull request on GitHub should check that line of the usage page before it picks a plan.

OpenAI's Plus page names GPT-6.1 Sol and GPT-6 Luna, and its message-limit table also covers GPT-6 Sol and GPT-6 Astra. Claude Code's product page does not name a default model, so the vendor pages we read do not say which model a Claude plan runs by default. For model quality on programming tasks, our ChatGPT vs Claude for coding page covers the chat side.

## Free routes to try the agent

Claude has no free route to the agent. Its Free plan covers everyday chat and excludes Claude Code, which the Pro plan includes. Codex on OpenAI's Free plan runs GPT-6 Luna at Standard speed in the desktop app, subject to rollout. For a developer who wants to test an agent before paying, Codex is the route that costs nothing, so check that access when you sign up.

## Who should pick which

Pick Claude Code if you work in a terminal or VS Code and want the agent to finish at a pull request. Annual Pro is $3 a month under Plus at its monthly rate, which makes it the cheaper of the two standard plans.

Pick Codex if you already pay for ChatGPT and want the agent in that same account, or if you want a free route to test one before you spend anything. Go at $8 a month is the cheapest paid entry in this comparison, with the lighter limits described above.

For a five-seat team on annual billing, the bill is identical at $1,200 a year. Decide on the usage meter you can predict, the Premium seat if you need it, and whether you need SAML single sign-on, which Business includes.

If Cursor is also on your list, our Claude Code vs Cursor comparison covers that pair, and our OpenCode vs Claude Code page covers the open-source route. If Cursor's own plans matter to the decision, our Cursor pricing explainer lists them.

## How we ran the numbers

Plan prices and plan facts come from vendor pages read on October 10, 2026: Claude's pricing page and Claude Code product page at claude.com, ChatGPT's pricing page at learn.chatgpt.com, and OpenAI's API pricing page at developers.openai.com. The ranking figure comes from Toolradar's AI coding page, October 2026. Every cost on this page is our arithmetic on list prices, with no discount, promotion or credit purchase applied. Vendor pricing changes, so check both pricing pages before you buy.

Cite this: Devshot, "Codex vs Claude Code 2026: Plans, Limits and Real Costs", October 2026. Prices verified on vendor pages, October 10, 2026.

## FAQ

### Is Codex cheaper than Claude Code?

Codex's cheapest tiers cost less, but they are lighter. Codex is free on the Free plan and $8 a month on Go, against $17 a month for Claude Pro billed annually. Among the two standard plans, Claude Pro is $3 a month cheaper than ChatGPT Plus at the monthly rate. The heavy plans, Claude Max 5x and ChatGPT Pro, start at the same price.

### Does Claude's free plan include Claude Code?

No. Claude's Free plan covers everyday chat and excludes Claude Code, which starts with Pro. Codex on OpenAI's Free plan runs GPT-6 Luna in the desktop app, subject to rollout.

### How many messages does Codex allow in five hours?

On Plus, OpenAI prints 350 to 3,000 local messages per five-hour period for GPT-6 Luna, 15 to 160 for GPT-6.1 Sol, 15 to 150 for GPT-6 Sol and 5 to 45 for GPT-6 Astra. It calls these estimates, not fixed limits. Anthropic publishes no message counts for Claude Pro, only the five-hour window and weekly limits.

### Is Codex or Claude Code cheaper through the API?

Per million tokens, Sonnet 5.5 and GPT-6.1 Sol list at $2 input and $10 output, with cached input at $0.10. At 30 million input and 3 million output tokens a month, both come to $90 without caching and $44.40 with 80% of input cached, before Anthropic's cache-write charge. Opus 5.5 costs twice as much on each side.

### Does OpenAI have a team plan for Codex?

Yes. Business costs $20 per user a month billed annually for two or more users, or $25 billed monthly. Claude Team Standard costs the same per seat on the same billing terms and runs from 2 to 150 seats.

Devshot's newsletter covers these tools as they change. Sign up at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Datadog Alternatives 2026: Cheaper Monitoring Compared at 10 Hosts

URL: https://devshot.co/blog/datadog-alternatives
Type: blog
Published: 2026-10-13
Updated: 2026-10-13
Summary: At 10 hosts, Datadog Infrastructure Pro costs $150 a month billed yearly. Grafana Cloud and New Relic free plans cover this load. Prices checked October 2026.

Guide

## Datadog Alternatives 2026: Cheaper Monitoring Compared at 10 Hosts

At 10 hosts, Datadog Infrastructure Pro costs $150 a month billed yearly. Grafana Cloud and New Relic free plans cover this load. Prices checked October 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 13, 2026 · 11 min read

For 10 hosts sending infrastructure metrics, Grafana Cloud's free plan and New Relic's free plan both cost $0 a month, while Datadog Infrastructure Pro costs $150 a month on annual billing. Grafana Cloud Pro is the cheapest paid option at $19 a month on telemetry billing, and it keeps metrics for 13 months against Datadog's 15. Prices were checked on each vendor's own pricing page on October 10, 2026, in USD.

Grafana's free plan fits this load exactly at its series cap, and New Relic's fits while data stays under 100 GB. The table shows where each one stops.

Toolradar data: of the 187 monitoring tools we rank, 115 (61%) have a free plan, as of October 2026, according to our [monitoring guide](https://toolradar.com/best/monitoring). Free tiers are common in this category. Their ceilings decide which one fits your host count. The wider field, ranked by billing model, is in best monitoring tools.

## Cost at 10 hosts, cheapest first

Tool | Entry price | Cost at scenario | Free plan (limits) | Billing basis | Hidden costs |

[Grafana Cloud](https://toolradar.com/tools/grafana-cloud) Free | Free | $0 a month | 10k active series, 14-day retention, 3 active users | Free, no card | The scenario uses the full 10k series allowance |

[New Relic](https://toolradar.com/tools/new-relic) Free | Free | $0 a month if data stays under 100 GB | 100 GB data ingest, one full-platform user, unlimited basic users | Data ingest plus full users | $0.40 per GB past 100 GB; a second full user needs Standard, at $10 for the first and $99 for each more |

[Grafana Cloud](https://toolradar.com/tools/grafana-cloud) Pro | From $19 a month | $19 a month | None; 10k active series and 3 active users sit inside the fee | Platform fee plus usage | $6.50 per 1,000 series past 10k; host-hour billing would make it $69.68 |

[Better Stack](https://toolradar.com/tools/betterstack) Telemetry Nano | $25 a month, billed yearly | $25 a month if metrics fit 40 GB | No free Telemetry bundle; 40 GB each of logs, metrics and traces | Bundle, $30 billed monthly | Usage past the bundle is billed by GB; team members are free |

[SigNoz](https://toolradar.com/tools/signoz) Teams | $49 a month | $49 a month | Community edition is open source, run by you | Usage included, then per sample | $0.10 per million metric samples past the included usage; $0.30 per GB of logs or traces |

[Datadog](https://toolradar.com/tools/datadog) Infrastructure Pro | $15 per host a month, billed yearly | $150 a month ($180 on-demand) | Up to 5 hosts, 1-day metric retention | Per host | APM from $31 per host a month; logs $0.10 per GB ingested |

Datadog's $150 is the annual figure. The same 10 hosts cost $180 a month on on-demand billing. Both free plans cost nothing here, but neither has much headroom: the 10-host load uses all of Grafana's series allowance, and New Relic's free data limit holds only if each sample is small. The arithmetic follows.

## Our 10-host scenario

The buyer runs 10 always-on Linux hosts. Each host sends about 1,000 active metric series, sampled once a minute, which comes to 10,000 series and about 432 million samples a month. Three people need access, and one of them builds the dashboards and alerts. The scenario leaves out APM traces and logs, because every vendor prices those separately. Logs get their own section below.

The 1,000 series per host is our assumption, not a vendor figure. It drives the usage bills of Grafana Pro and SigNoz, so the flip points further down show what happens at other sizes.

Ten hosts is also the size where Datadog's free plan stops working. That plan covers up to 5 hosts, so a 10-host team on Datadog pays for Infrastructure Pro.

## The arithmetic, shown once

Datadog. Ten hosts at $15 is $150 a month, or $1,800 a year, on annual billing. On-demand billing is $18 per host, which makes $180 a month.

Grafana Cloud Free. Ten hosts at 1,000 series each use all 10,000 active series the free plan allows, the same method as our cheaper Vercel alternatives comparison. There is no room for growth, and the free plan keeps only 14 days of data.

Grafana Cloud Pro. The platform fee includes 10,000 active series and 3 active users, so usage is zero and the bill is the platform fee alone. Each extra 1,000 series adds $6.50.

SigNoz Teams. Ten thousand series at 43,200 samples each (one per minute for 30 days) is 432 million samples. That is $43.20 of usage, which fits inside the plan's included usage, so the bill is $49 a month.

New Relic Free. The one full-platform user is free, and the two other people use basic users at no charge. Data counts toward the 100 GB monthly allowance. At 432 million samples, 100 GB holds as long as each sample averages less than about 230 bytes. New Relic does not publish bytes per sample, so the free price holds only under that threshold.

Better Stack Telemetry Nano. $25 a month billed yearly, or $30 billed monthly. The bundle includes 40 GB of metrics. Better Stack does not convert samples into GB, so the price holds only if your 10 hosts' metrics fit inside that bundle.

Grafana host hours. Grafana also bills by host hour if you opt into its curated experience. Ten hosts running 730 hours each make 7,300 host hours. The fee includes 2,232, which leaves 5,068 at $0.01 each, or $50.68. Add the platform fee and the bill is $69.68 a month. Opt out of the curated experience and billing follows telemetry alone, which is the Pro path above.

## Datadog: the annual rate is the one to compare

Datadog's Infrastructure Pro is $15 per host a month billed annually and $18 on-demand, according to its pricing page. Pro includes 1,000-plus integrations, out-of-the-box dashboards and 15 months of metric retention. Enterprise is $23 per host. Some roundups quote $23 for Pro, but on the pricing page that is the Enterprise rate.

APM starts at $31 per host a month billed annually. Adding it to these 10 hosts adds $310 a month. Logs are a separate meter, billed per GB ingested with indexing on top. Our Datadog review covers the platform beyond its pricing.

## Grafana Cloud: the Pro fee covers 10,000 series, with one trap

Grafana's free plan allows 10,000 active series, 14 days of retention and 3 active users. Pro starts at $19 a month, includes the same 10,000 series and 3 users, and keeps 13 months of data. Above the included series, Grafana bills $6.50 per 1,000 series. Extra active users cost $8 each beyond three. For the free allowances of hosting platforms, see cloud free tiers compared.

The trap is the billing mode. Grafana's FAQ defines a host as any operating system instance sending signals, and it says host-hour billing applies when you opt into the curated experience. For this scenario that switch moves the bill from the platform fee to $69.68 a month. Check which mode your account uses before you roll agents out across the fleet.

## New Relic: free for one full user, with a data ceiling

New Relic's free plan includes 100 GB of data ingest a month, one full-platform user and unlimited basic users. Basic users can access custom dashboards, queries and alerts, which covers two of the three people in this scenario. The builder needs the full-platform user, and the free plan provides one.

On the Standard edition, the first full user costs $10 and each additional full user costs $99, up to five. New Relic's default data retention is at least 8 days. Standard support is ticketed, with a 2-business-day response target. A team that wants a support queue with a fixed response time should price that in.

## SigNoz and Better Stack: flat bundles, usage on top

SigNoz Teams starts at $49 a month and includes that much usage. SigNoz says you can add unlimited teammates and monitor any number of hosts. Past the included usage, SigNoz bills $0.30 per GB of logs or traces and $0.10 per million metric samples. The open-source Community edition is the free route if you can run the stack yourself on a cheap VPS.

Better Stack's Telemetry Nano bundle is $25 a month billed yearly and $30 billed monthly. It includes 40 GB each of traces, logs and metrics, with 30 days of log retention. Team members are free on Telemetry. Better Stack also sells uptime and incident tools, which this scenario does not use. The bundle carries logs, traces and metrics under one price, which matters if you were going to buy logs anyway. Its metrics are billed by GB, so check that figure against your own agents.

## Where the ranking flips

The order above holds for this scenario. Three changes would move the order or the margins.

- Grafana Pro against SigNoz. Grafana Pro costs its platform fee plus $6.50 for each 1,000 series above 10,000. SigNoz costs $0.00432 per series a month at one-minute sampling once usage passes its included amount. The two lines cross at about 21,100 series, roughly 2,100 per host across 10 hosts. Below that, Grafana Pro is cheaper. Above it, SigNoz is.
- Host count. Datadog scales with hosts on annual billing. Take the same 1,000 series per host to 20 hosts, and Grafana Pro is $84, SigNoz is $86.40 and Datadog is $300 a month.
- New Relic's data line. The free plan holds while data stays under 100 GB. Past that, New Relic bills $0.40 per GB, so the data footprint, not the host count, decides when the free plan ends.

## What the cheaper plans give up

- Retention. Datadog keeps metrics for 15 months on Pro. Grafana Pro keeps 13 months, and Grafana's free plan keeps 14 days. New Relic's default is at least 8 days. SigNoz's metrics pricing offers retention from 1 to 13 months.
- Ecosystem. Datadog lists more than 1,000 integrations. Grafana's pages name native OpenTelemetry and Prometheus support, which matters if your agents already speak those formats.
- Support. Grafana Pro includes 8x5 email support. New Relic Standard support is ticketed with a 2-business-day response target. Datadog's support tiers sit outside this scenario.
- Error tracking. Errors are a separate job. If you need them, price them with the metrics, starting from best error tracking tools or the Sentry review.

## Logs change the answer

This scenario leaves out logs, and every vendor meters them differently. Datadog bills $0.10 per GB ingested, with indexing on top. SigNoz bills $0.30 per GB of logs. Better Stack includes 40 GB in its Nano bundle. Grafana bills logs through three meters: processing from $0.05 per GB, writes from $0.40 per GB and retention at its own rate. New Relic's 100 GB free allowance counts logs too, so a log-heavy team reaches that ceiling fast. The wider field is covered in best log management tools. Price your real log volume on each tool before you choose, because a plan that wins on metrics can lose on logs.

## Frequently asked questions

### What is the cheapest Datadog alternative for 10 hosts?

Grafana Cloud's free plan and New Relic's free plan both cost nothing for this load, as long as Grafana stays at 10,000 active series and New Relic stays under 100 GB of data a month. The cheapest paid plan on telemetry billing is Grafana Cloud Pro, which includes 10,000 series and keeps metrics for 13 months.

### How much does Datadog cost per host in 2026?

Datadog Infrastructure Pro costs $15 per host a month billed annually or $18 on-demand, according to its pricing page, checked October 10, 2026. APM starts at $31 per host a month billed annually. Logs are billed separately per GB ingested, with indexing on top.

### Is there a free Datadog alternative that covers 10 hosts?

Yes, with limits. Grafana Cloud Free allows 10,000 active series, 14 days of retention and 3 active users, which is exactly this scenario's series load. New Relic Free allows 100 GB of data ingest a month, one full-platform user and unlimited basic users. Neither plan leaves much room to grow.

### When does SigNoz become cheaper than Grafana Cloud Pro?

At about 21,100 active series with one-minute sampling, or roughly 2,100 series per host across 10 hosts. Below that, Grafana Pro is cheaper. The crossover moves with the sample interval, so recompute it if your agents scrape every 15 seconds.

### Does the 10-host result hold once logs are added?

Not necessarily. Datadog bills per GB ingested, SigNoz bills $0.30 per GB, Better Stack includes 40 GB in its Nano bundle, and Grafana meters processing, writes and retention separately. Price the same log volume on each tool before you choose.

Cite this: Devshot, "Datadog Alternatives 2026: Cheaper Monitoring Compared at 10 Hosts", October 2026.

## How we checked these prices

We read the pricing pages of [Datadog](https://www.datadoghq.com/pricing/), [Grafana Cloud](https://grafana.com/pricing/), [New Relic](https://newrelic.com/pricing), [Better Stack](https://betterstack.com/pricing) and [SigNoz](https://signoz.io/pricing/) on October 10, 2026, and used the list rates each publishes in USD, on annual billing where the page offers it. The monitoring category count comes from [Toolradar's monitoring guide](https://toolradar.com/best/monitoring), October 2026.

The scenario inputs (10 hosts, 1,000 series per host, one sample a minute, three users) are ours. The vendor rates are theirs. We left out Honeycomb, which prices by events and lists no per-event rate above its Pro allowance, and Axiom, which prices by data loaded in GB and does not map onto a host count.

Devshot covers the tools and bills developers have to act on. For pricing changes like these in your inbox, subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# How to Get GitHub Copilot for Free in 2026: Free Plan, Students, Teachers and Open Source

URL: https://devshot.co/blog/how-to-get-github-copilot-for-free
Type: blog
Published: 2026-10-13
Updated: 2026-10-13
Summary: GitHub Copilot is free on the Free plan, for students, and for teachers and maintainers who qualify. Limits, sign-up status and claim steps.

Guide

## How to Get GitHub Copilot for Free in 2026: Free Plan, Students, Teachers and Open Source

GitHub Copilot is free on the Free plan, for students, and for teachers and maintainers who qualify. Limits, sign-up status and claim steps.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 13, 2026 · 7 min read

GitHub Copilot is free in three ways in 2026: the Free plan, the Student plan for verified students, and Copilot Pro for teachers and maintainers of popular open source projects who qualify. Copilot Free gives 2,000 completions and 50 chat requests a month. Pro lists at $10 a month, so a free Pro grant is worth $120 a year at that price.

One caveat comes before the steps. GitHub's June 1 changelog paused new sign-ups for the Student, Pro, Pro+ and Max plans and said sign-ups would reopen "in the coming weeks", with no date. The plans page we read on October 10 shows sign-up buttons for Free, Pro, Pro+ and Max and no pause notice, so the pause may have ended. The changelog does not say whether it covers the free Pro grant for teachers and maintainers. Confirm the sign-up status on GitHub before you plan around a route.

## Which free route fits you

Start with the row that matches your account. The limit column is what the route caps, and the claim column is where GitHub sends you.

You are | Route | What you get | Monthly limit | Claim it at |

Any individual developer | Copilot Free | Auto-selected models, limited agent use | 2,000 completions and 50 chat requests, including Copilot Edits | [GitHub Copilot plans](https://github.com/features/copilot/plans) |

Verified student | Copilot Student | Unlimited code completions, auto-selected models | An allowance of AI credits, amount not published | [GitHub Education: students](https://github.com/education/students) |

Verified teacher | Copilot Pro, free | The full Pro plan | 1,500 credits, unlimited completions | [GitHub Education: teachers](https://github.com/education/teachers) |

Maintainer of a popular open source project | Copilot Pro, free | The full Pro plan | 1,500 credits, unlimited completions | [Free Copilot Pro access](https://docs.github.com/en/copilot/how-tos/manage-your-account/get-free-access-to-copilot-pro) |

On Free and Student you cannot choose the model. The plan picks the model for each request.

## Copilot Free: a small allowance you can use

Free gives 2,000 completions a month, which is about 67 a day in a 30-day month. The 50 chat requests come to fewer than two a day. GitHub's plans FAQ states both limits in one sentence. Agent mode is listed as limited in the plan table, and the page prints no cap for it.

Free is for individual developers. Anyone assigned a Copilot Business or Enterprise seat is excluded.

With a chat cap and limited agent use, Free is enough to judge the editor experience and to answer the occasional question. It will not carry a full day of agent work.

Count your own usage before you decide. At ten chat questions a day, the monthly chat allowance runs out in five days. Track one normal week of chat: if it stays under eleven requests, the monthly allowance should hold. Heavier users should look at the paid steps first.

If you outgrow it, the upgrade to Pro happens on the Copilot settings page. The paid tiers, the usage-based billing rules and the monthly cost scenarios are on our GitHub Copilot pricing page. This page covers only the free routes.

## Students: unlimited completions, with a sign-up question

Verified students get Copilot Student. It includes unlimited code completions, which the Free plan does not, so it suits someone who writes code for hours and rarely opens chat. Like Free, it runs on auto-selected models. Its allowance of AI credits is not printed as a number.

To verify, you must be at least 13 years old and enrolled in a degree or diploma-granting course. Proof is a school-issued email address, official dated proof of enrollment, or both. The student page sends you to a "Get verified" button in your education benefits settings.

The open question is the sign-up pause. The changelog that paused new Student sign-ups gives no reopening date, and the plans page we read in October lists Free, Pro, Pro+ and Max with no Student row. Before you plan around the route, confirm on GitHub's plans page that new Student sign-ups are open. If they are closed, Copilot Free is the fallback. Cursor's student options work differently, and its paid plans are priced in our Cursor pricing explainer. The Cursor student discount guide covers its free Hobby tier, and the Cursor review covers the product.

## Teachers and maintainers: free Pro, checked every month

GitHub's teacher page calls the benefit Copilot Pro, free for verified educators. You must be a currently employed educator at an accredited educational institution. Apply through the Teacher Application with your academic email and proof of current employment, such as a faculty ID or an employment verification letter. Rejections cite insufficient proof or a non-degree-granting institution. You can update your details and reapply.

The maintainer route covers maintainers of popular open source projects. GitHub's docs do not define "popular," so you find out by checking. Open your profile picture menu, choose Copilot settings, and look for a GitHub Copilot Pro page. If it appears, choose the option, set your Copilot use policies, and save. If you do not qualify, the page lists the paid plans or Copilot Free instead.

Both routes are rechecked every month. GitHub says you cannot cancel the plan once free access is granted, so read what you are accepting before you click. A granted Pro plan saves the $10 monthly list price, which is $120 a year if you would otherwise pay for it. The grant lasts only as long as the monthly check passes, so treat it as a benefit that can lapse, not a fixed discount.

## Gotchas before you claim

- Credits reset on the first of the month. Included credits go back to the full monthly amount at 00:00 UTC on the first day of each month, and unused credits do not carry over.
- Spending past the included credits needs a budget. GitHub's changelog says extra spending past the included credits needs an additional budget that you set.

## If none of the routes fits

Toolradar data: our AI coding ranking evaluated 156 tools in October 2026, and eight of its ten top picks have a free plan. Claude Code is one of the two without one. Its [Toolradar record](https://toolradar.com/tools/claude-code) and [Cursor record](https://toolradar.com/tools/cursor) are the next stop for tool detail. The Claude Code pricing explainer shows what its paid plans cost, and the Claude Code review covers the product.

If you need a paid step anyway, Pro at $10 a month is the entry point, and it carries the same 1,500 credits a granted teacher receives. Pro+ at $39 a month adds $70 of usage. Our GitHub Copilot pricing page runs the monthly cost scenarios, and our GitHub Copilot review gives the product verdict.

For a full field of options, start with the AI coding assistants ranking. If you are choosing between an editor and a plugin, Cursor against GitHub Copilot weighs the two side by side.

## FAQ

### Is GitHub Copilot free in 2026?

Yes, in limited form. Copilot Free includes a monthly allowance of completions and chat requests, with limited agent use. Verified students, teachers and some open source maintainers can also get free access, subject to GitHub's current sign-up status for the Student plan.

### How do students get GitHub Copilot for free?

Verify your student status through GitHub Education. You need to be 13 or older and enrolled in a degree or diploma-granting course, with a school-issued email or dated proof of enrollment. Verified students get the Copilot Student plan, which includes unlimited code completions. Check the sign-up status first.

### Can teachers get GitHub Copilot for free?

Yes, if you are a currently employed educator at an accredited educational institution. Apply with proof of current employment, such as a faculty ID or an employment verification letter, sent from your institutional email. Verified educators get Copilot Pro free, which includes 1,500 credits a month.

### Do open source maintainers get GitHub Copilot for free?

Some do. GitHub may grant free Copilot Pro to maintainers of popular open source projects, and it rechecks eligibility every month. Open Copilot settings from your profile menu to see whether you qualify. If you do not, the paid plans or Copilot Free are the options.

### How many completions does GitHub Copilot Free include?

2,000 a month, which is about 67 a day in a 30-day month. Free also includes 50 chat requests a month, including Copilot Edits, according to GitHub's plans FAQ.

### Can I choose the model on Copilot Free?

No. Free and Student use auto model selection only, so the plan picks the model for each request.

Cite this: Devshot, "How to Get GitHub Copilot for Free in 2026: Free Plan, Students, Teachers and Open Source", October 2026.

## How we checked these facts

Louis Corneloup, founder of Dupple and Toolradar, wrote this page. On October 10, 2026, we read GitHub's [Copilot plans page](https://github.com/features/copilot/plans), its [individual plans docs](https://docs.github.com/en/copilot/concepts/billing/individual-plans), its [usage-based billing docs](https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-individuals), the [June 1 billing changelog](https://github.blog/changelog/2026-06-01-updates-to-github-copilot-billing-and-plans/), GitHub Education's [student](https://github.com/education/students) and [teacher](https://github.com/education/teachers) pages, and the [free access docs](https://docs.github.com/en/copilot/how-tos/manage-your-account/get-free-access-to-copilot-pro). The 156-tool count comes from Toolradar's [AI coding ranking](https://toolradar.com/best/ai-coding), updated October 2026.

Eligibility and sign-up rules change, so read GitHub's claim page on the day you apply. Devshot covers the tools and bill changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# LaunchDarkly Alternatives 2026: Feature Flag Prices Compared

URL: https://devshot.co/blog/launchdarkly-alternatives
Type: blog
Published: 2026-10-13
Updated: 2026-10-13
Summary: LaunchDarkly costs about $833 a month at 100,000 users. For 10 developers, PostHog and Statsig are free for flags, and ConfigCat Pro is $110 a month.

Guide

## LaunchDarkly Alternatives 2026: Feature Flag Prices Compared

LaunchDarkly costs about $833 a month at 100,000 users. For 10 developers, PostHog and Statsig are free for flags, and ConfigCat Pro is $110 a month.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 13, 2026 · 10 min read

For 10 developers, 100,000 monthly users and 50 flags, LaunchDarkly's Foundation plan costs about $833 a month, billed yearly. PostHog and Statsig cost nothing for flags on their free tiers, and ConfigCat Pro is the cheapest paid plan at $110 a month. These are list prices from each vendor's pricing page, read on October 10, 2026.

Leaving LaunchDarkly mostly saves money when the bill grows with your users. LaunchDarkly does not charge per seat. It charges per 1,000 client-side monthly active users and per server connection, so a small team serving a large app pays the most. Toolradar data: Toolradar tracks 22 feature flag tools, and it lists LaunchDarkly as an alternative to 20 other tools as of October 2026. For the wider field, including Unleash, see the feature flag platforms ranking.

## Cost at the scenario

The scenario is 10 developers, 100,000 monthly active users, 3 server connections, 50 flags and 100,000 flag requests a month. The table shows monthly list cost, cheapest first.

Tool | Entry price | Cost at scenario | Free plan | What the bill counts | Watch for |

PostHog | Pay as you go, no base fee | Free | 1M flag requests and 1M product events a month, 1 project | Flag requests and events | $0.0001 per request above 1M |

Statsig | Pro, $150 a month | Free (flags only) | 2M metered events, unlimited flag checks, unlimited seats | Metered events, not seats | $0.05 per 1K events above 5M on Pro |

ConfigCat | Pro, billed monthly | $110 | 10 flags, 5M config downloads | Config downloads; seats and users unlimited | Pro caps the account at 100 flags |

GrowthBook | Pro, $40 per seat a month | $400 | Starter, up to 3 users | Seats | Google-only SSO on Pro |

Flagsmith | Start-Up, billed yearly | $500 on Scale-Up, billed yearly | 50,000 requests, 1 member | Team members and requests | $50 per extra member, yearly billing |

LaunchDarkly | Foundation, pay as you go | $833 billed yearly | Developer: 5 connections and 1K client-side users | Client-side users and server connections | $10 per connection beyond 5 |

PostHog and Statsig come out free because their free tiers absorb flag traffic at this size. The events they meter elsewhere are where their bills grow, and that is covered below.

## What the scenario assumes

- 10 developers. Seats cost money at GrowthBook and Flagsmith only. LaunchDarkly, ConfigCat, PostHog and Statsig charge nothing per seat.
- 100,000 monthly active users on a client SDK. LaunchDarkly bills every 1,000 of them. ConfigCat bills downloads, and PostHog and Statsig bill requests and events.
- 3 server connections in one production environment. That sits inside the first five connections LaunchDarkly includes on Foundation.
- Serverless and short-lived processes. They are billed by connection time, so a function-heavy stack changes the bill. The serverless platforms ranking covers where those workloads run.
- 50 flags. Only ConfigCat caps flags on its pricing page, at 100 on Pro.
- 100,000 flag requests a month, one per user. That is the low end for a client app that checks flags more than once a session, and request-metered plans scale with it.
- Excluded: product analytics, session replay and experiments that send their own events. Each platform meters those differently, and they can change the ranking.

At 100,000 users, PostHog's 1 million free requests cover 10 requests per user a month before anything is billed. Statsig's 2 million free events work out to about 20 per user a month at this size, and every metered event counts toward that, including the product events you log.

## How much does LaunchDarkly cost at 100,000 users?

LaunchDarkly's Developer plan is free, with unlimited seats, five server connections and 1,000 client-side users a month included. That covers a prototype or a small app, not a product with 100,000 users. Foundation is pay as you go, with no seat fees. Its pricing card lists $8.33 per 1,000 client-side monthly active users, and its plan table lists $10 per 1,000 for the same unit.

At the card rate, 100,000 users is 100 blocks of 1,000, which comes to $833 a month, or $9,996 a year, billed yearly. At the table rate the same users cost $1,000 a month. This page uses the card rate because it is the lower figure, and the gap matters. Every extra 1,000 users adds $8.33 to $10 a month, while adding developers adds nothing.

The three server connections sit inside the five that Foundation includes, so they add nothing here. Enterprise is quote-only, and Guardian, its release safety add-on, is priced separately.

## What is the cheapest LaunchDarkly alternative?

PostHog and Statsig are the cheapest for flags alone at this scenario. Both free tiers hold until your product events grow. Among the paid plans, ConfigCat Pro is the cheapest from four developers upward, because its seats and users cost nothing. Cloud free-tier limits are compared in cloud free tiers compared.

Below four developers the picture changes. Three developers can run GrowthBook's free Starter plan, which allows up to three users. Flagsmith's Start-Up plan is billed yearly and includes three members.

## Cost by team size

Developers | LaunchDarkly | PostHog | Statsig | ConfigCat | GrowthBook | Flagsmith |

3 | $833 | Free | Free | $110 | Free (Starter) | $40 (Start-Up, billed yearly) |

10 | Same | Free | Free | Same | $400 | $500 (Scale-Up, billed yearly) |

25 | Same | Free | Free | Same | $1,000 (Pro) | Quote only (Scale-Up holds 20) |

Monthly list cost at 100,000 users and 50 flags, in US dollars. Free means the allowance covers the scenario with no charge. LaunchDarkly and ConfigCat do not change with team size.

Two crossovers matter. GrowthBook Pro passes LaunchDarkly at 21 developers, at $840 against $833, because its price rises with every seat. Flagsmith's Scale-Up stays under LaunchDarkly up to 16 developers, at $800, and passes it at 17, at $850.

## Is GrowthBook cheaper than LaunchDarkly?

Yes, until the seat count catches up. GrowthBook Pro charges $40 per seat a month for up to 30 users, so 10 developers pay $400, under half of LaunchDarkly's bill at this scale. GrowthBook's Starter plan is free for up to three users, with unlimited flags and experiments.

Pro includes 2 million CDN requests a month, which is 20 per user at this scenario, and charges $10 for each extra million. SSO on Pro is Google only. SAML and SCIM sit on the quote-only Enterprise plan.

## Vendor by vendor: what PostHog, Statsig, ConfigCat and Flagsmith charge beyond the table

### PostHog

PostHog bills a flag request as one feature flag evaluation, from the client or the server. The first million a month are free, experiments bill on the same meter, and the next band costs $0.0001 a request. The free plan needs a close read. Past its limits, flags return a quota-limited default response instead of billing, while pay-as-you-go charges for usage above the allowance.

Product analytics is a separate meter, with 1 million events free and then $0.00005 per event in the next band. PostHog's free plan also includes 100,000 exceptions of error tracking. For dedicated tools, see the error tracking tools ranking and the Sentry review. Neither plan charges per seat. The paid plan adds six projects and seven years of data retention. Source: [PostHog pricing](https://posthog.com/pricing).

### Statsig

Statsig's Developer plan is free, with unlimited seats, unlimited flag and config checks, and 2 million metered events a month. The metering rule matters more than the free limit. Statsig charges for a flag exposure only when Metric Lifts are switched on and the flag is in a partial rollout. Exposures at 0% or 100% are not charged by default.

Logged events count toward the 2 million, so a product that sends many events uses the allowance quickly. Past it, Developer keeps gates and configs running but stops metric lift calculations. Pro is $150 a month with 5 million events included, then $0.05 per 1,000 events. Source: [Statsig pricing](https://statsig.com/pricing).

### ConfigCat

ConfigCat says its Forever Free plan has the same feature set as the paid plans. That holds for features but not for limits. The free plan allows 10 flags, two environments and 5 million config downloads a month, so 50 flags need Pro. Pro is $110 a month, with 100 flags, 25 million downloads and 100 GB of network traffic. It charges nothing for seats or monthly active users, and annual billing saves up to 10%, per the page. Smart, at $325 a month, removes the flag cap. Source: [ConfigCat pricing](https://configcat.com/pricing).

### Flagsmith

Flagsmith's free plan allows 50,000 requests and one team member, so this scenario outgrows it. Start-Up costs $40 a month billed yearly ($45 monthly) and covers 1 million requests and three members. Beyond three members, the plan is Scale-Up at $250 a month billed yearly ($300 monthly), with five members included. Up to 20 members cost $50 each a month on yearly billing ($60 monthly). Extra API calls cost $50 per million.

Above 20 members, Flagsmith quotes an Enterprise plan, which includes self-hosting. Source: [Flagsmith pricing](https://www.flagsmith.com/pricing).

## When LaunchDarkly is still the right answer

LaunchDarkly's case is governance more than price. Its Enterprise plan adds custom roles and teams, workflows with scheduling and approvals, and SCIM user provisioning. Guardian adds guardrail metrics and automatic rollback on top of Enterprise. If your release process depends on those controls, the cheaper tools above need extra work to match them. If it does not, price should decide.

Stay on the free Developer plan while you have 1,000 or fewer client-side users, and reprice before you pass that limit. Teams often connect flag changes to monitoring, so the best monitoring tools ranking is worth a look before you pick a replacement. If flags ship through a pipeline, the CI/CD tools ranking covers that side.

## Methodology

- Every price comes from the vendor's own pricing page, read on October 10, 2026, in US dollars. Annual figures are the monthly price times 12.
- The scenario and the arithmetic are ours. Each vendor is priced on its own billing unit, and the assumptions are listed above.
- Where a vendor lists two prices, both are shown: LaunchDarkly's card rate and table rate, and Flagsmith's yearly and monthly prices.
- Plans that need a sales quote stay out of the ranking.
- The first-party figure comes from Toolradar's live [feature flag ranking](https://toolradar.com/best/feature-flags), checked October 10, 2026.

## FAQ

### Is there a free alternative to LaunchDarkly?

Yes. PostHog's free plan covers 1 million flag requests a month, and Statsig's Developer plan covers unlimited flag checks with 2 million metered events. At 100,000 users and 50 flags, both cost nothing for flags alone. ConfigCat's Forever Free plan is free too, but it allows only 10 flags.

### How much does LaunchDarkly cost?

The Developer plan is free, with unlimited seats, five server connections and 1,000 client-side users a month. Foundation is pay as you go, at $8.33 per 1,000 client-side users on its card and $10 per connection beyond five. At 100,000 users, that is 100 blocks of 1,000 users at $8.33 each a month, billed yearly. Enterprise is quoted by sales.

### Is GrowthBook cheaper than LaunchDarkly?

For small teams, yes. GrowthBook Pro costs $40 per seat a month, so 10 developers pay $400 at this scenario. The crossover with LaunchDarkly's Foundation plan comes at 21 developers, where GrowthBook costs $840. Its Starter plan is free for up to three users.

### What is the cheapest paid feature flag tool at this size?

ConfigCat Pro, at $110 a month, for four or more developers at 100,000 users and 50 flags. Its price does not change with seats or users. Below four developers, Flagsmith's Start-Up plan, billed yearly, is cheaper.

### Does PostHog charge per seat?

No. Pay-as-you-go has no seat charge, and the free plan lists unlimited team members. The meters are flag requests and product events, and each has a free allowance of 1 million a month.

### Does the number of flags change the price?

Only on ConfigCat, whose Pro plan caps the account at 100 flags. LaunchDarkly, GrowthBook and Flagsmith list unlimited flags on the plans discussed here. PostHog and Statsig do not list a flag cap on their pricing pages.

Cite this: Devshot, "LaunchDarkly Alternatives 2026: Feature Flag Prices Compared", prices checked October 10, 2026.

Devshot covers the tools and changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cheapest Twilio Alternative 2026: SMS API Pricing Compared

URL: https://devshot.co/blog/cheapest-sms-api
Type: blog
Published: 2026-10-11
Updated: 2026-10-11
Summary: Bird is the cheapest of four US SMS APIs compared for 50,000 one-time codes a month, ahead of Telnyx, Plivo and Twilio. Prices checked October 10, 2026.

Guide

## Cheapest Twilio Alternative 2026: SMS API Pricing Compared

Bird is the cheapest of four US SMS APIs compared for 50,000 one-time codes a month, ahead of Telnyx, Plivo and Twilio. Prices checked October 10, 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 11, 2026 · 9 min read

For 50,000 US one-time codes a month, Bird has the lowest list rate of the four SMS APIs ranked here: $175.50 a month at $0.0035 per segment with one number, before carrier fees. Twilio, the default comparison point, costs $416.15 for the same volume. Telnyx comes second at $201.00 and adds no carrier fee on carriers outside its named list. Every price is in USD, checked on each vendor's own pricing page on October 10, 2026.

Vonage, Sinch, AWS SNS, Infobip and Bandwidth are left out. The reasons are in the section "Who is left out, and why", and none of their prices appear in the tables.

Toolradar data: the [Twilio alternatives list on Toolradar](https://toolradar.com/alternatives/twilio) names 8 tools: 5 direct alternatives and 3 in the same category, updated October 2026. Each vendor in the table also has its own tool page on Toolradar: [Twilio](https://toolradar.com/tools/twilio), [Plivo](https://toolradar.com/tools/plivo), [Telnyx](https://toolradar.com/tools/telnyx) and [Bird](https://toolradar.com/tools/bird).

## Cost comparison at 50,000 one-time codes

Every row uses the same scenario: 50,000 one-segment messages and one sending number a month, at list rate, with no volume discount.

Tool | Entry price per segment | Monthly cost | Monthly cost with AT&T fee | Free plan | Hidden costs |

[Bird](https://bird.com/en-us/pricing) | $0.0035 | $175.50 | $350.50 | No SMS allowance listed | Carrier fees not listed by carrier; setup and registration fees can apply |

[Telnyx](https://telnyx.com/pricing/messaging) | $0.004 | $201.00 | $376.00 | No SMS allowance; $0 platform fee | Carrier fees on a named list only; $1.00 per number |

[Plivo](https://www.plivo.com/sms/pricing/) | $0.0077 | $385.50 | $560.50 | Trial credits, amount not stated | Carrier fee on every route; $0.50 per long code |

[Twilio](https://www.twilio.com/en-us/sms/pricing/us) | $0.0083 | $416.15 | $591.15 | Free start, no card needed; no SMS allowance listed | $1.15 per leased number; $0.001 per failed message |

"Monthly cost" includes one number. The AT&T column adds the AT&T carrier fee explained below to each vendor's rate. Bird publishes no per-carrier fee table, so its AT&T figure comes from the US example on its own pricing page, which prices 10,000 segments with AT&T fees included at $70, or $0.007 each.

## The scenario, and why it is one segment

The buyer is a team sending sign-in codes, password resets or order confirmations, each short enough to travel as a single segment. Twilio and Bird bill per segment, Telnyx bills per message part and Plivo bills per SMS. One segment per message keeps all four on the same footing, and a message that needs two segments is billed twice.

At this volume the number rental is under 1% of every bill, so the per-message rate decides the order. No volume discount is applied. Telnyx's volume tiers start at 100 million message parts a month, far above this scenario. This comparison does not model the volume tiers at Bird or Twilio, so the ranking is a verdict for 50,000 messages, not for a program many times larger.

## The arithmetic, shown once

Twilio: $0.0083 per segment x 50,000 segments = $415.00, plus $1.15 for one leased long code = $416.15 a month. Bird: $0.0035 x 50,000 = $175.00, plus $0.50 for one US local number = $175.50. The gap is $240.65 a month, 58% less than Twilio's bill, or $2,887.80 a year.

## Carrier fees: the part the headline rate leaves out

Carrier fees are charged per segment on top of the vendor's rate, and they depend on who receives the message. AT&T charges $0.0035 on Twilio, Plivo and Telnyx alike. T-Mobile charges $0.0045 on all three. Verizon is $0.005 on Twilio and Plivo and $0.0045 on Telnyx, while U.S. Cellular is $0.005 on Twilio and Telnyx. Plivo groups U.S. Cellular with other carriers at $0.005, and Twilio charges $0.0042 on all other carriers.

Telnyx is the exception. It charges nothing on carriers outside its named list, so on that traffic its $0.004 rate is the whole cost. That is the one route where a vendor with a higher list rate can still come out cheaper for the same message.

Telnyx's own US example also prices carrier pass-through at a blended $0.0038 per part. At 50,000 messages that adds $190.00, for a total of $391.00 with the number. Bird's rate card shows carrier fees separately and does not list them by carrier, so its total is confirmed only for AT&T traffic.

## Where the ranking flips

Three things change the order at this volume.

Bird against Telnyx turns on the carrier. Bird's list rate is $0.0005 below Telnyx's, so on traffic where Telnyx charges no carrier fee, any Bird carrier fee above $0.0005 per segment makes Telnyx cheaper for that recipient. Bird's fees by carrier are not on its pricing page, so ask Bird for its per-carrier fees before routing traffic on that basis. On AT&T traffic Bird still wins, at $0.007 against Telnyx's $0.0075 per segment.

Twilio against Plivo is nearly a tie. Plivo's list rate is $0.0006 lower, which is about $30 a month at 50,000 messages. On traffic to carriers outside AT&T, T-Mobile, Verizon and U.S. Cellular, though, Plivo charges $0.005 and Twilio charges $0.0042. That puts Twilio at $626.15 a month and Plivo at $635.50 on that traffic, so the order reverses.

Volume is the third trigger. Telnyx's volume tiers start at 100 million message parts a month, and this comparison leaves out Bird's and Twilio's. Read the order as a verdict for this scenario only.

## Which one to pick at this volume

Pick Bird when the list rate decides it and its carrier fees check out on your own recipients. It has the lowest base rate in the table, and at AT&T traffic it still totals less than Telnyx. The risk is that its pricing page does not list carrier fees by carrier, so you cannot see its cost on every route before you sign.

Pick Telnyx when a large share of your traffic goes to carriers outside its named list. On that traffic it charges only the $0.004 list rate, and it publishes a per-carrier table, so you can check the rest of its costs before committing.

Pick Plivo if your main goal is to move off Twilio for a modest saving. It is the closest match to Twilio's price, at about $30 a month less on list rate, and it comes out cheaper than Twilio on AT&T, T-Mobile and Verizon traffic. Twilio keeps the edge on traffic to carriers outside those names, so check your recipient mix first.

## Twilio's hidden lines

Twilio's headline rate is the number people quote, but other lines sit on its invoice. Its leased long code costs $1.15 a month, or $0.50 if you bring your own number. Each message that ends in a Failed status adds a $0.001 processing fee. US application-to-person texts also need 10DLC registration, and Twilio says those registrations carry onboarding fees without giving the amount on its rate card.

Bird says setup and registration fees can apply, and none of the four totals in the table includes a registration charge. Budget for those fees with whichever vendor you pick, and ask each vendor for the exact amount before you compare totals.

## Who is left out, and why

Five vendors are not in the table, and none of their prices appear on this page.

- Vonage: this page has no verified US SMS rate to compare.
- Sinch: its public SMS rate card lists per-message prices without naming a country, so they cannot be attributed to the US.
- AWS SNS: its pricing page lists 10DLC fees, $4 one-time per company and $1 a month per 10DLC number, while its per-message US rate sits in a separate per-country table on the [AWS SNS SMS pricing page](https://aws.amazon.com/sns/sms-pricing/).
- Infobip: its SMS rates sit behind a per-country estimator, and its pricing page shows no US figure.

Bandwidth is the fifth name left out, because this page ranks only vendors whose US rate it verified on their own pricing page.

## How we checked these prices

Devshot checked each vendor's own pricing page on October 10, 2026: [Twilio](https://www.twilio.com/en-us/sms/pricing/us), [Plivo](https://www.plivo.com/sms/pricing/), [Telnyx](https://telnyx.com/pricing/messaging) and [Bird](https://bird.com/en-us/pricing). Prices are list rates in USD, per segment or message part, and none is a quote. Totals are computed for 50,000 one-segment messages and one number, with carrier fees shown separately because they depend on the recipient. Devshot's editorial rules are on the about page.

## Where to read next

For the rest of a developer's bill, see cheaper Vercel alternatives, object storage pricing compared and cloud free tiers compared. For hosting and backend costs, read Supabase pricing, Vercel pricing, best serverless platforms and best backend-as-a-service tools. For API tools in general, see free LLM APIs, best text-to-speech APIs and best API gateways.

## FAQ

### What is the cheapest SMS API in 2026?

Among the four SMS APIs ranked here, Bird has the lowest list rate. At 50,000 US one-time codes a month, it costs $175.50 with one number, before carrier fees. Telnyx comes second at $201.00. Both prices were checked on October 10, 2026.

### How much does Twilio cost per SMS in the US?

Twilio's US long-code rate is $0.0083 per outbound segment, plus a carrier fee that depends on the recipient's carrier. A leased long code adds $1.15 a month. At 50,000 messages, the list-rate total is $416.15 a month, checked on October 10, 2026.

### Is there a free SMS API?

None of the four rate cards lists a free, ongoing SMS allowance. Plivo gives new accounts trial credits, and Twilio offers a free start with no card required, but neither page states the amount. Bird and Telnyx list no SMS allowance, so budget for every message from the first one.

### What is a cheaper alternative to Twilio for US texts?

Bird and Telnyx both list US SMS rates below Twilio's, and Plivo is the closest, about $30 a month less at 50,000 messages. Bird's total depends on carrier fees it does not publish by carrier, so ask Bird for those fees before you switch.

### Why does the carrier fee change the ranking?

Carrier fees are charged per segment on top of the vendor's rate, based on who receives the message, so they change the total for the same traffic. Each $0.001 per segment adds $50 a month at 50,000 messages. Telnyx charges nothing on carriers outside its named list, which can make it cheaper than a vendor with a lower base rate on that traffic.

### Does Plivo cost less than Twilio?

On list rate, yes: $0.0077 per segment against Twilio's $0.0083, about $30 a month less at 50,000 messages. The order can reverse on traffic to carriers outside AT&T, T-Mobile, Verizon and U.S. Cellular, where Twilio's fee is lower.

Cite this: Devshot, "Cheapest Twilio Alternative 2026: SMS API Pricing Compared", October 2026.

Subscribe to Devshot at devshot.co for the developer bills that change.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cursor Student Discount 2026: Is Cursor Still Free for Students?

URL: https://devshot.co/blog/cursor-student-discount
Type: blog
Published: 2026-10-11
Updated: 2026-10-11
Summary: Cursor closed its student discount to new sign-ups on June 25, 2026. The free Hobby plan remains, and undergraduates can claim campus credits this fall.

Guide

## Cursor Student Discount 2026: Is Cursor Still Free for Students?

Cursor closed its student discount to new sign-ups on June 25, 2026. The free Hobby plan remains, and undergraduates can claim campus credits this fall.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 11, 2026 · 8 min read

No, not as a discount. Cursor discontinued new sign-ups for its legacy student discount on June 25, 2026, according to its own help page, checked October 10, 2026. The Hobby plan is still free with no credit card, undergraduates can claim campus credits starting this fall, and Pro lists at $20 a month with no separate student price.

## What a student can get right now

Start with the row that matches you. The price column is the one to check before you sign up, because the student routes do not publish dollar amounts.

You are | What you can get now | Price | Source |

Any student | Hobby: Agent, Chat and Tab on the Auto model, Composer access, limited Agent requests | Free, no card | [Cursor pricing](https://cursor.com/pricing), [pricing help](https://cursor.com/help/account-and-billing/pricing) |

Undergraduate | Credits and discounts at on-campus and online events | Not published | [Cursor help](https://cursor.com/help/account-and-billing/student-discount) |

Master's, PhD, researcher or educator | Credits through Cursor's request form | Not published, no approval terms | [Cursor help](https://cursor.com/help/account-and-billing/student-discount) |

Redeemed the discount before the closure | Your current rate until the plan expires | Regular monthly rate after that | [Cursor help](https://cursor.com/help/account-and-billing/student-discount) |

Anyone who needs Pro now | Pro, the individual paid plan | $20/mo list | [Cursor pricing help](https://cursor.com/help/account-and-billing/pricing) |

Most students should start on Hobby today. Move to a paid plan only after a real week of work runs into the free limits, and treat the campus credits as a bonus rather than a budget line.

## What changed when the discount closed

Cursor gives two reasons for the change. The program had become a target for fraud, and it kept Cursor from reaching students in more countries. The routes that replace it are campus events and a request form, both covered below. Nothing on Cursor's help page suggests the old sign-up path will return.

If you already redeemed the discount, your rate holds until the plan expires. After the student discount year, Pro continues at the regular monthly rate unless you cancel. That renewal date is the number to watch.

Some guides still describe the old year-long offer without mentioning the closure. Trust Cursor's dated help page over any guide that gives no date.

## Undergraduates: campus credits, with no amount yet

Cursor says undergraduates can claim credits and discounts at on-campus and online events, and its students page says those events start "this fall." Neither page publishes a credit amount, a list of events, or a date for each one. There is no number to plan around yet, so do not hold off on a project waiting for one.

The practical step is to join Cursor's campus newsletter. Cursor says the list gets early access to events, exclusive discounts and beginner guides. If your term starts before the first event near you, Hobby covers the gap, and you do not need to pay for Pro to begin.

## Graduate students, researchers and educators: the request form

Master's students, PhD candidates, academic researchers and educators can fill out Cursor's credit-request form to support their work. The help page gives no approval criteria, credit amount or response time. Treat the form as a chance, not a plan. Until a credit arrives, Hobby keeps you coding, and paying month to month for Pro is the fallback if you cannot wait.

## Hobby is free, with limits

Hobby needs no credit card. Cursor's pricing page lists limited Agent requests and access to Composer, and its pricing help article adds Agent, Chat and Tab completions on the Auto model. Cursor does not publish a request count for Hobby on any page we read, so the limit only shows up in use. Our Cursor review covers the product itself, separate from the plan question.

That is enough for a class project and a few evenings of edits. Heavier work, such as a thesis, a hackathon or a summer internship, is likely to reach the cap, and the choice then is Pro, on-demand usage, or a different free tool. Cursor's Start plan is sold only in India and priced in rupees, so it does not help a student anywhere else.

## What Pro costs a student, with the arithmetic

Cursor lists Pro at $20 a month, Pro+ at $60 a month and Ultra at $200 a month. It lists no separate student price, so the sums below use those list numbers. One four-month term on Pro is $80. A full year billed monthly is $240. The same term on Pro+ comes to $240, and on Ultra it comes to $800. Cursor's pricing page recommends Pro+ for daily agent users and Ultra for agent power users, so the step up depends on how many hours a week you run agents.

Pro includes two monthly usage pools, one for Cursor's own models and one for third-party models. Both reset with the billing cycle, and unused usage does not roll over. Turning on on-demand usage bills extra requests in arrears at the same API rates, so the list price is the floor of the invoice, not the total. Our Cursor pricing breakdown shows how the pools and on-demand charges add up.

You can move from monthly to yearly billing in the dashboard. Cursor's help page says there is no way to switch back to monthly in the middle of a plan. The yearly price is not on the pages we read, so this page does not put a figure on it.

## The traps that cost students money

The biggest trap is a discounted code sold by someone else. Cursor says it sells subscriptions only through cursor.com, and that subscriptions bought from any other source are unauthorized and may be suspended or terminated.

The second trap is the renewal date for anyone still on the old discount. The rate holds until the plan expires, then the regular monthly rate applies. Put that date in your calendar and decide before it arrives.

The third trap is on-demand usage. A student who runs long agent sessions can see extra charges even on Pro. Keep on-demand off until you have seen one full cycle of your own usage on the Spending dashboard.

## What to do this week

- Install Cursor from [the download page](https://cursor.com/download) and use Hobby for the coding you already do. A week of real work shows whether the free limits matter for your course or project.
- If you are an undergraduate, join the [campus newsletter](https://anysphere.typeform.com/to/Xq2p21jK) so you hear about the fall events first.
- If you are a master's student, PhD candidate, researcher or educator, fill out the [credit-request form](https://anysphere.typeform.com/to/bgTJGDeA) and keep working on Hobby until you hear back.
- If you already hold the discount, check when your plan renews and decide before then whether Pro is still worth the monthly rate.

## Free AI coding tools if Cursor's free tier is too small

Toolradar data: our October 2026 AI coding ranking evaluated 156 tools. Eight of its ten top picks have a free plan, and Cursor's is Hobby. Claude Code and Kluster AI are the two top picks without one, which matters if you plan to switch tools mid-term. The Claude Code review and the Claude Code against Cursor comparison cover that tool in more depth. For each tool, the [Cursor record](https://toolradar.com/tools/cursor) and the [Claude Code record](https://toolradar.com/tools/claude-code) on Toolradar are the next stop.

Gemini CLI, listed as open source, and Codium AI, with a Developer plan, are also free in that table. Check each vendor's page before you build a workflow around one, and start with our AI coding assistants ranking for the full field. Cursor against GitHub Copilot covers the choice between an editor and a plugin, which changes what a free tier actually buys you.

Our GitHub Copilot pricing page and its review cover GitHub's student route. Check GitHub's own plan page before you rely on that route. For GitHub's student route, the claim steps and sign-up status are in How to Get GitHub Copilot for Free in 2026. If you want to see what the paid tiers cost on the tool you are most likely to switch to, the Claude Code pricing explainer covers its plans and prices.

## FAQ

### Does Cursor still have a student discount in 2026?

No. Cursor's help page says it discontinued new sign-ups for the legacy student discount on June 25, 2026. Students who redeemed it before then keep their rate until the plan expires, and then the regular monthly rate applies unless they cancel.

### Is Cursor free for students?

Partly. The Hobby plan is free with no credit card and includes limited Agent requests and Composer access. Undergraduates can claim campus credits at events, and master's students, PhD candidates, researchers and educators can request credits through a form. Cursor publishes no credit amount for either route.

### How much is Cursor Pro for a student?

Cursor lists Pro at its list price and publishes no separate student price. A four-month term costs $80, and a full year at the monthly rate costs $240. Pro+ is $60 a month and Ultra is $200 a month.

### Can graduate students get Cursor for free?

Possibly. Cursor says master's students, PhD candidates, academic researchers and educators can request credits through its form, but it gives no approval criteria, amount or timeline. Plan on Hobby, or a paid month, until a credit arrives.

### Does the student discount still apply to existing subscribers?

Yes, if you redeemed it before June 25, 2026. You keep your current rate until the plan expires, then pay the regular monthly rate unless you cancel.

### Are student Cursor codes sold by third parties legitimate?

No. Cursor sells subscriptions only through cursor.com and calls any other source unauthorized, which can lead to suspension. Do not pay a third party for a Cursor student code.

## How we checked these facts

Louis Corneloup, founder of Dupple and Toolradar, wrote this page. We read Cursor's [student discount help page](https://cursor.com/help/account-and-billing/student-discount), its [students page](https://cursor.com/students), its [pricing page](https://cursor.com/pricing), its [pricing help article](https://cursor.com/help/account-and-billing/pricing) and its [usage doc](https://cursor.com/help/models-and-usage/usage-limits) on October 10, 2026. The Toolradar figures come from our [AI coding ranking](https://toolradar.com/best/ai-coding), updated October 2026.

Cursor changes its terms, so check the help page before you buy. The prices here are list prices, before taxes, which Cursor says are extra.

Cite this: Devshot, "Cursor Student Discount 2026: Is It Still Available?", October 2026.

Some offers on this page may be paid placements or contain affiliate links.

---

# GitHub Copilot Pricing 2026: Every Plan, Usage Billing and Real Monthly Cost

URL: https://devshot.co/blog/github-copilot-pricing
Type: blog
Published: 2026-10-11
Updated: 2026-10-11
Summary: GitHub Copilot pricing in 2026: Pro is $10/mo, Pro+ $39/mo, Max $100/mo, Business $19/seat. Usage now bills in AI credits at $0.01 each.

Guide

## GitHub Copilot Pricing 2026: Every Plan, Usage Billing and Real Monthly Cost

GitHub Copilot pricing in 2026: Pro is $10/mo, Pro+ $39/mo, Max $100/mo, Business $19/seat. Usage now bills in AI credits at $0.01 each.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 11, 2026 · 10 min read

Key facts7 plans compared: Free, Student, Pro, Pro+, Max, Business, Enterprise

- Updated: October 11, 2026
- Prices as of: October 2026

- 7 plans compared: Free, Student, Pro, Pro+, Max, Business, Enterprise
- Free (best for: Trying Copilot before paying): $0
- Student (best for: Students with a verified account): $0 for verified students
- Pro (best for: Completions plus occasional chat and agent): $10/mo
- Pro+ (best for: Agent mode on most workdays): $39/mo

GitHub Copilot Pro costs $10/mo, Pro+ is $39/mo, Max is $100/mo, and Free costs nothing, verified on GitHub's pricing page in October 2026. Since June 1, 2026, every plan bills in GitHub AI credits at $0.01 each, so the sticker is a floor and heavy agent use is what lifts the invoice. Business is $19 per seat and Enterprise is $39.

These figures come from GitHub's [plans page](https://github.com/features/copilot/plans), its [plans documentation](https://docs.github.com/en/copilot/concepts/billing/individual-plans) and its [usage billing docs](https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-individuals), all read in October 2026. The short version for a budget: code completions stay unlimited on paid plans, while chat, agent mode, the CLI and the cloud agent draw from a monthly pool of credits that resets on the first of the month and does not roll over.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 155 tools as of October 2026, and in the scenario table below Pro+ becomes cheaper than Pro once a month's usage passes $44. The product verdict lives in the GitHub Copilot review, and the directory record is [GitHub Copilot on Toolradar](https://toolradar.com/tools/github-copilot).

Plan | Price | Included monthly credits | Best for |

Free | $0 | Limited chat and agent usage, 2,000 completions | Trying Copilot before paying |

Student | $0 for verified students | An allowance of credits | Students with a verified account |

Pro | $10/mo | 1,500 credits ($15 of usage) | Completions plus occasional chat and agent |

Pro+ | $39/mo | 7,000 credits ($70 of usage) | Agent mode on most workdays |

Max | $100/mo | 20,000 credits ($200 of usage) | Heavy agent and CLI users |

Business | $19/seat/mo | 1,900 credits per seat, pooled | Teams that need admin controls |

Enterprise | $39/seat/mo | 3,900 credits per seat, pooled | Organizations that need the top governance tier |

## What changed with Copilot usage-based billing?

Before June 1, 2026, each model interaction cost one premium request unit and a multiplier was applied depending on the model. GitHub's own billing docs say that as of June 1, 2026, the cost of an interaction depends on the model and the number of tokens consumed instead. One AI credit equals $0.01, and rates follow each model's listed API price per million tokens.

Code completions and Next Edit Suggestions stay unlimited on paid plans, so a developer who mostly tabs through suggestions is never metered. What moved is the meaning of "included": you now get a pool of money-equivalent credits rather than a count of requests, which rewards light chat users and penalizes anyone who runs long agent sessions on a frontier model.

Code review picked up a second meter. GitHub's June 1 changelog says Copilot code review now consumes GitHub Actions minutes in addition to AI credits, so a team running it on every pull request should not assume the seat price is the whole cost. AI for code review covers the alternatives.

## How do AI credits turn into a bill?

A credit is a cent, and what each interaction burns depends on input, output and cached tokens, priced per model. GitHub's model table lists Claude Sonnet 5.5 at $2.00 per million input tokens and 10 dollars per million output tokens, while GPT-6 Luna is $0.10 input and $0.50 output. A million output tokens on Sonnet therefore burns 1,000 credits, which is two-thirds of a Pro allowance in one go.

That spread is why model choice, not seat choice, decides most overage: pinning a lightweight model for routine chat saves the pool for the occasions when an agent really needs a frontier model. Opus 5.5 is listed at $4.00 input and 20 dollars output per million tokens, double the Sonnet output rate, so a long agent run on it empties a Pro pool quickly.

When the pool runs dry you have three options, all from GitHub's docs: upgrade the plan, set a budget for additional usage, or wait for the reset. Every 1,000 extra credits you budget for cost ten dollars, so the budget you set is your ceiling on extra spend.

## Is GitHub Copilot free?

Yes, with limits. Copilot Free costs nothing and covers limited chat and agent usage plus 2,000 completions a month, and GitHub restricts it to developers who have no organizational access. That is enough to judge the editor experience, but it will not carry a week of agent work.

Two other free routes exist: Copilot Student is free for verified students, and GitHub says verified teachers and maintainers of popular open source projects may be eligible for free Pro. If you qualify for either, check that route before paying. The claim steps for each free route, and the current sign-up status, are in How to Get GitHub Copilot for Free in 2026.

GitHub's June 1 changelog also said new sign-ups for the Student, Pro, Pro+ and Max plans were paused while safeguards went in, with a reopening promised in the following weeks. The plans page now lists a sign-up button for each plan, but confirm availability at checkout rather than assuming it.

## Copilot Pro: the plan most people mean

Pro is the plan behind the quoted headline price. It includes unlimited completions and 1,500 credits a month, made of 1,000 base credits plus a 500-credit flex allowance. GitHub states that flex allotments may change over time, so treat the full allowance as a current benefit, not a contract.

Stay on Pro when completions are your main use and chat is an occasional question, because a developer who opens agent mode a few times a week will rarely strain 1,500 credits. A daily agent user can drain the pool mid-month, which is where the next plan comes in. Cursor pricing shows how a rival handles the same trade-off.

## Pro+ and Max: when the pool is the product

Pro+ carries 7,000 credits, made of 3,900 base credits and a 3,100 flex allowance. Max carries 20,000 credits, split evenly between base and flex. GitHub launched Max on June 1 as an upgrade for existing Student, Pro and Pro+ subscribers, aimed at more intensive workflows and higher spending limits.

The math favors the bigger plans sooner than the labels suggest. Pro+ gives $70 of usage and Max gives $200, so each step up buys more credits per dollar than Pro does. The catch is that headroom only pays off if you spend it, since an unused allowance is forfeited at month's end.

## Copilot Business and Enterprise: per-seat math

Business includes 1,900 credits per user and Enterprise includes 3,900, each priced per granted seat. Credits are pooled at the billing entity level, so a heavy user can draw on teammates' unused allowance. Additional usage is enabled by default for organizations, at the same one cent per credit.

The detail that surprises finance teams is that organization seats carry no markup on credits: 1,900 credits is exactly $19 of usage, and 3,900 matches the Enterprise seat price too. The seat price is the allowance, and any usage beyond it adds dollar for dollar. Admins can cap that with user-level, cost-center, organization and enterprise budgets, which GitHub made generally available for user-level limits on June 1. For how Copilot's seat price compares with GitLab's Premium seat, see GitHub vs GitLab 2026.

Choose Business unless you need Enterprise's governance features, because the extra credits alone do not justify the jump. Both plans also bill non-licensed users' code review usage straight to the organization, so audit who triggers reviews.

## What does Copilot cost for a real month?

These scenarios apply GitHub's published allowances and the one-cent credit rate. The usage levels illustrate monthly model spend rather than measured averages, and they assume the full flex allowance is available and a budget is set for overage.

Monthly model usage | Pro | Pro+ | Max |

$30 (light agent use) | $25 | $39 | $100 |

$120 (daily agents) | $115 | $89 | $100 |

$250 (all-day agents) | $245 | $219 | $150 |

Pro+ beats Pro once usage passes $44, because Pro costs its base price plus whatever exceeds $15. Max beats Pro+ once usage passes $131. Below $44, the cheapest plan is the one whose allowance covers you.

A 10-seat team shows the organization side. Ten Business seats are $190 a month with 19,000 pooled credits, and the bill only rises if the team spends more than $190 of usage in total. Ten Enterprise seats double that. Per year, Business comes to $2,280 before any overage.

## How does Copilot compare with Cursor, Claude and Windsurf?

At the entry price Copilot is the cheapest paid option on this list, and the gap widens for teams. Cursor, Claude and Windsurf all start at $20/mo for an individual, while Copilot Pro costs half that. Entry price is not total cost, though, because each product bills a different way.

Tool | Entry paid plan | 10-seat team, per month |

GitHub Copilot | Pro, $10/mo | Business, $190 |

Cursor | Pro, $20/mo | Teams, $400 |

Claude (includes Claude Code) | Pro, $20/mo | Team Standard, $250 billed monthly |

Windsurf | Pro, $20/mo | Teams, $480 ($80 plus $40 per seat) |

Claude Code is a terminal agent included in paid Claude plans and shares one usage limit with the rest of the plan. For the head-to-head verdict, see Cursor vs GitHub Copilot, and for the terminal route read Claude Code pricing. The product pages are the Cursor review, the Windsurf review and the Claude Code review, and the broader shortlist is AI coding assistants. Directory records: [Cursor on Toolradar](https://toolradar.com/tools/cursor) and [Windsurf on Toolradar](https://toolradar.com/tools/windsurf).

## Is Copilot worth paying for?

For most individual developers, yes at the Pro level, because the entry price buys unlimited completions plus a usage pool that casual agent use rarely exhausts. The case weakens if you already pay for another assistant, because unlimited completions overlap with what rival tools bundle.

Move up when agent mode becomes a daily habit, which is the $44 crossover above, and stay on Free if you only need an occasional answer. Teams get the clearest value from Business, where pooled credits absorb uneven use across seats.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. Prices and allowances come from GitHub's [plans page](https://github.com/features/copilot/plans), [plans docs](https://docs.github.com/en/copilot/concepts/billing/individual-plans), [individual billing docs](https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-individuals), [organization billing docs](https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-organizations-and-enterprises), [model pricing table](https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing) and [June 1 changelog](https://github.blog/changelog/2026-06-01-updates-to-github-copilot-billing-and-plans/), read in October 2026.

Competitor prices come from [Cursor](https://cursor.com/pricing), [Claude](https://claude.com/pricing) and [Windsurf](https://windsurf.com/pricing) pricing pages read the same month, and the 155 tool count is Toolradar's live AI coding ranking. Scenario figures are our arithmetic on those published rates. More on how we publish is on the Devshot about page.

## FAQ

### How much does GitHub Copilot cost in 2026?

Free is $0, Pro is $10/mo, Pro+ is $39/mo and Max is $100/mo for individuals, verified in October 2026. Business is $19 per seat per month, and Enterprise costs the same per seat as Pro+ does per month. Usage beyond each plan's included credits bills at $0.01 per credit, so heavy agent users pay more than the sticker.

### How does Copilot usage-based billing work?

Since June 1, 2026, every plan bills in GitHub AI credits, where one credit equals $0.01, instead of premium requests. Chat, agent mode, the CLI and the cloud agent consume credits based on the model and tokens used, while code completions stay unlimited. Included credits reset on the first of each month and do not carry over.

### Is GitHub Copilot free?

Yes. Copilot Free has limited chat and agent usage and 2,000 completions a month, and Copilot Student is free for verified students. Verified teachers and maintainers of popular open source projects may be eligible for free Pro.

### What is the difference between Copilot Pro and Pro+?

Pro gives 1,500 credits a month, and Pro+ gives 7,000. Both include unlimited completions and access to the available models. Pro+ becomes the cheaper plan once your monthly usage passes $44.

### How much is Copilot Business versus Enterprise per seat?

Business has 1,900 pooled credits per seat, and Enterprise has 3,900. Choose Enterprise only if you need its governance features.

### What happens when I run out of Copilot credits?

You can upgrade the plan, set a budget for extra usage, or wait for the monthly reset. A budget of ten dollars adds 1,000 credits. Organizations have additional usage enabled by default, with budgets available to cap it.

### Is Copilot cheaper than Cursor?

Yes at entry level, since Cursor Pro is $20/mo against half that for Copilot Pro. For ten seats, Copilot Business is $190 and Cursor Teams is $400, so the gap compounds with headcount.

Cite this: Devshot, "GitHub Copilot Pricing 2026: Every Plan, Usage Billing and Real Monthly Cost", October 2026.

Devshot covers the tools and bill changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# OpenCode vs Claude Code 2026: Cost, Models and Which to Use

URL: https://devshot.co/blog/opencode-vs-claude-code
Type: blog
Published: 2026-10-11
Updated: 2026-10-11
Summary: OpenCode is free and MIT-licensed, while Claude Code costs $20 to $200 a month. Compare list prices, a monthly token scenario, login rules and model choice.

Guide

## OpenCode vs Claude Code 2026: Cost, Models and Which to Use

OpenCode is free and MIT-licensed, while Claude Code costs $20 to $200 a month. Compare list prices, a monthly token scenario, login rules and model choice.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 11, 2026 · 9 min read

Short answer: the OpenCode software costs nothing, so your bill is the model you choose. Claude Code costs $20 a month on Pro and $100 to $200 on Max, and it runs only Anthropic's Claude models. For a flat price on Claude, pick Claude Code. For model choice and per-token billing, pick OpenCode.

Most comparisons stop at the sticker price, and here the sticker misleads. On a Claude plan, Claude Code runs under usage limits that Anthropic sells as multiples of Pro usage, and no token allowance is published. OpenCode charges nothing for the software and passes the model's price through to you. So this page prices one month of agent work both ways, using the list prices each company published on October 10, 2026.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) evaluates 156 tools, updated October 2026. Both products have a listing there: [Claude Code](https://toolradar.com/tools/claude-code) and [OpenCode](https://toolradar.com/tools/opencode). For the wider field, start with AI for developers and the best AI coding assistants guide.

## OpenCode vs Claude Code at a glance

| OpenCode | Claude Code |

Software price | Free, MIT licensed | Included in a Claude plan |

Entry paid option | Go, from $10 a month | Pro, see the plan table below |

Heavy use | Go Plus, higher limits, or per-token billing | Max 5x and Max 20x, see the plan table below |

Models | 75+ providers through Models.dev, local models included | Anthropic's Claude models |

Subscription login | GitHub Copilot, ChatGPT Plus or Pro | Claude Pro, Max, Team or Enterprise |

Privacy position | Stores no code or context data (OpenCode's claim) | Consumer Terms for Free, Pro and Max, Commercial Terms for Team, Enterprise and API |

Where it runs | Terminal, desktop app, IDE extension | Terminal, IDE extensions, desktop app, web, GitHub, Slack |

## What each one costs

Claude Code is included in the Claude Pro, Max, Team and Enterprise plans. These are the list prices on [claude.com/pricing](https://claude.com/pricing) as of October 10, 2026. The Claude Code review covers the product itself, and the Claude Code pricing guide goes deeper on the plan grid.

Claude plan | Price | What changes |

Pro | $20 a month, or $17 a month billed annually ($200 up front) | Short sessions in small codebases |

Max 5x | $100 a month | Five times Pro's usage |

Max 20x | $200 a month | Twenty times Pro's usage |

Team seats add a per-seat price, and Team Premium gives each seat five times Standard's usage. Claude Code can also run on a Console account, which bills usage as pay-as-you-go API credits. That is the route OpenCode takes when you bring your own key.

OpenCode's options are a free download and several ways to pay for models, all listed on [opencode.ai](https://opencode.ai):

Option | Price | What you get |

OpenCode software | Free, [MIT licensed](https://github.com/anomalyco/opencode/blob/dev/LICENSE) | The agent itself, under the license in the repository |

[Go](https://opencode.ai/go) | $10 a month | Curated models, including Claude Haiku 5.5, with request estimates per five hours |

Go Plus | $40 a month | Higher limits, with everything in Go |

[Zen](https://opencode.ai/docs/zen/), pay as you go | Prepaid balance | Pay per request at list rates, zero markup beyond card fees |

Your own API key | The provider's rates | The provider bills the tokens directly, not OpenCode |

Two details change the real bill. Zen passes card fees through at cost, and its homepage shows a $1.23 fee on the starting balance. An auto top-up runs when the balance reaches $5, and you can set monthly spend limits. Go's limits are request estimates per five hours plus a monthly dollar amount, not token counts, so a Go plan cannot be converted into a per-token price.

## One month of agent work, priced both ways

We priced one developer's workload at list rates: 20 million input tokens and 2 million output tokens in a month, with no prompt caching. Anthropic's pricing page lists lower rates for cached input, so an agent that re-reads the same files pays less than these numbers. The table leaves caching out so that every row is priced the same way.

Each row is input tokens times the input price, plus output tokens times the output price. Sonnet 5.5 comes to 20 × $2 + 2 × $10 = $60.

Option | Rates per million tokens (input, output) | Monthly cost |

Claude Haiku 5.5 via API or Zen (prompts up to 100K tokens) | $0.10, $0.50 | $3.00 |

DeepSeek V4 Flash via Zen | $0.14, $0.28 | $3.36 |

Claude Sonnet 5.5 via API or Zen | $2, $10 | $60 |

Claude Opus 5.5 via API or Zen | $4, $20 | $120 |

Claude Code Pro, flat | Usage limits apply | $20 a month |

Claude Code Max 5x, flat | Usage limits apply | $100 a month |

Claude Code Max 20x, flat | Usage limits apply | $200 a month |

Three things follow. At this volume, Sonnet 5.5 billed per token costs $60, below the Max 5x plan, while Opus 5.5 at $120 costs more than Max 5x. Pro's $20 pays for about 6.7 million input and 0.67 million output tokens of Sonnet 5.5 at list rates, using a 10-to-1 input-to-output mix (10 million input plus 1 million output costs $30). The cheapest model in the table, Haiku 5.5 at $3.00, costs under a sixth of Pro's price, which is the cost case for OpenCode in one line.

The flat rows are not comparable token for token, because Anthropic publishes no token allowance for Pro or Max. The Sonnet and Opus rows are per-token bills, paid through the API or Zen, and they are not a substitute for a Claude plan, because the plan does not work inside OpenCode. The table says nothing about answer quality. Dupple has no test lab, so this page prices models and does not rank them.

## Can OpenCode use my Claude subscription?

Not as a login. OpenCode's homepage lists two subscription logins, GitHub Copilot and ChatGPT Plus or Pro, and no Claude subscription. Its Zen docs describe bringing your own API keys, billed by the provider.

Anthropic's rule comes from its own [Claude Code legal and compliance page](https://code.claude.com/docs/en/legal-and-compliance). Subscription sign-in is "intended exclusively" for purchasers of Claude Free, Pro, Max, Team and Enterprise plans, and it is "designed to support ordinary use of Claude Code and other native Anthropic applications." Developers building products must use API key authentication, and third-party developers may not route requests through Free, Pro or Max plan credentials on behalf of their users. The same page confirms that you can sign in to the unmodified Claude Code program with your own subscription.

In practice, a Claude model in OpenCode is a metered API bill, and a Max plan does not cover it. Anthropic's terms set this, and they change, so check the legal page before you rely on it.

## Models and where each one runs

OpenCode connects to 75 or more providers through Models.dev, including local models, according to its homepage. Claude Code runs Anthropic's Claude models only. That difference drives most of the price gap. On OpenCode you can run a model at $0.14 per million input tokens next to one at $4 per million, and you choose per task. Claude Code keeps the whole workflow inside one vendor's lineup. If you are weighing the Cursor side of the same question, our Claude Code vs Cursor comparison and the Cursor pricing page cover it.

Claude Code runs in the terminal, as an extension for VS Code and JetBrains, in the Claude desktop app, on the web and on iOS and Android, and inside GitHub and Slack. OpenCode runs as a terminal interface, a desktop app and an IDE extension. Its homepage also says it loads the right language servers for the model automatically and runs several agents in parallel on one project.

## Privacy and terms

OpenCode says it does not store any of your code or context data, and it presents that as a design for privacy-sensitive environments. The model provider still receives each prompt, so that provider's policy governs what happens to the data. According to Anthropic's Claude Code legal page, Consumer Terms apply to Free, Pro and Max users, and Commercial Terms apply to Team, Enterprise and API users.

## Which one to use

- Claude at a flat price, mostly in the terminal: Claude Code on Pro for light use, Max 5x for daily work. Usage limits apply. The how to use Claude Code guide covers the setup that makes a flat plan go further.
- Model choice or per-token billing: OpenCode with your own API key, or Zen. Your bill follows the model, so check the scenario table first.
- The cheapest working setup, if the model fits the task: OpenCode with Claude Haiku 5.5 or DeepSeek V4 Flash, the two cheapest rows in the table, or the Go plan if its request estimates cover your work.
- You already pay for ChatGPT Plus or Pro, or GitHub Copilot: check whether OpenCode's login covers your use. GitHub's [plans page](https://github.com/features/copilot/plans) lists Copilot Pro at $10 a month and Pro+ at $39 a month, and our GitHub Copilot pricing and GitHub Copilot review cover the product. For the OpenAI side, see the OpenAI Codex review.

## Gotchas

- The Zen balance is prepaid. Card fees are passed through at cost, and the homepage example shows a $1.23 fee on the starting balance.
- Haiku 5.5 costs more above 100K-token prompts. Zen lists $0.50 input and $2.50 output for prompts over 100K tokens. The scenario uses the lower tier, so long agent sessions cost more than the table shows.
- Team seats add up. Five Team Standard seats cost $125 a month billed monthly, or $100 billed annually.
- Caching lowers the Claude bill. The scenario table leaves it out on purpose, so treat those figures as a ceiling.

## FAQ

### Is OpenCode free?

Yes. The software is free and MIT licensed. You pay for the model, either through a Zen balance billed per request, the Go plan, or your own API key billed by the provider.

### Is OpenCode cheaper than Claude Code?

It depends on the model. At 20 million input and 2 million output tokens a month, Claude Haiku 5.5 costs $3.00 and Sonnet 5.5 costs $60 at list rates. Pro and Max 5x are flat monthly plans, but they only work inside Claude Code.

### Can I use my Claude Pro or Max subscription in OpenCode?

No. OpenCode lists GitHub Copilot and ChatGPT Plus or Pro logins, and Anthropic's legal page says Pro and Max credentials are for its own applications. A Claude model in OpenCode is billed per token through an API key.

### Does OpenCode support Claude models?

Yes, through an API key or Zen. Zen lists Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 at the same input and output prices as Anthropic's pricing page.

### Which costs less for a team of five?

Five Claude Code Team Standard seats cost $125 a month billed monthly, or $100 billed annually. OpenCode has no seat fee, so a team using it pays only for model usage. The scenario table prices one developer's month.

## How we checked the prices

We checked the prices on October 10, 2026, on opencode.ai, its Zen and Go pages, claude.com/pricing and the Claude Code product page, Anthropic's Claude Code legal and compliance page, GitHub's Copilot plans page, and the MIT license in OpenCode's repository. The arithmetic uses list rates and no caching, and each row shows its inputs. No model was run for quality, because Dupple has no test lab.

Cite this: Devshot, "OpenCode vs Claude Code 2026: Cost, Models and Which to Use", October 2026.

Some offers on this page may be paid placements or contain affiliate links.

---

# Claude Code vs Cursor 2026: Which Is Better for Developers?

URL: https://devshot.co/blog/claude-code-vs-cursor
Type: blog
Published: 2026-10-10
Updated: 2026-10-10
Summary: Claude Code vs Cursor in 2026: terminal agent or AI editor, with current plan prices, a 10-seat team cost, usage limits, and a verdict by how you work.

Guide

## Claude Code vs Cursor 2026: Which Is Better for Developers?

Claude Code vs Cursor in 2026: terminal agent or AI editor, with current plan prices, a 10-seat team cost, usage limits, and a verdict by how you work.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 10, 2026 · 11 min read

Short answer: Claude Code wins for developers who work in the terminal and want an agent that edits a whole repo, runs commands, and handles git. Cursor wins if you want AI built into an editor you work in all day, with team controls on top. On price, a solo developer pays the same $20 a month for either entry plan on monthly billing.

The editor question matters less than it looks. Anthropic publishes a Claude Code extension that installs in Cursor as well as VS Code, so you can run the agent inside Cursor's window if you prefer it. The real split is where the agent lives by default: in a terminal session or the Claude desktop app for Claude Code, and inside the editor for Cursor.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) we run evaluates 156 tools as of October 2026. Both products sit in that set, with listings for [Claude Code](https://toolradar.com/tools/claude-code) and [Cursor](https://toolradar.com/tools/cursor), and Cursor is the only one of the two that lists a free plan there, which matches the vendor pages.

## Quick verdict

Dimension | Claude Code | Cursor |

Where it runs | Terminal, VS Code and JetBrains extensions, desktop app, web | Its own editor, which you install and work in |

Entry paid plan | Pro, from $17 a month billed annually | Pro at the monthly price; no annual price printed |

Free tier | None for Claude Code; Claude's Free plan excludes it | Hobby, free with limited agent requests |

Heavier individual plan | Max, from $100 a month | Pro+ at $60 a month, Ultra at $200 a month |

Team seat | Standard, billed annually; Premium is the higher-usage tier | Teams at $40 a user a month billed monthly, or $32 billed annually; Premium at $120, or $96 billed annually |

Models | Claude models; the terminal, VS Code and JetBrains can also point at third-party providers | OpenAI, Anthropic, Google and others, plus Cursor's own Composer 2.5 |

Usage limits | Rolling five-hour windows and weekly caps, shared across Claude apps | Two monthly pools; third-party models draw from one at API rates |

Best for | Terminal-first work and changes that end in a pull request | Developers who want the AI inside their editor |

## Where each tool lives

Claude Code runs in four places. You can use it in your terminal, as an extension in VS Code or JetBrains (the JetBrains plugin needs the CLI installed separately), in a desktop app that requires a paid subscription, or on the web. Every surface reads the same project settings, so a CLAUDE.md file you write once works in all of them. Our Claude Code review and Cursor review cover each product on its own.

Cursor's agent has one home: its own editor, where it works on the files you have open. That is the right trade if you want the AI to feel like part of the editor. It is the wrong one if your team is standardized on JetBrains, because the Claude Code plugin keeps you in that IDE while Cursor asks you to move.

## What one developer pays

On list prices, a single developer on an entry plan pays $240 a year on Cursor Pro billed monthly. Claude Pro billed monthly costs the same $240, but Claude Pro billed annually comes out $40 cheaper for the year, paid up front. Cursor's Pro price is listed by the month only, so the annual comparison runs in Claude's favor. Our Cursor pricing and Claude Code pricing explainers list each vendor's full plan set. This page compares only the entry and heavier tiers.

Plan | Price | Per year | What it adds |

Cursor Pro | $20 a month | $240 | Extended agent limits, cloud agents, frontier models |

Claude Pro, billed annually | $17 a month | $200 | Claude Code included, billed up front |

Claude Pro, billed monthly | $20 a month | $240 | Claude Code included |

Cursor Pro+ | $60 a month | $720 | Higher individual usage than Pro |

Claude Max | From $100 a month | $1,200 | Five times Pro usage; the page does not price the 20x tier |

Cursor Ultra | $200 a month | $2,400 | The top individual tier |

The first step up is where the two diverge. Cursor Pro+ costs $60 a month, while Claude Max is the pricier first step. Claude's step is a fixed multiple of Pro usage, so it suits someone who already knows the Pro allowance runs out. Cursor's page does not print a usage multiplier for Pro+, so judge that step by the plan you would actually use.

## What a 10-seat team pays

Ten developers on team plans is where the gap grows on paper. Cursor Teams costs $40 per user a month billed monthly, or $32 billed annually. On monthly billing, ten seats are $400 a month, or $4,800 a year. Claude Team Standard is cheaper per seat when billed annually. Billed monthly, the same seats cost $25 each, or $3,000 a year for ten.

Seat type | Price per seat | 10 seats per month | 10 seats per year |

Claude Team Standard, billed annually | $20 | $200 | $2,400 |

Claude Team Standard, billed monthly | $25 | $250 | $3,000 |

Cursor Teams Standard, monthly rate | $40 | $400 | $4,800 |

Claude Team Premium, billed annually | $100 | $1,000 | $12,000 |

Claude Team Premium, billed monthly | $125 | $1,250 | $15,000 |

Cursor Teams Premium, monthly rate | $120 | $1,200 | $14,400 |

Read the gap with care. Cursor's rows use its monthly seat rate (its annual Standard seat is $32 a month, and Premium is $96 a month), and Claude's annual rate is its cheapest option, so compare that row first. Claude's team plan also runs from 2 to 150 seats, and Claude Team Premium gives each seat five times the usage of a Standard seat. The cheaper seat only wins if its allowance covers the work your team really runs.

## Usage limits are where the bill surprises people

Claude's limits reset on a rolling five-hour session window, and paid plans add weekly limits on top. The web app, the desktop app, mobile, and Claude Code all draw from one pool. A long agent session in the terminal can therefore eat the allowance you planned to use for chat.

Cursor uses two monthly pools. The Cursor Models pool covers its own models, and the Other Models pool covers third-party models, which bill at each model's API rate. Your model choice sets how fast the included usage runs down, and once both pools are gone, on-demand usage bills at the same API rates. If you keep a frontier model selected all day, read the usage page before you pick the cheapest plan.

There is a third route for Claude Code. With a Claude Console account, it consumes API tokens at standard API rates, which swaps the subscription allowance for a per-token bill. Pick it when you already know your daily token volume, because a per-token bill punishes guesses.

## Models and what you can switch between

Claude Code is built around Anthropic's Claude models. Anthropic's docs also say the terminal, VS Code and JetBrains surfaces can point at third-party providers, which helps if one job needs a model outside the Claude family.

Cursor lets you choose among models from OpenAI, Anthropic, Google and others, and it ships its own Composer 2.5 model. If you want to compare a Claude model against a GPT model on the same task inside one tool, Cursor is the simpler setup. Our ChatGPT vs Claude for coding breakdown covers how the underlying models differ on programming work.

## Working across a repo

Both tools handle changes across many files, and this is where the choice starts to matter. Claude Code works directly with git. It stages changes, writes commit messages, creates branches, and opens pull requests, so a single task can run from a failing test to a reviewable diff. You can run several sessions side by side in the desktop app and review each diff visually before you commit anything. For the setup side, see our how to use Claude Code guide.

Cursor's agent runs inside the editor, which keeps the feedback loop short when you want to watch each edit land in a file. Its paid plans include cloud agents, and the Teams plan adds Bugbot code reviews.

The pattern follows from the design. A developer who wants one agent to own a ticket from start to pull request will find Claude Code the more direct tool. A developer who wants to stay in the editor, see every change as it happens, and switch models per task will find Cursor fits better.

## Team controls and admin

Cursor's Teams plan bundles the admin layer in one place: centralized billing, team-wide privacy mode, SAML and OIDC single sign-on, usage analytics, and Bugbot code reviews. For a team that wants one vendor for the editor and the review step, that bundle is the real reason to pay the higher seat price.

Claude's team plans are simpler on the seat side. You can mix Standard and Premium seats in one team and pool up to $500 a month of API credits across members. Enterprise bills the seat plus usage at API rates and lets admins set spend limits per user and per organization, so finance will want the usage forecast before the contract, not after.

## Free tiers

Cursor's Hobby plan is free and includes limited agent requests and access to Composer. That is enough to judge the editor before you spend anything. Claude's Free plan covers everyday chat, but it does not include Claude Code. To try the agent on Claude, you need at least Pro, so Cursor is the cheaper way to test an agent workflow.

## Winner by dimension

Dimension | Winner | Why |

Individual entry price | Claude, on annual billing | Claude Pro is $17 a month billed annually; Cursor Pro is priced by the month |

Free way to try the agent | Cursor | Hobby includes limited agent requests with no card needed |

Team seat price at 10 seats | Claude | Claude Team Standard is cheaper than Cursor Teams at either billing term |

Terminal and git workflow | Claude Code | Stages, commits, branches and opens pull requests from the CLI |

Editor-native experience | Cursor | The agent runs inside the editor where your files are open |

Model choice in one tool | Cursor | Models from several providers, switchable in the same interface |

Team admin | Cursor | SSO, usage analytics and Bugbot ship on the Teams plan |

## Who should pick which

Pick Claude Code if you already work in a terminal and a git repo, you want one agent that touches many files and opens the pull request itself, or you already pay for Claude Pro or a Team plan. The agent comes with the subscription, so there is no second line item to justify. For the Codex side of the same choice, our Codex vs Claude Code 2026 comparison covers plan prices, usage meters and API costs.

Pick Cursor if you want the AI built into the editor, you prefer to switch between model providers inside one tool, or you are buying for a team that needs SSO, usage analytics, and review tooling in one contract. If Copilot is also on your shortlist, our Cursor vs GitHub Copilot comparison covers that pair.

If you still cannot decide, spend a week on Cursor's free Hobby plan for everyday editing, then try Claude Code for the larger changes if you already pay for Claude. A week of real tasks settles the question faster than another comparison. If you want the wider field first, the AI-assisted coding primer and the best AI coding assistants ranking put both tools next to the alternatives.

Devshot sends one short email a day on the AI tools that change how developers ship. Subscribe here if you want the useful ones without the launch-day noise.

## How we ran the numbers

Prices come from the vendor pages read on October 10, 2026: [Cursor pricing](https://cursor.com/pricing), [Cursor's account pricing docs](https://cursor.com/docs/account/pricing), [Claude pricing](https://claude.com/pricing), and the [Claude Code product page](https://claude.com/product/claude-code). Feature and usage facts come from [Cursor's model docs](https://cursor.com/docs/models) and [Anthropic's Claude Code overview](https://code.claude.com/docs/en/overview). Every cost on this page is our arithmetic: list price times seats times months, with no discounts or promotions applied.

The Claude 20x tier is not priced as a plan on Anthropic's page, so this page shows Max only from $100 and leaves the 20x figure out. Cursor's rows use monthly prices. Vendor prices change, so check both pricing pages before you buy.

Cite this: Devshot, "Claude Code vs Cursor 2026: Which Is Better for Developers?", October 2026. Prices verified on vendor pages, October 2026.

## FAQ

### Is Claude Code better than Cursor in 2026?

Neither wins every task. Claude Code is the better fit for terminal-first work and changes that end in a pull request. Cursor is the better fit for developers who want the AI built into their editor and a choice of model providers in one tool. Choose by where you work, then check the price against your seat count.

### Which is cheaper, Claude Code or Cursor?

For one person, the entry prices sit close together. Claude Pro is $17 a month billed annually and includes Claude Code, while Cursor Pro is priced by the month. For ten people on team plans, Claude Team Standard billed annually costs $2,400 a year, against $4,800 for Cursor Teams at its monthly rate.

### Is there a free version of Claude Code or Cursor?

Cursor has one: its Hobby plan is free, with limited agent requests. Claude's Free plan does not include Claude Code, so trying the agent on Claude means paying for Pro at the least. Free allowances change, so check both pricing pages before you commit.

### Can I use Claude Code inside Cursor?

Yes. Anthropic publishes the Claude Code extension with an install link for Cursor. You still need a Claude subscription or an Anthropic Console account to use it, the same as in VS Code.

### Can Cursor use Claude models?

Yes. Cursor lists Anthropic models alongside OpenAI, Google and others, so you can pick a Claude model inside Cursor. Third-party models draw from Cursor's Other Models pool at each model's API rate, which is the line to check on your bill.

### Which is better for a team of ten?

On list prices, Claude Team Standard is the cheaper seat at either billing term. Cursor's team plan adds SSO, usage analytics and Bugbot code reviews, so compare the feature list as well as the seat. If your team runs heavy agent sessions every day, price the usage pools too, since the seat price does not set the cap.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cloud Free Tiers Compared (2026): AWS, Google Cloud, Azure, Oracle

URL: https://devshot.co/blog/cloud-free-tiers-compared
Type: blog
Published: 2026-10-10
Updated: 2026-10-10
Summary: Oracle's Always Free compute gives one server 2 Arm OCPUs and 12 GB with no end date. AWS credits close at six months, Google's trial at 90 days.

Guide

## Cloud Free Tiers Compared (2026): AWS, Google Cloud, Azure, Oracle

Oracle's Always Free compute gives one server 2 Arm OCPUs and 12 GB with no end date. AWS credits close at six months, Google's trial at 90 days.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 10, 2026 · 9 min read

For one always-on small server, Oracle Cloud's Always Free compute costs $0 with no end date: 2 Arm OCPUs and 12 GB of memory for a full month, per Oracle's own docs checked October 10, 2026. Google Cloud's e2-micro also renews free each month, but only in three US regions, and its $300 trial ends at 90 days. AWS gives up to $200 over six months, and Azure gives $200 for 30 days plus a 12-month small VM.

Every vendor figure below was read on the vendor's own page on October 10, 2026, in US dollars. Free tiers change without notice, so check the vendor page before you open an account.

## The four free tiers at a glance

Provider | Free credit | Window | Still free afterwards | Free server | Source |

AWS | $100 at signup, up to $100 more | 6 months, or until the credit runs out | 30+ always-free services within monthly limits | None named on the free page | [AWS Free Tier](https://aws.amazon.com/free/) |

Google Cloud | $300 | 90 days | Free Tier monthly limits | One e2-micro a month in three US regions | [Free features](https://cloud.google.com/free/docs/free-cloud-features), updated October 7, 2026 |

Azure | $200 | 30 days | 65+ always-free services | B1s, B2pts v2 and B2ats v2 VMs, 750 hours each for 12 months (new customers) | [Azure free account](https://azure.microsoft.com/en-us/free/) |

Oracle Cloud | $300 | 30 days | Always Free, no time limit | 1,500 Arm OCPU-hours and 9,000 GB-hours a month, or up to two AMD micro VMs | [Oracle Free Tier](https://www.oracle.com/cloud/free/) |

Tool pages for the four clouds: [AWS](https://toolradar.com/tools/aws), [Google Cloud](https://toolradar.com/tools/google-cloud), [Azure](https://toolradar.com/tools/azure) and [Oracle Cloud](https://toolradar.com/tools/oracle-cloud).

## AWS: $200 over six months, then the account closes

New AWS accounts get $100 in credits at signup and can earn up to $100 more by exploring key services. AWS says the account closes on its own six months after you open it, or when the credits run out, whichever comes first. You are not charged unless you convert to a Paid plan, so the real deadline is the six-month window, not the first bill. AWS dates this model to July 15, 2025, when new customers began receiving the larger credit, according to its [S3 pricing page](https://aws.amazon.com/s3/pricing/).

The always-free list is real but narrower than the headline. AWS says 30 or more services stay free within monthly limits on both plans. [Lambda](https://aws.amazon.com/lambda/pricing/) is the clearest example, with 1 million requests and 400,000 GB-seconds of compute every month. The free page does not name an always-free virtual machine, so an AWS server draws on the credit until the account closes. That is the trap for anyone who plans a permanent server on AWS credits.

## Google Cloud: a 90-day trial, then one e2-micro

The trial lasts 90 days, and Google does not bill usage during it. If you upgrade early, you keep the unused credit until day 90 and the Free Tier carries on. If you do nothing, or spend the credit first, the trial billing account closes and all of its projects and resources are stopped. A trial server stops with the account, not at the end of a month, so move anything you need onto an upgraded billing account before day 90.

The lasting benefit is the Free Tier, and it is narrower. You get one non-preemptible e2-micro VM a month in us-west1 (Oregon), us-central1 (Iowa) or us-east1 (South Carolina), with 30 GB-months of standard disk and 1 GB of outbound data from North America. Google counts hours across all your e2-micro instances against the hours in the month, so one VM running all month stays free. Cloud Storage adds 5 GB-months of US regional storage, 5,000 Class A and 50,000 Class B operations, and 100 GB of egress. [Cloud Run](https://cloud.google.com/free/docs/free-cloud-features) adds 2 million requests a month.

## Azure: a 30-day credit, then 12 months of a small VM

Azure gives a credit to use within 30 days, and the clock is what bites. If you want to keep going, move to pay-as-you-go before the window closes. Azure says that without the move, your account and services are disabled until you upgrade. The list that matters for a server is the 12-month one: 750 hours each of B1s, B2pts v2 (Arm) and B2ats v2 (AMD) burstable VMs, for new customers. Azure Functions adds 1 million requests a month as an always-free item.

Always-free services keep their monthly amounts after the first year. The 12-month list does not, so an Azure VM that you keep running past month 12 starts billing at the pay-as-you-go rate.

## Oracle Cloud: the free server with no end date

Oracle's trial credit lasts 30 days or until it is used up, whichever comes first, and covers all eligible services. Always Free is a separate allowance with no time limit, as long as the account stays in use. After the trial, Always Free keeps running, and after an upgrade Oracle bills only for usage above the free amounts.

Compute is what matters for a server. Oracle's [Always Free docs](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm) give every tenancy 1,500 OCPU-hours and 9,000 GB-hours of Arm Ampere A1 compute each month, and say that equals 2 OCPUs and 12 GB of memory. Older guides quote 4 Arm cores and 24 GB, but the current docs give 2 OCPUs and 12 GB. The other option is up to two VM.Standard.E2.1.Micro AMD instances, which can only be created in one availability domain.

Every tenancy also gets 200 GB of block storage with five volume backups, 20 GB of combined object and archive storage once the trial has ended, and one Flexible Load Balancer at 10 Mbps.

## Does one server fit the free limits?

Take the common case: one server that runs all month, which is 730 hours. The table shows whether it fits the free allowance without any payment.

Provider and server | Free allowance | Used in 730 hours | Result |

Oracle Arm A1, 2 OCPUs and 12 GB | 1,500 OCPU-hours and 9,000 GB-hours a month | 1,460 OCPU-hours and 8,760 GB-hours | Fits, $0, no end date |

Google Cloud e2-micro | Hours equal to the month's total, 744 in October | 730 hours | Fits, $0, in one of three US regions |

Azure B1s | 750 hours for 12 months (new customers) | 730 hours | Fits, $0 for 12 months |

AWS EC2 | No always-free VM named | Draws on the credit | Runs until the account closes |

Oracle leaves 40 OCPU-hours and 240 GB-hours to spare. Going above 2 OCPUs and 12 GB for the whole month uses up the allowance, so a bigger VM is billed. Azure leaves 20 hours of headroom. After Azure's 12 months, a B1s on pay-as-you-go Linux in East US costs $0.0104 an hour, or $7.59 for 730 hours. Windows in the same region is $0.014 an hour, or $10.22 for the month. These rates come from [Microsoft's retail prices API](https://prices.azure.com/api/retail/prices).

## Which free VPS can you keep?

For a free server with no end date, Oracle's Always Free compute is the strongest option. It fits a full month of 2 OCPUs and 12 GB, and the allowance does not expire while the account stays in good standing. Google's e2-micro also renews each month, but it is one small VM in three regions, and it lives inside a billing account that stops if the trial closes without an upgrade.

Azure's VM is free for a year, which suits a learning project or a test server. AWS's free page names no always-free VM, so a server there runs on credit. Pick Oracle for a permanent server, Google if you already run on Google Cloud and can live with the region limit, and Azure for a one-year trial. For the paid VPS that follows a free year, compare the best VPS hosting for developers, Hetzner and DigitalOcean.

## Free plans are not free servers

Devshot data: of the 198 hosting and deployment tools Toolradar ranks, 142 (72%) have a free plan, per its [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment), updated October 2026. A free plan is not the same as a server you can keep running. The four clouds above cap different things first: credit, time, hours or requests. For the platform layer on top of a server, see Vercel pricing, Supabase pricing, cheaper Vercel alternatives and best serverless platforms.

## Gotchas with exact figures

- AWS: the account closes at six months or when the credits run out, whichever comes first.
- Google Cloud: the trial closes at day 90 and its projects and resources stop. The e2-micro is free in us-west1, us-central1 and us-east1 only.
- Azure: the account is disabled if you do not move to pay-as-you-go within 30 days. The 12-month VM list ends at month 12, and a B1s costs $7.59 a month after that in East US.
- Oracle: Oracle may place a temporary authorization hold on your card. An account left idle for 30 days or more can become eligible for suspension or termination.
- Oracle Arm: the allowance covers 2 OCPUs and 12 GB for a full month. Going above that size uses more than the monthly cap.

Once a free tier ends, storage and transfer are the next bill. The object storage pricing comparison shows what S3-style buckets cost once the free egress runs out, and best Postgres hosting covers the managed database side.

## How we checked the numbers

Each vendor figure comes from the vendor's own page, read on October 10, 2026: [AWS Free Tier](https://aws.amazon.com/free/), [AWS Lambda pricing](https://aws.amazon.com/lambda/pricing/), [AWS S3 pricing](https://aws.amazon.com/s3/pricing/), [Google Cloud free program](https://cloud.google.com/free) with its [free features doc](https://cloud.google.com/free/docs/free-cloud-features), [Azure free account](https://azure.microsoft.com/en-us/free/), [Oracle Cloud Free Tier](https://www.oracle.com/cloud/free/) and Oracle's [Always Free resources doc](https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm). The scenario uses 730 hours, the average month, and Oracle's own equivalence of 2 OCPUs and 12 GB. The Azure rates come from Microsoft's retail prices API for East US, consumption pricing. Free tiers change, so re-check the vendor page before you sign up.

## FAQ

### Which cloud has the best free tier in 2026?

For a permanent small server, Oracle Cloud. Its Always Free Arm compute gives 1,500 OCPU-hours and 9,000 GB-hours a month, which covers 2 OCPUs and 12 GB for a full 730-hour month with no end date. Google's e2-micro also renews monthly, but only in three US regions.

### Is the AWS free tier actually free?

Partly. New accounts get the signup credit plus a bonus for exploring key services, within six months at most. The account closes at six months or when the credit runs out, whichever comes first. The always-free services stay free within monthly limits on both plans, and Lambda gives 1 million requests and 400,000 GB-seconds a month.

### Is there a free VPS that does not expire?

Oracle's Always Free compute has no time limit: up to two AMD micro instances, or the Arm allowance of 2 OCPUs and 12 GB. Google's e2-micro renews monthly in us-west1, us-central1 or us-east1 while the account stays open. Azure's free VM runs for 12 months only.

### What does an Azure VM cost after the free 12 months?

A B1s costs $0.0104 an hour on pay-as-you-go Linux in East US, which is $7.59 for a 730-hour month. The same VM on Windows costs $0.014 an hour, or $10.22 a month.

### Does Google Cloud charge after the trial?

Not during the trial. If you upgrade, the Free Tier carries on and you pay only for usage beyond it. If you do nothing before day 90, or spend the credit first, the trial billing account closes and its projects and resources stop.

Cite this: Devshot, "Cloud Free Tiers Compared (2026): AWS, Google Cloud, Azure, Oracle", October 2026.

Free tiers move on each vendor's schedule, not ours. Devshot's daily brief flags the pricing changes developers act on. Subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Free LLM APIs in 2026: Rate Limits and Training Terms

URL: https://devshot.co/blog/free-llm-apis
Type: blog
Published: 2026-10-10
Updated: 2026-10-10
Summary: OpenRouter, Cerebras, Google Gemini and Mistral free LLM APIs compared on requests, token caps, training terms and first paid prices, checked October 2026.

Guide

## Free LLM APIs in 2026: Rate Limits and Training Terms

OpenRouter, Cerebras, Google Gemini and Mistral free LLM APIs compared on requests, token caps, training terms and first paid prices, checked October 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 10, 2026 · 10 min read

Short answer: OpenRouter gives the most daily requests, 1,000 a day once you have bought $10 of credit. Cerebras gives the largest token budget, 1M tokens a day on GPT OSS 120B, but its $5 trial credit expires 30 days after activation. When free access ends, Mistral Small 4 is the cheapest paid route we verified, at $0.15 input and $0.60 output per 1M tokens.

Free tiers change often, and each vendor states its caps in a different place. Apart from the two labelled counts below, this page uses only what each vendor says on its own pricing, rate-limit or changelog page, checked on 10 October 2026. Where a cap sits behind a sign-in, the table says so instead of guessing.

Devshot data: 22 of the 659 models in OpenRouter's public model list carry the :free suffix today, counted from its models API on 10 October 2026. Toolradar data: the [AI model deployment ranking](https://toolradar.com/best/ai-model-deployment) evaluated 42 tools.

## Free tiers at a glance

Route | Requests per minute | Requests per day | Tokens per day | Models | Training on free-tier data | First paid price per 1M tokens (input / output) |

OpenRouter free models | 20 | 50, or 1,000 after buying credit | Not published | 22 models with the :free suffix | Not logged by default; providers that log, or whose policy is unconfirmed, are skipped unless the training toggle is on | Model list price, no markup |

Cerebras Free Trial | 5 | Not published | 1M a day and 1M an hour | GPT OSS 120B, Qwen 3.8 27B | Not stated on the pricing page | $0.35 / $0.75 (GPT OSS 120B) |

Google Gemini API free tier | Not in the docs (AI Studio, sign-in) | Not in the docs | Not in the docs | Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite | Free plan content used to improve products; paid plan not used | $0.25 / $1.50 (3.1 Flash-Lite, text) |

Mistral API free mode | Not public (Limits page, sign-in) | Not public | Not public (monthly cap on the Limits page) | Mistral Small 4 and others | "Opt-out" listed, not split by plan | $0.15 / $0.60 (Small 4) |

## OpenRouter: the most daily requests, once you buy $10

Free models on OpenRouter are the ones whose ID ends in :free. Every account gets 20 requests a minute on them. An account with no credit gets 50 requests a day. Buying at least $10 of credit lifts that ceiling to 1,000 a day. Failed attempts count against the daily cap, and a negative balance blocks free models until you add credit.

Buying credit carries a fee. OpenRouter charges 5.5% on credit purchases, with a $0.80 minimum on card payments. On a $10 top-up, 5.5% would be $0.55, so the $0.80 minimum applies. Inference carries no markup of its own: the FAQ says it is billed at the provider's list price on every plan. Paid usage therefore costs whatever the model you route to lists at, and the cost changes by model.

The cap counts requests, not tokens, and OpenRouter publishes no token cap for free models. A 20-token prompt and a 20,000-token prompt both use one request. For a gateway that adds its own fee on top of model prices, see the OpenRouter alternatives ranking, and the best API gateways list covers the wider field. That fee is a separate cost from the free tier here. The [OpenRouter entry on Toolradar](https://toolradar.com/tools/openrouter) has the directory details.

OpenRouter says prompts are not logged by default. It skips providers that log prompts, or whose policy it cannot confirm, unless you switch on the model training toggle in your privacy settings. That is a routing rule, so check the policy of the specific model you pick before you send private text.

## Cerebras: the largest token budget, on a 30-day clock

Cerebras has no permanent free tier. New accounts get a one-time $5 Free Trial credit after they add a valid payment method. The credit expires 30 days after activation. Cerebras says you are not charged or enrolled in paid usage automatically, so access stops when the credit runs out until you buy more.

While the trial lasts, GPT OSS 120B and Qwen 3.8 27B each allow 5 requests a minute, 30,000 uncached tokens a minute (90,000 total), and 1 million tokens an hour and a day. The per-minute request cap is the tighter limit. A chat app with a few users at the same time hits 5 requests a minute before the daily token cap matters.

The token budget is generous, but the credit is not. At list price, GPT OSS 120B costs $0.35 per 1M input tokens and $0.75 per 1M output tokens on Cerebras's developer tier. A workload of 1M tokens a day, split evenly, costs $0.55 a day, so the $5 credit covers about nine days of it. Qwen 3.8 27B lists at $0.99 input and $1.49 output, which puts it at more than twice the cost of GPT OSS 120B on the same mix.

Cerebras's pricing page does not say whether prompts from free accounts train its models. Read its data terms before you run customer text through the trial.

## Google Gemini API: four free models, caps behind a sign-in

Four current models are free of charge on the Standard tier: Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.1 Flash-Lite. The [Gemini model on Toolradar](https://toolradar.com/tools/gemini-3-1-flash-lite) is the 3.1 Flash-Lite listing.

Google publishes per-model free caps only in Google AI Studio, after sign-in. Its rate-limit docs explain the rules: limits apply per project, and daily quotas reset at midnight Pacific time. The numbers themselves are not in the docs, so this page gives no requests-per-minute figure for Gemini.

The training line matters more than the caps. Google's pricing page says content on the Free plan is used to improve its products, and the Paid plan says it is not. For a prototype that handles real user text, that single line decides the route. Gemini is the only provider in this list whose pricing page states the paid-tier rule in writing.

On price, Gemini 3.1 Flash-Lite is the cheapest Gemini model at $0.25 input (text, image and video; $0.50 for audio) and $1.50 output per 1M tokens. Gemini 3.8 Flash and 3.6 Flash cost $0.75 input and $3.75 output through December 31, 2026. From January 1, 2027, both double to $1.50 and $7.50. A team that builds on 3.8 Flash now should budget for that step.

## Mistral: free mode with caps behind the console

Mistral's free mode lets you create API keys and use "included monthly usage within the limits shown on the Limits page" in the admin panel. The public docs give no requests-per-second, tokens-per-minute or monthly figure, so this page cannot quote one. Accounts move up tiers by billed usage, not by prepaid credit. Tier 2 starts after more than $20 billed, Tier 3 after $100, and Tier 4 after $500. Prepaid credit does not count toward those thresholds. [Mistral AI on Toolradar](https://toolradar.com/tools/mistral-ai) lists the vendor record.

Mistral Small 4 is the cheapest paid model in this list on both input and output. Cached input drops to $0.015 per 1M tokens, and batch processing costs half. The price table is in dollars and gives no effective date, so re-check it before you budget.

Mistral's plan table lists model training as "Opt-out" and does not say which plans that covers. Read the data terms before you send customer text through the free mode.

## Cohere trial keys: good for testing, not for production

Cohere's trial keys are free, with 20 requests a minute on chat models and a cap of 1,000 API calls a month. Cohere says trial keys cannot be used for production or commercial purposes, so a free key is for evaluation only. Its pricing page lists no current price for Command A, so this page quotes none. [Cohere on Toolradar](https://toolradar.com/tools/cohere) is the directory record.

## GitHub Models is gone

GitHub Models was retired on July 30, 2026. Its playground, model catalog, inference API and bring-your-own-key option no longer work for any customer. GitHub points people to Microsoft Foundry for model access and to GitHub Copilot instead. Any list that still names GitHub Models as a free route is out of date.

## What a prototype costs after the free tier

Free caps are for testing. To compare real costs once free access ends, this page prices one scenario: 1 million tokens a day for 30 days, half input and half output, at each route's first paid list price. For Mistral Small 4, half the tokens at the input price plus half at the output price comes to $0.375 a day, and $0.375 x 30 = $11.25 a month.

Route | Input per 1M | Output per 1M | Cost for 30 days |

Mistral Small 4 | $0.15 | $0.60 | $11.25 |

Cerebras GPT OSS 120B (developer tier) | $0.35 | $0.75 | $16.50 |

Gemini 3.1 Flash-Lite (text) | $0.25 | $1.50 | $26.25 |

Cerebras Qwen 3.8 27B (developer tier) | $0.99 | $1.49 | $37.20 |

Mistral Small 4 is the cheapest route at any input share, because it is lower on both input and output. Gemini 3.1 Flash-Lite has the cheaper input, so it beats Cerebras GPT OSS 120B only when input is more than about 88% of the tokens. At the even split used here, Gemini costs $9.75 more over 30 days. OpenRouter is not in the table: it bills the model you choose at that model's list price, so its cost depends on the model.

## Which free route to use

For many daily calls with cheap models, OpenRouter is the route. Top up once to raise the daily cap, and accept that paid usage then follows each model's list price. For a short test measured in tokens, Cerebras's trial gives the most volume, but at a 1M-token-a-day workload the credit runs out in about nine days. For Google models on the free tier, use them only when training on your text is acceptable, and move to the paid tier when it is not.

For a steady workload you will keep paying for, the cheapest scenario here is Mistral Small 4. Choose the model on your own test set first, because this ranking says nothing about quality. For the coding side of the bill, see AI for developers, Claude Code pricing, Cursor's plans, the best AI code generators and AI-assisted coding. Coding assistants with a free plan, and the cap each one publishes, are in Best Free AI Coding Tools in 2026.

## How this page was checked

- Each price, cap and date comes from a vendor's own pricing, rate-limit or changelog page, read on 10 October 2026. Gemini's pricing page was last updated on 9 October 2026 (UTC).
- The free-model count comes from OpenRouter's public models API, and the tool count from Toolradar's ranking, both read on the same day.
- A cap appears only when the vendor states it. Caps behind a sign-in are named as such, with no number.
- The cost scenario is 1 million tokens a day for 30 days, half input and half output, at first paid list prices, with no discount and no caching. It follows the per-million-unit pricing used in the text to speech API ranking. Another cost comparison on this site is object storage pricing compared.
- Rankings are by cost for that scenario only. They do not measure model quality.

## Frequently asked questions

### Which free LLM API has the most requests per day?

OpenRouter. Its free models allow 50 requests a day with no credit, and 1,000 a day after you buy $10 of credit. Every account gets 20 requests a minute on them.

### Is there a free LLM API with a token limit?

Cerebras. Its Free Trial allows 1 million tokens an hour and a day on GPT OSS 120B and Qwen 3.8 27B, but the $5 credit expires 30 days after activation.

### Do free LLM APIs train on my prompts?

It depends on the vendor. Gemini's free plan uses content to improve Google's products, and its paid plan does not. OpenRouter does not log prompts by default and skips providers that log them unless you switch on the training toggle. Cerebras does not say on its pricing page. Mistral's plan table lists "Opt-out" without naming the plans it covers.

### Can I run a product on a free LLM API trial key?

Not with Cohere. Its trial keys are capped at 1,000 API calls a month and cannot be used for production or commercial purposes.

### What does an LLM API cost after the free tier?

For 1 million tokens a day over 30 days, split evenly between input and output, Mistral Small 4 costs $11.25 a month at its first paid list price, the lowest in this comparison. Gemini 3.1 Flash-Lite costs $26.25 on the same mix.

### Is GitHub Models still a free option?

No. GitHub retired GitHub Models on July 30, 2026, and points users to Microsoft Foundry and GitHub Copilot instead.

Cite this: Devshot, "Free LLM APIs in 2026: Rate Limits and Training Terms", October 2026.

Some offers on this page may be paid placements or contain affiliate links.

---

# GitHub vs GitLab 2026: Pricing, CI Minutes and Which to Pick

URL: https://devshot.co/blog/github-vs-gitlab
Type: blog
Published: 2026-10-10
Updated: 2026-10-10
Summary: GitHub vs GitLab in 2026: seat prices, CI minutes per plan, Copilot and GitLab AI credits, and the yearly cost for 5, 25 and 100 developers.

Guide

## GitHub vs GitLab 2026: Pricing, CI Minutes and Which to Pick

GitHub vs GitLab in 2026: seat prices, CI minutes per plan, Copilot and GitLab AI credits, and the yearly cost for 5, 25 and 100 developers.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 10, 2026 · 10 min read

Key facts2 products compared: GitHub, GitLab

- Updated: October 10, 2026
- Prices as of: October 2026

- 2 products compared: GitHub, GitLab
- GitHub (best for: Teams that want the cheapest paid seat and the most free CI): Free plan; $4 per user a month on Team (labeled first 12 months)
- GitLab (best for: Teams with long pipelines, or teams that want a free self-managed install): Free plan; $29 per user a month on Premium, billed annually

For most developer teams in 2026, pick GitHub. Team is listed at $4 per user a month, and with Copilot Business at $19 the whole stack comes to $23, against $29 a month for GitLab Premium billed annually. GitLab wins on paid CI minutes (10,000 against 3,000) and on a free self-managed install.

Two caveats can change that answer. GitHub's pricing page labels its Team and Enterprise figures "for the first 12 months" and gives no renewal rate, so the gap may narrow after year one. GitLab's AI credits are a limited-time offer. All prices below are verified on vendor pages, October 2026. The sections below work the numbers at 5, 25 and 100 developers, and show where the verdict flips.

Toolradar data: Toolradar, the software directory we run, evaluated 114 CI/CD tools for its October 2026 ranking. GitHub and GitLab each have a profile on the site: [GitHub](https://toolradar.com/tools/github) and [GitLab](https://toolradar.com/tools/gitlab). The seat math below uses the vendors' own pricing pages, not that ranking.

## At a glance

Product | Best for | Starting price (USD) | Standout |

GitHub | Teams that want the cheapest paid seat and the most free CI | Free plan; $4 per user a month on Team (labeled first 12 months) | 2,000 Actions minutes a month free on private repos |

GitLab | Teams with long pipelines, or teams that want a free self-managed install | Free plan; $29 per user a month on Premium, billed annually | Self-Managed Free at $0; 10,000 compute minutes on Premium |

GitHub and GitLab both sell hosted Git, a CI system and paid seats on top. They differ in where the money goes. GitHub charges less per seat and includes less pipeline time on its paid tier, so a team with heavy CI pays for the difference in overage. GitLab charges more per seat and bundles the minutes, and some AI credits, into that seat.

If the Git host is only one of your choices, our CI/CD tools ranking covers the pipeline systems that sit beside it, and best Git clients covers the desktop side.

## Winner by dimension

Dimension | Winner | The numbers |

Free plan CI minutes | GitHub | 2,000 a month on Free, against 400 on GitLab's Free plan |

Entry paid seat | GitHub | Team against Premium billed annually; GitHub's Team figure carries the first-12-months label |

Paid CI minutes | GitLab | 10,000 a month on Premium, against 3,000 on GitHub Team |

AI included in the seat | GitLab, for now | $12 of GitLab Credits per user a month on Premium; GitHub's Copilot Business is a separate seat |

Free self-hosting | GitLab | Self-Managed Free is listed as free on GitLab's pricing page; GitHub's page lists hosted plans only |

Yearly cost for 25 developers | GitHub | $6,900 with Copilot Business, against $8,700 for GitLab Premium |

No single column settles it. A team that runs little CI and wants AI in every seat should pick GitHub. A team that runs long pipelines, or wants to self-host for nothing, should look hard at GitLab. The rest of this piece shows the numbers behind each row.

## What the seats cost at 5, 25 and 100 developers

The scenario is a developer team on GitHub's Team plan with Copilot Business for AI, against GitLab Premium billed annually. GitHub's pricing page does not state a billing basis for its seat price, so the GitHub figures are the monthly rate times 12. For 25 developers the arithmetic is $4 x 25 x 12 = $1,200 for Team, plus $19 x 25 x 12 = $5,700 for Copilot Business, a total of $6,900 a year.

Developers | GitHub Team | GitHub Team + Copilot Business | GitLab Premium, billed annually | GitHub saving per year |

5 | $240 a year ($20 a month) | $1,380 a year ($115 a month) | $1,740 a year ($145 a month) | $360 |

25 | $1,200 a year ($100 a month) | $6,900 a year ($575 a month) | $8,700 a year ($725 a month) | $1,800 |

100 | $4,800 a year ($400 a month) | $27,600 a year ($2,300 a month) | $34,800 a year ($2,900 a month) | $7,200 |

The GitHub stack comes to $6 per developer a month less than GitLab Premium, or $72 a year for each person. That gap is the whole price argument for GitHub, and it holds only while the Team figure holds after year one. Copilot Business is the larger line at every size, so a team that does not want AI in its seats should compare Team alone. At five developers, Team costs $240 a year.

The gap closes if GitHub's Team price rises. The GitHub stack matches GitLab Premium only at a Team price of $10 a user a month, which is the GitLab Premium price minus the Copilot Business seat. Above $10, GitLab is cheaper on list prices and the verdict flips.

## The first-12-months label

GitHub's pricing page shows its Team price and an Enterprise price "starting at" $21, and labels both "for the first 12 months". The page gives no renewal price for either. Read the page on the day you buy, and get the renewal rate in writing before you put year two in a budget.

GitLab's Premium price is billed annually, and its page shows no monthly rate. That makes the price easy to budget but hard to start small on. Annual billing means a team that wants to try Premium first still commits to a full year.

## CI minutes: where each plan runs out

Both platforms meter pipeline time each month, with a fixed allowance per plan. GitHub's Free plan gives 2,000 Actions minutes a month on private repositories, and Team gives 3,000 with 2 GB of package storage. GitLab's Free plan gives 400 compute minutes, Premium 10,000 and Ultimate 50,000. Standard GitHub-hosted runners are free on public repositories, but larger runners are charged there too.

The overage is where Team gets expensive. GitHub bills $0.006 a minute for a Linux 2-core x64 runner, and $0.005 a minute for arm64. A team that runs 10,000 minutes a month on Team pays for the 7,000 minutes above the 3,000 included, which is $42 a month on top of the seats. Larger runners cost more per minute, so treat $42 as the floor on Linux.

Count your pipeline minutes for the last three months before you choose. Under 3,000 a month, Team's allowance covers the work. Above it, each extra 1,000 minutes costs $6 on Linux 2-core runners. For 25 developers the seat saving is $1,800 a year, and 25,000 overage minutes a month at $0.006 a minute use that up in twelve months. A 25-developer team running roughly 28,000 minutes a month on Team would spend the whole seat saving on CI.

GitHub also announced a fee for self-hosted runners, the machines a team runs itself. It postponed that change on 15 December 2025 and set no new start date. Treat it as an open question, not a line item, and check GitHub's billing docs before you sign. Both platforms let you run your own runners, and our self-hosting tools guide covers what running your own hardware involves.

## AI seats: bundled or bolted on

GitHub sells AI as a separate subscription. Copilot Pro is $10 a month for an individual, Copilot Business is $19 per granted seat a month, and Copilot Enterprise is $39. Copilot is billed per granted seat, so the AI line follows the people you assign it to, not the headcount on the invoice. A GitHub team that wants AI for every developer pays $23 per person a month on Team plus Business.

GitLab bundles credits into its paid seats instead. Its page lists GitLab Credits at $1 each, with $12 of credits per user a month on Premium and $24 on Ultimate, and calls that a promotional offer for a limited time. Do not budget the $12 as permanent until GitLab confirms it past the promotion.

Compare like with like. Right now a GitLab Premium seat includes AI credits and the GitHub stack includes Copilot Business, which puts the comparison at $23 against $29 with AI in both. If the credits lapse, the GitLab seat carries no included AI credits, and the two prices stop being like for like. Our GitHub Copilot review covers what the Copilot seat includes, and Cursor vs GitHub Copilot is the comparison for teams weighing a Copilot alternative. Cursor's own plan prices are in Cursor pricing.

## Who should pick which

- Pick GitHub when the team has 5 to 100 developers, the pipelines fit 3,000 minutes a month, and you want Copilot in the same bill. Team is the cheapest paid seat on the two pricing pages, as long as its renewal rate holds. If Copilot is not the assistant you want, best AI coding assistants lists the rest.
- Pick GitLab when pipelines run past 3,000 minutes a month, when you want a self-managed install on the free Self-Managed plan, or when AI credits inside the seat matter more than the seat price.
- Wait on Ultimate. GitLab Ultimate has no published price, so a like-for-like comparison with GitHub Enterprise plus Copilot Enterprise cannot be costed from the vendor pages. GitHub Enterprise is "starting at" $21 with the same 12-month label, and Copilot Enterprise is $39.

## Gotchas that change the bill

- GitHub's Team and Enterprise prices carry the first-12-months label, and the page gives no renewal rate.
- Copilot Business is billed per granted seat, so every developer you assign it to adds one seat's price, not one flat amount.
- GitHub's overage is $0.006 a minute on Linux 2-core runners. At 10,000 minutes on Team, that is $42 a month before seats.
- GitLab's $12 of monthly AI credits per user is a limited-time offer.
- GitHub's self-hosted runner fee is postponed with no new date. A team that runs its own runners should ask before it budgets.

## How we compared

Every price on this page comes from a vendor page read on 10 October 2026: [GitHub's pricing page](https://github.com/pricing), [GitHub's Copilot plans page](https://github.com/features/copilot/plans), [GitHub's Copilot plan docs](https://docs.github.com/en/copilot/get-started/plans), [GitHub's Actions billing docs](https://docs.github.com/en/billing/managing-billing-for-github-actions/about-billing-for-github-actions), and [GitLab's pricing page](https://about.gitlab.com/pricing/). [GitHub's update on Actions pricing changes](https://github.com/resources/insights/2026-pricing-changes-for-github-actions) is the source for the self-hosted runner status. The seat totals are multiplication, and the 25-developer example shows the sums. Toolradar's [GitHub vs GitLab guide](https://toolradar.com/blog/github-vs-gitlab) is the broader read. This page covers the cost side. For the same kind of head-to-head on other developer tools, see Supabase vs Firebase and Vercel vs Netlify.

## FAQ

### Is GitHub or GitLab cheaper in 2026?

GitHub, on the prices its pricing page shows. Team is $4 per user a month, against $29 for GitLab Premium billed annually. Add Copilot Business and GitHub comes to $23, still $6 a month under GitLab Premium. The Team figure is labeled "for the first 12 months", so compare renewal rates before you decide.

### Which has more free CI minutes, GitHub or GitLab?

GitHub. Its Free plan includes 2,000 Actions minutes a month on private repositories, against 400 compute minutes a month on GitLab's Free plan. Standard GitHub-hosted runners are free on public repositories.

### Which is better for a team that runs a lot of CI?

GitLab on the numbers. Premium includes 10,000 compute minutes a month, and GitHub Team includes 3,000. A team that needs 10,000 minutes on GitHub Team pays $0.006 a minute for the 7,000 extra on Linux 2-core runners, which is $42 a month before seats.

### Can I run GitLab for free on my own servers?

Yes. GitLab lists a Self-Managed Free plan on its pricing page at no charge. GitHub's pricing page lists only hosted plans, so the free self-managed route is GitLab's alone on that page.

### Is GitHub Copilot included in the Team plan?

No. Copilot is a separate subscription. On GitHub's plans page, Copilot Business is $19 per granted seat a month, and Copilot Pro is $10 a month for an individual. Our GitHub Copilot review covers what each tier includes.

### Has GitHub started charging for self-hosted runners?

Not yet. GitHub announced a fee for self-hosted runners and postponed it on 15 December 2025, with no new start date. Check GitHub's billing docs before you budget for it.

Cite this: Devshot, "GitHub vs GitLab 2026: Pricing, CI Minutes and Which to Pick", October 2026.

Devshot covers the tools and changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Object Storage Pricing Compared (2026): S3 vs R2 vs B2 vs Wasabi at 1 TB

URL: https://devshot.co/blog/object-storage-pricing-compared
Type: blog
Published: 2026-10-06
Updated: 2026-10-06
Summary: Wasabi is the cheapest object storage at $7.99/mo for 1TB stored and 5TB egress; AWS S3 costs $464.90/mo for the same load, verified September 2026.

Guide

## Object Storage Pricing Compared (2026): S3 vs R2 vs B2 vs Wasabi at 1 TB

Wasabi is the cheapest object storage at $7.99/mo for 1TB stored and 5TB egress; AWS S3 costs $464.90/mo for the same load, verified September 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 6, 2026 · 14 min read

For 1 TB stored, 5 TB of egress and 1.1 million requests a month, Wasabi is the cheapest S3-compatible object storage at $7.99/mo, verified on Wasabi's pricing page, September 2026. Hetzner Object Storage is the runner-up at $12.79/mo priced in USD through its own site currency switcher, and Cloudflare R2 follows at $14.85/mo, the pick if you want free requests baked into a real free tier rather than a flat rate. AWS S3 costs $464.90/mo for that same load, almost entirely because of its per-GB egress fee.

Every price below was read on the vendor's own pricing page today, in USD, using each vendor's own published dollar figure (a currency switcher, in Hetzner's case) rather than a conversion this page performed. This page compares 10 S3-compatible and object-storage-adjacent services developers actually put a bucket on: hyperscaler S3 clones, storage-first challengers and the object storage add-on on three VPS platforms.

Devshot data: of the 30 tools Toolradar's [cloud storage ranking](https://toolradar.com/best/cloud-storage) evaluated (updated September 2026, the closest built category to developer object storage on that site), the 10 developer-facing services priced here all convert to a USD figure their own site publishes, and the spread between the cheapest and the most expensive runs 76x.

## The scenario, and why it matters

1 TB stored, 5 TB of monthly egress, 1 million GET requests and 100,000 PUT requests describes a mid-size app bucket: user uploads, a CDN origin, or a backup target that gets read a lot more than it gets written. That mix is what separates the vendors below, because storage is cheap everywhere and egress is not. A provider with a low sticker price on storage alone can still lose badly once 5 TB a month starts leaving the network, the same lesson cursor pricing and Claude Code pricing teach about usage-metered developer tools generally: the sticker price is never the whole invoice.

## Cost comparison at the scenario

Tool | Entry price | Cost at scenario | Free plan (limits) | Billing basis | Hidden costs |

[Wasabi](https://toolradar.com/tools/wasabi) | $7.99/TB/mo | $7.99/mo | None; 30-day free trial only | Flat per TB, no egress or request meter | 90-day minimum storage duration, billed for remaining days on early delete |

[Hetzner](https://toolradar.com/tools/hetzner) Object Storage | $7.99/mo | $12.79/mo | 1TB storage + 1TB egress included in the base fee; no free-forever tier | Flat monthly base plus per-TB overage, billed hourly | Additional storage $8.86/TB/mo, additional egress $1.20/TB past the included 1TB |

[Cloudflare R2](https://toolradar.com/tools/cloudflare-r2) | Free | $14.85/mo | 10GB storage, 1M Class A ops, 10M Class B ops/mo | Per GB storage, per-million ops, zero egress | Class A (write) ops $4.50 per million past the free tier |

[Tigris Data](https://toolradar.com/tools/tigris-data) | Free tier | $20.90/mo | 5GB storage, 10k Class A, 100k Class B/mo | Per GB storage, per-1,000 ops, zero egress | Class A request fee $0.005 per 1,000 past the free 10,000 |

[Backblaze](https://toolradar.com/tools/backblaze) B2 | $6.95/TB/mo | $26.88/mo | 10GB storage, 3x storage in free egress | Per GB storage, metered egress past the 3x allowance | Download overage $0.01/GB once the free 3x-storage egress runs out |

[Linode](https://toolradar.com/tools/linode) (Akamai) Object Storage | $0.02/GB/mo | $40.00/mo | 1TB outbound transfer/mo | Pure usage, no base fee | Transfer $0.005/GB past the 1TB included each month |

[DigitalOcean](https://toolradar.com/tools/digitalocean) Spaces | $5/mo | $54.95/mo | 250GiB storage, 1TiB transfer | Flat base plus per-GiB overage | Storage overage $0.02/GiB and transfer overage $0.01/GiB past the included quota |

[Azure Blob Storage](https://azure.microsoft.com/en-us/pricing/details/storage/blobs/) (Hot, LRS) | $0.0208/GB/mo | $448.00/mo | 100GB egress/mo (account-wide) | Per GB storage, per-10,000 ops, metered egress | Egress $0.087/GB past the 100GB free allowance |

[AWS S3](https://toolradar.com/tools/aws-s3) Standard | $0.023/GB/mo | $464.90/mo | 100GB egress/mo (account-wide) | Per GB storage, per-1,000 requests, metered egress | Egress $0.09/GB once the 100GB free allowance is spent |

[Google Cloud Storage](https://toolradar.com/tools/google-cloud) (Standard) | $0.020/GB/mo | $608.76/mo | 100GB egress/mo, 5GB storage, 5k Class A / 50k Class B ops (Always Free) | Per GB storage, per-1,000 ops, metered egress | Egress $0.12/GB, the steepest per-GB rate on this page |

Hetzner Object Storage defaults to EUR (a base fee of €6.49/mo covering 1TB of storage and 1TB of egress, plus €1.00/TB for egress past that allowance, €10.49/mo at this scenario), but its pricing page carries a currency switcher with a published USD price, the base fee and per-TB egress overage figures used in the table above, never derived by converting the euro figure ourselves.

## The arithmetic, shown once

Wasabi charges one flat rate per TB regardless of usage, so its cost at this scenario is just that sticker price, with no egress or request line ever added, which is the entire pitch.

AWS S3 is the opposite: storage is a rounding error next to egress. Storage is $0.023/GB x 1,000GB = $23.00. Requests add 100,000 PUT / 1,000 x $0.005 = $0.50 and 1,000,000 GET / 1,000 x $0.0004 = $0.40. The first 100GB of egress is free across an AWS account, so the billable egress is 4,900GB: 4,900GB x $0.09/GB = $441.00. Total: $23.00 + $0.50 + $0.40 + $441.00 = $464.90/mo, with egress alone accounting for 95% of the bill.

Google Cloud Storage follows the same shape at a steeper egress rate: Google's Always Free tier nets out 5GB of storage, 5,000 Class A and 50,000 Class B operations, and 100GB of egress every month, so storage is $19.90, requests add $0.86 combined, and the billable 4,900GB of egress at $0.12/GB is $588.00, for a $608.76/mo total, the most expensive vendor on this page.

## Where the ranking flips

This order holds at 5 TB of egress against 1 TB stored. Drop the egress volume to well under 3 TB a month and Backblaze B2's free-egress allowance (3x average storage, so 3TB free on a 1TB bucket) starts to cover the whole month: at near-zero egress, B2's bill is just its storage charge, about $6.88/mo, which undercuts Wasabi's flat rate. Past roughly 10 TB stored, DigitalOcean Spaces and Linode Object Storage both move fully onto their per-GB overage rate from the first byte, since their included quotas stop scaling with usage, so they lose ground to Wasabi, R2 and Tigris faster the bigger the bucket gets. At egress far above 5 TB a month, the hyperscalers' per-GB egress rate keeps compounding while Wasabi, R2, Tigris and Hetzner's included tier stay flat, so the gap between the cheapest and most expensive rows only widens with more traffic.

## Wasabi: flat pricing, with a 90-day catch

[Wasabi](https://toolradar.com/tools/wasabi)'s entire pitch is one flat per-TB rate (see the table above), no egress fee and no per-request charge, confirmed on its pricing and FAQ pages. That works cleanly for a bucket that stays roughly the same size all month. The catch shows up on deletion: Wasabi enforces a 90-day minimum storage duration, and deleting an object before that window bills you for the remaining days as if it were still stored. Objects under 4KB are also billed as if they were 4KB. Neither trap matters for a stable media or backup bucket; both matter for a bucket full of short-lived temp files.

## Cloudflare R2: the free tier does real work

[R2](https://toolradar.com/tools/cloudflare-r2) charges $0.015/GB for storage and nothing for egress, and its free tier, 10GB storage plus 1 million write and 10 million read operations a month, absorbs this scenario's entire request volume. The only line that shows up on the invoice is storage past the free 10GB, which is why R2 lands at $14.85 rather than a flat $15.00. Push past 1 million writes a month and Class A operations cost $4.50 per million, the one number worth watching if the workload is upload-heavy rather than read-heavy.

## Tigris Data: R2's shape, a slightly higher entry price

Tigris also charges nothing for egress and prices storage per GB, but its free tier is smaller (5GB storage, 10,000 Class A requests, 100,000 Class B requests a month), so more of this scenario's usage falls outside it. The result, $20.90/mo (storage $20.00 plus $0.45 in Class A and $0.45 in Class B overage), sits between R2 and Backblaze B2 rather than matching either. Tigris is newer and smaller than both, which matters for teams weighing ecosystem maturity against price.

## Backblaze B2: cheap storage, egress that runs out

[Backblaze](https://toolradar.com/tools/backblaze) B2 prices storage at $6.95/TB/mo and gives every account free egress equal to three times its average monthly storage, 3TB on a 1TB bucket here. This scenario's 5TB of egress blows past that allowance by 2TB, adding $20.00 at $0.01/GB and pushing the total to $26.88. A bucket that stays under its 3x-storage egress ceiling effectively pays B2's storage price and nothing else, which is why the ranking flip above puts B2 ahead of Wasabi at lower traffic.

## Linode and DigitalOcean: VPS-style object storage, not a hyperscaler clone

Linode (Akamai) Object Storage has no monthly base fee: $0.02/GB storage plus $0.005/GB egress past a 1TB-included allowance, landing at $40.00/mo here. DigitalOcean Spaces flips that model, a $5/mo flat fee that bundles 250GiB of storage and 1TiB of transfer, then meters storage at $0.02/GiB and transfer at $0.01/GiB past those quotas; at 1TB stored and 5TB moved, the overage adds up to $54.95/mo. Both are the object storage bolted onto a VPS provider's console rather than a dedicated storage product, which suits a team that already runs compute there and wants one invoice, not a team shopping storage on price alone; Devshot's best VPS hosting for developers roundup covers the compute side of that tradeoff. The same platform-versus-piece-by-piece tradeoff shows up in Vercel's own pricing: bundling usually loses on price once one meter runs hot, and wins on invoice count.

## AWS S3, Azure Blob and Google Cloud Storage: the egress bill dominates

All three hyperscalers price storage within a few tenths of a cent of each other and would look nearly tied on a storage-only comparison. What separates them here is the egress rate: Google Cloud Storage charges $0.12/GB, AWS S3 charges $0.09/GB, and Azure Blob Storage (Hot tier, LRS, East US) charges $0.087/GB for outbound bandwidth on Microsoft's premium global network, per the same [bandwidth pricing page](https://azure.microsoft.com/en-us/pricing/details/bandwidth/) used for this calculation. Azure's storage rate, $0.0208/GB-month, is read from Microsoft's retail Prices API ([prices.azure.com](https://prices.azure.com/api/retail/prices?$filter=armRegionName+eq+'eastus'+and+skuName+eq+'Hot+LRS'), Hot LRS, East US), not the marketing pricing page, because the two can disagree and the API is what actually bills. At this scenario's 5TB of egress, that rate difference alone is worth over $150/mo between Azure and Google. None of the three waive egress at any real volume in this range; all three have a 100GB-a-month free allowance (AWS and Azure perpetually, Google under its Always Free tier), effectively a rounding error against 5TB. A Vercel or Netlify deployment that proxies file downloads through one of these three buckets inherits that egress bill on top of its own hosting invoice, which is why teams serving a lot of static assets often put R2 or Wasabi in front instead.

## Hetzner: cheap, and it does quote a dollar price

[Hetzner](https://toolradar.com/tools/hetzner)'s Object Storage pricing page defaults to EUR, a base fee of €6.49/mo covering 1TB of storage and 1TB of egress, with additional egress at €1.00/TB and no charge at all for S3 API requests (PUT, GET, DELETE). Switch the page's own currency selector to USD, though, and the base package matches Wasabi's headline number exactly (see the table above), with additional storage at $8.86/TB and additional egress at $1.20/TB, a published first-party dollar price, not a conversion this page performed. At this scenario, 4TB of egress past the included 1TB adds 4 x $1.20 = $4.80 to the base fee, for $12.79/mo total, second only to Wasabi. Objects under 64KB are billed as if they were 64KB, and the base fee is billed hourly with no monthly minimum below a full active month.

## Object storage is rarely the only meter

A bucket doesn't run alone. A Supabase project bills its own storage and egress separately from whatever sits behind it, and the Supabase vs Firebase comparison covers that invoice in full; a team choosing a backend and an object store in the same quarter should price both before committing to either. Devshot's statistics page tracks how often developer tooling costs move month to month, which is the same reason every price above carries a verification date instead of a "starting at" figure with no month attached. AWS, Google Cloud and Azure each run a free tier with its own limits, and Cloud Free Tiers Compared lists what those include, with Oracle Cloud.

## Hidden costs that do not show up on the plan card

- Wasabi: 90-day minimum storage duration; deleting early bills the remaining days. Objects under 4KB are billed as 4KB.
- Cloudflare R2: Class A (write) operations cost $4.50 per million past the 1 million included each month; Infrequent Access storage doubles that to $9.00 per million.
- Tigris Data: the free tier caps at 10,000 Class A and 100,000 Class B requests a month, and the free storage allowance is 5GB, half of R2's; past that, small per-request fees apply.
- Backblaze B2: the free egress allowance is 3x a bucket's average monthly storage; usage past that meters at a per-GB overage rate, which added $20.00 to this scenario's total.
- Linode Object Storage: there is no base fee cushioning a spiky month, so egress bills from the first GB past the 1TB included each month.
- DigitalOcean Spaces: storage and transfer overage both apply past the 250GiB/1TiB the $5/mo base includes, which is most of the $54.95/mo total here.
- AWS S3: egress is $0.09/GB after the first 100GB, and PUT/COPY/POST/LIST requests run $0.005 per 1,000, a separate line most storage-only comparisons skip.
- Azure Blob Storage: egress is $0.087/GB on the premium network after 100GB free; the cheaper transit-ISP routing option is $0.08/GB but is not the default.
- Google Cloud Storage: egress is $0.12/GB after the first 100GB free each month under the Always Free tier, the highest per-GB egress rate of the three hyperscalers here.
- Hetzner: the base fee is billed hourly, so a bucket that exists for even one hour in a month is billed for that hour; objects under 64KB are billed as 64KB.

## How we checked these prices

Louis Corneloup, founder of Devshot, Dupple and Toolradar, read each vendor's own pricing page on September 28, 2026: [AWS S3](https://aws.amazon.com/s3/pricing/), [Cloudflare R2](https://developers.cloudflare.com/r2/pricing/), [Backblaze B2](https://www.backblaze.com/cloud-storage/pricing), [Wasabi](https://wasabi.com/pricing) and its [pricing FAQ](https://wasabi.com/pricing/faq), [Tigris Data](https://www.tigrisdata.com/pricing/), [Linode/Akamai Object Storage](https://www.akamai.com/cloud/pricing/north-america), [DigitalOcean Spaces](https://www.digitalocean.com/pricing/spaces-object-storage), [Google Cloud Storage](https://cloud.google.com/storage/pricing), [Azure Blob Storage](https://azure.microsoft.com/en-us/pricing/details/storage/blobs/) plus [Azure bandwidth pricing](https://azure.microsoft.com/en-us/pricing/details/bandwidth/), and [Hetzner Object Storage](https://www.hetzner.com/storage/object-storage/pricing/). Azure's storage rate is read from Microsoft's retail Prices API ([prices.azure.com/api/retail/prices](https://prices.azure.com/api/retail/prices?$filter=armRegionName+eq+'eastus'+and+skuName+eq+'Hot+LRS')), Hot tier, LRS, East US, rather than the marketing page. Costs at the scenario, 1TB stored, 5TB of egress, 1 million GET and 100,000 PUT requests a month, are computed from each vendor's own published per-unit rates; they are not vendor quotes. Editorial rules are on the Devshot about page.

## FAQ

### What is the cheapest S3-compatible object storage in 2026?

For 1TB stored and 5TB of monthly egress, Wasabi is cheapest, its flat per-TB rate in the table above, verified September 2026, because it charges nothing for egress or requests. Hetzner Object Storage is next at $12.79/mo priced in USD, and Cloudflare R2 follows at $14.85/mo and gives more free requests before any metering starts.

### How much does AWS S3 cost for 1TB with 5TB of egress?

$464.90/mo: $23.00 for storage, $0.90 combined for PUT and GET requests, and $441.00 for 4,900GB of billable egress once the 100GB free allowance is used. Egress is the line that dominates the AWS bill, not storage.

### Is there a free way to run S3-compatible object storage?

Cloudflare R2's free tier covers 10GB of storage and 1 million write plus 10 million read operations a month, which is a real workload for a small app, not a demo. Past that, R2 bills $0.015/GB with no egress fee at all.

### Does Cloudflare R2 really have no egress fees?

Yes, on every storage class, standard and Infrequent Access alike. The costs that do apply are storage per GB and per-operation charges once the free tier of 1 million Class A and 10 million Class B operations a month is spent.

### Why is Google Cloud Storage the most expensive vendor at this scenario?

Its $0.12/GB egress rate is the highest of the three hyperscalers compared here, and its egress free tier is negligible next to 5TB of monthly traffic. Storage pricing is close to AWS and Azure; egress is what separates the three.

### Is Wasabi actually free of hidden fees?

Storage and egress are flat, but Wasabi enforces a 90-day minimum storage duration: deleting an object earlier bills the remaining days as if it were still stored, and objects under 4KB are billed as 4KB regardless of actual size.

### How much does Backblaze B2 cost for a bucket that barely gets read?

Close to its storage price alone. B2 gives free egress up to three times a bucket's average monthly storage, so a 1TB bucket gets 3TB of free egress; only traffic past that adds the $0.01/GB overage that pushed this scenario's total to $26.88.

### Does Hetzner Object Storage price in dollars?

Yes. Hetzner's pricing page defaults to euros (€6.49/mo base for 1TB storage and 1TB egress, plus €1.00/TB after that) but has its own currency switcher that shows USD, matching Wasabi's headline base rate, plus $1.20/TB additional egress, which puts it second-cheapest in this comparison at $12.79/mo for the scenario.

Cite this: Devshot, "Object Storage Pricing Compared (2026): S3 vs R2 vs B2 vs Wasabi at 1 TB", September 2026.

Storage bills change quietly. Devshot's daily brief flags the pricing changes developers act on. Subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cheaper Vercel Alternatives in 2026: Hosting Costs for a 3-Developer Team

URL: https://devshot.co/blog/cheaper-vercel-alternatives
Type: blog
Published: 2026-10-05
Updated: 2026-10-05
Summary: Cloudflare Workers is the cheapest Vercel alternative at $5/mo for a 3-developer team in 2026; Vercel Pro costs $60/mo. Prices verified September 2026.

Guide

## Cheaper Vercel Alternatives in 2026: Hosting Costs for a 3-Developer Team

Cloudflare Workers is the cheapest Vercel alternative at $5/mo for a 3-developer team in 2026; Vercel Pro costs $60/mo. Prices verified September 2026.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated October 5, 2026 · 11 min read

For a 3-developer team pushing 1TB of bandwidth, 5 million function invocations and 100 GB-hours of compute a month, Cloudflare Workers is the cheapest Vercel alternative, at $5/mo, verified on Cloudflare's pricing page, September 2026. Vercel Pro costs $60/mo for that same team: three deploying seats at $20 each. If you want a platform that behaves like Vercel rather than a Workers rewrite, self-hosted Coolify on a $24/mo VPS is the runner-up.

These prices were checked on each vendor's own pricing page on September 24, 2026, in USD. This page prices nine of the ten platforms developers ask about as Vercel alternatives; Hetzner is excluded because its cloud pricing page loads instance prices through a client-side calculator with a location-based currency switch, and no USD figure could be confirmed on the vendor's page today.

Toolradar data: the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) evaluated 298 tools, updated September 2026. At this scenario, the cheapest paid alternative costs 92% less than Vercel Pro for the same team.

## The scenario, and why it matters

Three developers who can all deploy, 1TB of monthly bandwidth, 5 million function calls and 100 GB-hours of compute describes a real production Next.js app with a small team, not a hobby project and not an enterprise account. Every price below is what that workload costs on that vendor, not the vendor's cheapest advertised sticker. A low entry price can still lose this scenario once seats, bandwidth or invocations start metering. Transactional messages add their own line to a developer's bill: see the cheapest SMS API for 50,000 one-time codes a month, which compares Bird, Telnyx, Plivo and Twilio at list rate.

## Cost comparison at the scenario

Tool | Entry price | Cost at scenario | Free plan (limits) | Billing basis | Hidden costs |

[Cloudflare Workers](https://toolradar.com/tools/cloudflare) | Free | $5/mo | 100,000 requests/day, 10ms CPU per call | Flat minimum, then per-request | $0.02 per extra million CPU-ms past 30M included |

Coolify (self-hosted) | Free forever | $24/mo (VPS only) | Full features, no caps | Free platform, you pay for the server | Cloud tier caps out at 2 connected servers before per-server billing starts |

[Fly.io](https://toolradar.com/tools/flyio) | No plan fee | $31 to $131/mo | No ongoing free tier | Pure usage: per-second compute, per-GB egress | Egress rate varies by region |

[DigitalOcean](https://toolradar.com/tools/digitalocean) App Platform | Free (static only) | $41/mo | Static sites only, no compute | Per container instance plus bandwidth overage | Overage billed by the GiB past the container's allowance |

Heroku | $5/mo (Eco) | $50/mo | None; Eco sleeps after 30 min idle | Per dyno, per month | Moving to a performance dyno for real traffic costs 5x more |

Vercel Pro | Free (Hobby) | $60/mo | Non-commercial only, 100GB transfer, 1M invocations | Per deploying seat plus usage past a shared credit | Every extra deploying seat adds to the fee |

Railway | Free | $91.20/mo | One-time trial credit, no ongoing free plan | Workspace fee plus per-second RAM, CPU and egress | RAM, CPU and egress are all metered separately |

Netlify | Free | $140/mo | 300 usage credits/mo, 1 member | Workspace fee plus metered usage credits | Extra credits are sold in packs once the monthly pool runs out |

Render | Free | $196.25/mo | 5GB bandwidth, sleeps on idle | Workspace fee plus compute instance plus bandwidth overage | Bandwidth overage applies past a small included tier |

## The arithmetic, shown once

Vercel's Pro fee includes one deploying seat, so three developers who can all push code means three seats: $20/mo x 3 seats = $60/mo. The scenario's 1TB of transfer and 5 million invocations fit inside the included Flat Rate CDN tier and the per-seat usage credit, so nothing adds to that total.

Netlify's Pro workspace is a flat fee with unlimited members, but every gigabyte and compute-hour draws down a shared pool of 3,000 credits at 20 credits/GB of bandwidth and 10 credits/GB-hour of compute (1TB is treated as 1,000GB here). 1TB of transfer plus 100 GB-hours of compute needs 21,000 credits against that included pool: 20,000 credits + 1,000 credits = 21,000 credits. The shortfall, 18,000 credits at Netlify's $10-per-1,500-credit pack rate, adds $120/mo on top of the $20 base fee, for $140/mo total.

Render bills a workspace fee plus a compute instance, then meters bandwidth past a 25GB allowance at $0.15/GB. A single 2GB instance covers the compute, and the remaining 975GB of transfer (1,000GB minus the 25GB allowance) adds $146.25/mo, landing at $196.25/mo total: $25 workspace + $25 instance + $146.25 bandwidth = $196.25/mo.

## Where the ranking flips

This order holds for three deployers at this bandwidth and invocation volume. At one deployer, Vercel Pro and Netlify Pro both land at the same monthly fee, so seat count is not yet the deciding factor. Past roughly ten deployers, Vercel's per-seat model keeps multiplying while Netlify, Render and Railway stay on one flat workspace charge, so those platforms overtake Vercel earlier the larger the team gets. At low bandwidth, well under 200GB a month, Render's and DigitalOcean's included transfer covers the whole workload and their totals drop close to their base compute fee, ahead of Cloudflare Workers once you also count the migration work Workers needs. For a side project that only needs a small server, Cloud Free Tiers Compared shows what AWS, Google Cloud, Azure and Oracle keep free.

## Cloudflare Workers: cheapest, but not a drop-in replacement

Workers Paid includes 10 million requests and 30 million CPU-milliseconds a month for its flat minimum, comfortably above this scenario's invocation count, and [Cloudflare](https://toolradar.com/tools/cloudflare) charges no egress fee at all, so the bandwidth is free. The catch: Workers is not a Vercel-style build pipeline. Running Next.js on it needs the OpenNext Cloudflare adapter, and some Vercel-specific features, ISR write semantics among them, behave differently once ported. Teams whose app is already adapter-ready get the cheapest number on this page; teams mid-build should weigh the migration time against the savings.

## Coolify: free software, paid server

[Coolify](https://toolradar.com/tools/coolify) is open source and free forever when self-hosted, with no feature gate. Coolify Cloud adds a managed control plane for its own small monthly fee, connecting up to two servers before per-server billing kicks in. Either way, Coolify never bills for bandwidth or compute: that meter belongs to the server underneath it. A DigitalOcean Droplet sized for this scenario, 4GB RAM, 2 vCPUs and 4,000GiB of included transfer, costs $24/mo and comfortably absorbs 1TB of bandwidth with no overage. That server cost is the real Coolify bill, and it beats every managed platform on this page except Cloudflare Workers.

## Fly.io: the number depends on your region

Fly.io has no monthly plan fee; every dollar is usage. A shared vCPU with 2GB of RAM running continuously costs about $11/mo at Fly's per-second compute and memory rates. Bandwidth is the wildcard: Fly's own pricing page lists egress between $0.02 and $0.12 per GB depending on region, so 1TB (1,000GB) of transfer lands anywhere from $20 to $120/mo. Total cost at this scenario runs $31 to $131/mo, with the low end reserved for apps whose traffic stays in cheap regions. The same kind of bill applies to search, and Algolia alternatives prices the search API meters for 100,000 records and 500,000 searches a month.

## Vercel Pro's seat math is the whole story

Once a team ships from more than one account, Vercel's invoice grows with headcount before it grows with traffic. The base fee buys one deploying seat and a matching usage credit, and each additional seat adds another fee and another credit, which is why bandwidth and invocations barely move the number for three seats. Vercel vs Netlify covers where that seat math changes the comparison with Netlify's flat workspace fee. Viewers who only open preview URLs stay free, so trimming deploy rights from people who do not merge code usually fixes a bloated invoice faster than switching hosts.

## DigitalOcean, Heroku and Railway: usage-metered middle ground

DigitalOcean's App Platform charges per container instance rather than per seat: a 1 vCPU, 2GB container covers the compute, and the remaining transfer past its included allowance adds roughly $16/mo more at its per-GiB overage rate. It is not serverless and does not autoscale on invocations, so it suits a steady workload better than a spiky one. The DigitalOcean review covers the rest of the platform.

Heroku charges per dyno with no bandwidth meter published on its pricing page, so the scenario's cost is really the dyno tier: the cheapest paid dyno sleeps on idle and is not a production choice, and a mid-tier always-on dyno is the closer match to a small production app, with heavier traffic pushing teams to a performance dyno at several times the price. Heroku alternatives covers the migration case in more depth.

[Railway](https://toolradar.com/tools/railway) has no per-seat charge and unlimited workspace members on Pro, and instead meters RAM, CPU and egress per second against an included usage credit. For this scenario, a 2GB/1vCPU service plus 1TB of egress lands near $91/mo once that credit is spent, which is why Railway suits variable workloads that scale down between bursts better than it suits a team that is only adding developers. Best VPS hosting for developers and best serverless platforms cover the wider field this scenario sits inside.

## Hidden costs that do not show up on the plan card

- Vercel: each extra deploying seat adds another full seat fee; SAML SSO is a $300/mo add-on; bandwidth overage past the included Flat Rate CDN tier is $0.15/GB.
- Netlify: additional usage credits are sold in 1,500-credit packs for $10 (about two-thirds of a cent each), and bandwidth alone consumes 20 credits per GB.
- Render: bandwidth overage is $0.15/GB past the plan's included transfer; persistent disks add $0.25/GB/mo; SAML and SCIM require the $499/mo Scale workspace.
- Railway: RAM, CPU and egress are all metered per second with no flat compute allowance beyond the included credit.
- Fly.io: egress swings from $0.02 to $0.12/GB by region; a dedicated IPv4 address is a small add-on on top of compute.
- DigitalOcean: egress overage is billed by the GiB past the container's allowance; a dedicated egress IP adds a flat monthly fee per app.
- Coolify: the software is free, but Cloud's base tier stops at two connected servers, and self-hosting means you patch and monitor the box yourself.
- Heroku: no bandwidth rate is published at all, and moving to a performance dyno for real traffic is a multiple of the entry price.
- Cloudflare Workers: CPU time past the included allowance is billed per extra million milliseconds, and porting a Next.js app off Vercel's build pipeline is real engineering time, not a line item, but it is still a cost.

## How we checked these prices

Louis Corneloup, founder of Devshot, Dupple and Toolradar, checked each vendor's own pricing page and docs on September 24, 2026: [Vercel](https://vercel.com/pricing), Netlify, [Cloudflare Workers](https://developers.cloudflare.com/workers/platform/pricing/), [Render](https://render.com/pricing), [Railway](https://railway.com/pricing), [Fly.io](https://docs.fly.io/about/pricing/), [DigitalOcean](https://www.digitalocean.com/pricing/app-platform), [Coolify](https://coolify.io/pricing) and [Heroku](https://www.heroku.com/pricing/). Costs at the scenario are computed from each vendor's own published per-unit rates for 3 deploying developers, 1TB (treated as 1,000GB) of monthly bandwidth, 5 million function invocations and 100 GB-hours of compute; they are not quotes from the vendors. The same method, one stated scenario with dated vendor prices, applies to monitoring in Datadog alternatives at 10 hosts. Editorial rules are on the Devshot about page.

## FAQ

### What is the cheapest alternative to Vercel for hosting a Next.js app in 2026?

For a 3-developer team at 1TB of bandwidth, 5 million invocations and 100 GB-hours of compute a month, Cloudflare Workers is cheapest at $5/mo, verified September 2026. It needs the OpenNext adapter to run Next.js, so teams not already on that adapter should weigh the migration work against the savings.

### How much does Vercel cost for a 3-developer team?

Vercel Pro costs $20/mo per deploying seat, so three developers who can all push to production cost $60/mo before any usage past the included credit. Viewer-only teammates who cannot deploy stay free, so trimming deploy rights to the people who actually merge code is the fastest way to cut the bill.

### Is there a free way to run what Vercel runs?

Coolify is free forever when self-hosted, with no feature limits, but it still needs a server. A DigitalOcean Droplet sized for this scenario costs $24/mo, making that the real floor for a Vercel-like setup rather than nothing.

### Is Netlify cheaper than Vercel for a small team?

At one deploying developer, Netlify Pro and Vercel Pro cost the same. At three developers pushing 1TB of bandwidth, Netlify's flat workspace fee stays put on headcount, but its usage-credit system adds $120/mo for that bandwidth and compute load, landing near $140/mo, well above Vercel's fee for the same team.

### Does Heroku still make sense for a Next.js team?

Heroku charges per dyno rather than per seat or per gigabyte, and it publishes no bandwidth rate at all. A mid-tier always-on dyno covers a small production app, but Heroku was built before edge rendering and ISR were standard, so teams that lean on those Next.js features usually find Vercel, Netlify or Cloudflare a closer fit.

### Where does the ranking change at a bigger team?

Past roughly ten deploying developers, Vercel's per-seat fee keeps climbing while Netlify, Render and Railway keep charging a single flat workspace fee regardless of headcount, so those platforms overtake Vercel on cost the larger the deploying team gets.

### Is Fly.io cheaper than Vercel?

It depends on region. Fly.io has no plan fee, only usage: a small compute charge plus an egress bill that moves with where the traffic originates, for a total of $31 to $131/mo at this scenario. In cheap regions it undercuts Vercel; in expensive ones it does not.

### What does DigitalOcean App Platform cost against Vercel?

A container sized for this scenario runs about $25/mo and includes most of the transfer allowance; the remainder at this scenario's 1TB volume adds roughly $16/mo more, for about $41/mo total, well under Vercel's fee for the same team, with the tradeoff that App Platform runs a persistent container rather than Vercel's per-request serverless model.

Cite this: Devshot, "Cheaper Vercel Alternatives in 2026: Hosting Costs for a 3-Developer Team", September 2026.

Devshot is the daily brief on the tools and bills developers have to act on. Get it at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cursor Pricing in 2026: Every Plan, Real Costs and Hidden Fees

URL: https://devshot.co/blog/cursor-pricing
Type: blog
Published: 2026-09-27
Updated: 2026-09-27
Summary: Cursor pricing in 2026: Pro is $20/mo, Pro+ is $60/mo, and Ultra is $200/mo. Teams seats are $40 or $120 before on-demand usage.

Guide

## Cursor Pricing in 2026: Every Plan, Real Costs and Hidden Fees

Cursor pricing in 2026: Pro is $20/mo, Pro+ is $60/mo, and Ultra is $200/mo. Teams seats are $40 or $120 before on-demand usage.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 27, 2026 · 17 min read

Key facts7 plans compared: Hobby, Pro, Pro+, Ultra, Teams Standard, Teams Premium, Enterprise

- Updated: September 27, 2026
- Prices as of: September 2026

- 7 plans compared: Hobby, Pro, Pro+, Ultra, Teams Standard, Teams Premium, Enterprise
- Hobby (best for: A short trial of Tab and Agent): Free
- Pro (best for: Light agent use, not an all-day loop): $20/mo
- Pro+ (best for: Agent open on a normal workday): $60/mo
- Ultra (best for: Several agents or automations): $200/mo

Cursor Pro is $20/mo, and that sticker is the wrong budget if Agent stays open through a normal workday. Cursor's own pricing page points daily agent users at Pro+ ($60/mo) and power users at Ultra ($200/mo). A Teams Standard seat is $40/user/mo, or $32 on the annual plan.

These figures were read on Cursor's pricing page and docs on September 24, 2026. Cursor is an AI editor, forked from VS Code, sold as a monthly seat plus two usage pools that reset with the billing cycle. The seat is the number in the ad, while the pool, the model you pin, and on-demand usage after the pool are what show up on the invoice.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 153 tools, updated September 2026. The product writeup on this site is the Cursor review, and the directory record is [Cursor on Toolradar](https://toolradar.com/tools/cursor).

Plan | Price | Best for | What changes the bill |

Hobby | Free | A short trial of Tab and Agent | Limited Agent requests, Auto model, no card |

Pro | $20/mo | Light agent use, not an all-day loop | Two pools, unlimited Tab, then on-demand |

Pro+ | $60/mo | Agent open on a normal workday | The tier Cursor names for daily agent users |

Ultra | $200/mo | Several agents or automations | The tier Cursor names for power users |

Teams Standard | $40/user/mo, or $32 annual | Shared billing, SSO, privacy mode | The allowance stays on that seat |

Teams Premium | $120/user/mo, or $96 annual | The heavy users on the same team | 5x the Standard agent allowance |

Enterprise | Custom quote | Invoice, SCIM, pooled usage | Sales, not checkout |

## Hobby is a real install, with a short leash

Hobby is free and does not ask for a card. You get Agent, Chat, and Tab on the Auto model, plus access to Composer, with a limited number of Agent requests. That trial shows whether the editor is worth leaving VS Code for, and it will not cover a week of multi-file edits.

The Agent cap is the leash, not a missing editor. A refactor that touches a handful of files can spend the free allowance before the tests are green. Pro and above are where Cursor states that Tab completions are unlimited, which matters when Tab is the daily habit. Cursor also sells a Start plan in India that is not priced in USD, so a US checkout begins at Hobby or Pro. Students should check the free route first: Cursor's legacy student discount closed to new sign-ups on June 25, 2026, so see the Cursor student discount guide for the routes that remain.

## Pro is the number people repeat

Pro is the individual plan on the public table, billed monthly. It includes unlimited Tab completions, extended agent limits, Cloud Agents, and access to Bugbot. Usage splits into two pools that reset with your billing cycle: Cursor models (Grok 4.7, Grok 4.6, Grok 4.5, and Composer 2.5) and other models, which draw at that model's API rate. A third-party pick spends the other pool even while Cursor's own pool still has room.

Stay on Pro when Agent is occasional, living in Tab and opening Agent for a single function or a failing test. Cursor's docs put a daily agent user in a higher spend band than this plan's sticker, which is why the pricing page tells that person to step up. On-demand usage, if you turn it on, continues at the same API rates and does not downgrade the model, so you pay more for the same quality. The public table does not print a yearly dollar amount for Pro, Pro+, or Ultra. The help center lets you switch an individual plan to yearly, and it does not publish the discounted rate.

## Pro+ is the plan for a workday in Agent

Pro+ is the middle individual tier, and the editor matches Pro, so the higher bill buys a larger pool. Cursor's pricing FAQ recommends it for daily agent users, and the models doc says that habit typically runs $60 to $100 a month of total usage. Cursor does not publish how many included dollars separate the two tiers, so the step up is headroom before on-demand, not a new feature list.

Buy Pro+ when you can name the afternoon Pro ran out. That afternoon is a multi-file change, a test loop, or an agent that keeps a terminal open while you review the diff. If you cannot name that moment, the step up is idle capacity. Fast mode and higher effort on first-party models spend the pool faster than the default, so a Fast workday costs more than Composer at the standard rate.

## Ultra is a capacity plan

Ultra is the top individual plan, and the product surface matches Pro, so the editor does not change. Cursor recommends it for agent power users, and the models doc says people running several agents or automations are often past a spend that lines up with this sticker. Requests are not slowed down when you spill into on-demand, and you pay the API rate instead.

Skip Ultra until Pro+ is the thing blocking a merge. A solo maintainer who runs one careful agent pass per feature does not need this tier. Someone who fans out cloud agents, or who leaves automations running against a repo, is the buyer. An upgrade from a lower individual plan takes effect immediately, and any credit is based on unused included usage, not on days left in the month. A downgrade waits until the period ends, so you cannot drop the tier before renewal.

## Teams charges a seat that cannot lend its pool

Teams is the self-serve company plan for shared billing, and Standard is the monthly or annual rate in the table. Premium is three times that monthly seat for five times the Standard agent allowance, at $120/user/mo or $96 on the annual plan, for seats that already exhaust Standard. Cursor's June 2026 pricing note says it expects the first-party pool on Premium to cover a heavy agent month for 99% of users. Treat that as the company's forecast, not a measurement of your repo.

Each paid seat gets its own Cursor-models pool and its own third-party pool. Usage does not move to a teammate who shipped nothing that week. Enterprise is the plan where usage is pooled, so leftovers on Teams cannot be shared. On-demand is on by default for Teams, billed in arrears, so a seat that looks flat becomes a token invoice when someone pins a frontier model. Admins can set a team-wide spend limit, while per-member spend limits are an Enterprise control, so a self-serve admin can cap the company and not one person.

Teams also adds centralized billing, a marketplace for internal rules and plugins, shared cloud-agent context, usage analytics, team-wide privacy mode, and SAML or OIDC SSO. Buy Teams for those controls when individual seats would already cover the usage. An unpaid admin seat has no Cursor access and no seat fee, which is the right slot for finance or IT. Adding someone mid-cycle is pro-rated, and removing someone who already used credits keeps that seat occupied until the cycle ends, so a Friday offboarding does not shrink this month's invoice.

The annual rates ($32 and $96) come from Cursor's [Teams pricing update](https://cursor.com/blog/teams-pricing-june-2026), which applied to new customers immediately and to renewing customers on cycles starting July 1, 2026. You cannot flip a yearly plan back to monthly in the middle of the term, so the discount locks the year.

## Enterprise is a quote because the pool is shared

Enterprise has no public seat price, so sales has to name the seat. It is the tier for pooled usage, invoice or PO billing, SCIM, audit logs, service accounts, repository and MCP controls, and priority support. Cursor's recommendation is blunt: stay on Teams if you can self-serve, and talk to sales when you need the pool, the invoice, or the security controls.

The shared pool is the reason the quote exists. On Teams, one heavy user cannot spend a quiet teammate's leftover allowance, so you buy Premium seats for the people who burn them. On Enterprise, the allowance is shared, which helps a team whose usage is uneven and hurts a team that never watches the dashboard. Third-party models still pick up the token fee described below. Contact sales is the price, and there is no list rate to put next to a Teams seat.

## The model you pin moves more money than the seat

Auto has three modes, Cost, Balance, and Intelligence, and none of those names is a discount. Every mode bills at the list price of the model the router actually picked. On Teams and Enterprise, a third-party model also takes a Cursor Token Rate of $0.25 per million tokens, on input, output, and cache. The rate applies when Auto lands on that model and when you bring your own key. Grok and Composer are exempt from that rate, so that fee skips first-party models. First-party rates below are per million tokens, from Cursor's models doc, September 2026.

Model | Input | Output | Why it matters |

Composer 2.5 | $0.50 | $2.50 | The cheap first-party default for everyday agent work |

Composer 2.5 Fast | $3 | $15 | Much faster, and a different bill |

Grok 4.7 | $2 | $6 | Cursor's frontier model, jointly trained with SpaceXAI |

Pin Composer when the task is a known edit, such as a rename, a test, or a small migration. That work should not sit on a third-party frontier model that also owes the token rate. Grok 4.7 above 256,000 input tokens bills at twice the standard rate, up to 500,000, and its Fast mode is twice the standard price. Regional data residency adds 10% on eligible model pricing for any team that turns it on. Max Mode, which stretches context and adds 20% on top of the API rate, exists only on legacy request-based plans, so a current usage-pool account should ignore that line item.

Third-party models, including Claude and GPT variants in the picker, draw the other-models pool at the provider's API rate. On a team plan that rate is not the whole line. The $0.25 fee sits on top, which is how two seats with the same agent habit diverge once one of them leaves Auto's first-party route.

## What ten developers pay next to Copilot, Devin, and Claude

Seat math is where these tools stop looking interchangeable on a real invoice. Prices below were checked on each vendor's own pages on September 24, 2026. The product choice, editor versus plugin, is in Cursor vs GitHub Copilot, and this table is only the invoice.

Buyer | Cursor | GitHub Copilot | Devin Desktop | Claude Code |

Light individual use | Pro | Pro, $10/mo, 1,500 credits | Matches Cursor's individual monthly price | Monthly Pro matches Cursor; annual is $17/mo |

Agent most workdays | Pro+ | Pro+, $39/mo, 7,000 credits | Pro quota, then on-demand credits | Max 5x at $100/mo |

Agents all day | Ultra | Max, $100/mo, 20,000 credits | Max, weekly quota, no daily cap | The higher Max tier, same sticker as Ultra |

10 people, standard monthly | $400 | Business, $190 | Ten full seats, same monthly total as Cursor Standard | Team standard, $250 |

10 people, cheaper interval | $320/mo on annual Standard | Business stays a monthly seat | No annual rate on the billing doc | $2,400 for the year on annual standard seats |

[GitHub Copilot](https://docs.github.com/en/copilot/concepts/billing/individual-plans) Pro is the cheap individual entry: $10/mo buys 1,000 base credits plus a 500-credit flex allotment, 1,500 in total. Business is $19 per granted seat per month with 1,900 credits, which is why ten seats are $190 before extra credits, and one credit equals $0.01. Completions and next-edit suggestions are not billed in credits on any paid plan, and they stay unlimited.

Credits die at 00:00 UTC on the first of the month, which may not be the invoice date. Copilot is not sold for GitHub Enterprise Server, so that install cannot buy it. The GitHub Copilot review covers the product, and the directory page is [Copilot on Toolradar](https://toolradar.com/tools/github-copilot).

Devin Desktop is the editor on Cognition's current price list, and anyone still pricing Windsurf is looking at this product. Pro matches Cursor's individual monthly price, and Max matches Ultra's sticker. The [billing doc](https://docs.devin.ai/admin/billing/self-serve) sets a Teams minimum of $80/mo, so a smaller group cannot open Teams. A full seat is the same monthly rate as Cursor Standard and includes Desktop plus a Pro-like quota. A flex seat has no seat charge, spends shared on-demand credits, and does not include Desktop, so it is not an editor seat.

Once two full seats are active, the doc bills headcount times the full-seat rate, so ten full seats land on the same monthly total as ten Cursor Standard seats. The public card can be read as a platform fee plus seats, so budget from the billing doc, not from that reading. SSO is an Enterprise conversation, so self-serve will not include it. The product record, apart from this invoice, is [Devin on Toolradar](https://toolradar.com/tools/devin).

Claude Code is not a separate IDE subscription, and it is absent on the free plan. It is included on Claude's paid plans and shares one usage pool with chat, so a day in the terminal can exhaust the plan before anyone opens chat. Pro's annual rate is $17/mo, with the year paid up front. Max 5x is $100/mo and Max's higher tier matches Ultra, both monthly only, on Anthropic's [Max plan page](https://support.claude.com/en/articles/11049741-what-is-the-max-plan).

Team standard seats are $25/user/mo monthly, with a lower annual rate, a two-seat minimum, and a 150-seat cap. A solo developer cannot buy it, and a company past the cap has to leave. Allowances are per person: 1.25x Pro on a standard seat and 6.25x Pro on a premium seat. Ten monthly standard seats are $250, and ten annual standard seats are $2,400 for the year. How the terminal should be set up is in how to use Claude Code, the review is the Claude Code review, and the head-to-head with this editor is [Claude Code vs Cursor](https://toolradar.com/blog/claude-code-vs-cursor-2026). For a side-by-side with Claude Code, including what a 10-seat team pays on each, read Claude Code vs Cursor 2026.

## Fees that never appear on the plan card

The plan price is the floor, not the month you will recognize. On Teams, on-demand starts enabled, Bugbot reviews bill from on-demand spend rather than from a separate seat, and the token rate hits every third-party request. Individuals burn included usage first when Bugbot runs, then on-demand, so the review spends the pool before any extra charge. Bugbot moved to usage-based billing in May 2026, and the plan card still lists it as included, which is why a review-heavy month surprises people who trusted the card.

A removed teammate can keep costing money until the renewal date if they used credits before they left, and a new teammate costs a pro-rated seat immediately. Yearly individual discounts are not printed, and taxes are extra on every Cursor price. Subscriptions bought anywhere except cursor.com are unauthorized, and Cursor says those accounts can be suspended.

Privacy mode, on by an individual or enforced for a team, is the control that keeps code out of training by Cursor and by model providers, and it does not cap spend. Team-wide privacy is a Teams feature, and the training setting does not touch the invoice.

Devshot data: 28,000 developer readers get this brief, and the audience file updated September 20, 2026 puts the open rate at 30%. They compare a seat the way they compare a cloud bill, so a plan card with on-demand left on is the month-two surprise. Rate changes for this audience land in the Devshot brief.

## Which plan is worth paying for

For one person, start on Hobby, then Pro, and move to Pro+ only after a real week runs out. Ultra is for several concurrent agents, not for a nicer autocomplete. If Tab in the editor you already have is the whole job, Copilot Pro is the smaller bill, and the credit meter does not touch completions.

For a team, put Premium only on the people who will burn it. Allowances do not pool, so a Premium seat for someone who reviews pull requests twice a week is the usual waste. Turn on a team spend limit before the first on-demand cycle. If you need one shared pool, an invoice, or SCIM, price Enterprise only after you know which models the team will pin. A group that wants a terminal agent and can share a chat allowance should price Claude before they price a second editor.

If you are still choosing the tool, start with AI coding assistants, then AI-assisted coding for the workflow and AI for developers for where review time goes. Prompt-to-app tools are a different budget, covered in vibe coding tools.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We checked Cursor's [pricing page](https://cursor.com/pricing), the [models and pricing doc](https://cursor.com/docs/models-and-pricing), the [Teams pricing doc](https://cursor.com/docs/account/teams/pricing), the [June 2026 Teams note](https://cursor.com/blog/teams-pricing-june-2026), and the [Bugbot doc](https://cursor.com/docs/bugbot) on September 24, 2026.

Copilot figures come from GitHub's [individual plans](https://docs.github.com/en/copilot/concepts/billing/individual-plans) and [individual usage billing](https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-individuals). Devin figures come from [devin.ai/pricing](https://devin.ai/pricing) and the [self-serve billing doc](https://docs.devin.ai/admin/billing/self-serve). Claude figures come from [claude.com/pricing](https://claude.com/pricing), the Max article linked above, and the [Team plan article](https://support.claude.com/en/articles/9266767-what-is-the-team-plan). The 153 tool count is from Toolradar's live AI coding ranking that same month. More on how we publish is on the Devshot about page.

## FAQ

### How much does Cursor cost in 2026?

Hobby is free, and it only covers a short trial. Pro is $20/mo, Pro+ is $60/mo, and Ultra is $200/mo, all shown as monthly prices. Teams Standard is $40/user/mo, or $32 if you pay annually. Teams Premium is $120/user/mo, or $96 annually, for five times the Standard agent allowance. Enterprise is a custom quote with no public seat price. On-demand usage and, on Teams, a third-party token fee sit on top of those seats, so the card is the floor.

### Is Cursor free?

Yes, Hobby includes Agent, Chat, and Tab on Auto, plus Composer, with limited Agent requests and no card, which is a trial rather than a working week. Paid plans are where Tab is unlimited and the agent pool is large enough for daily work. The India-only Start plan is not a US price.

### What is the difference between Cursor Pro, Pro+, and Ultra?

All three include unlimited Tab, Cloud Agents, Bugbot access, and the two usage pools. Pro is the light-agent plan for someone who mostly stays in Tab. Pro+ is the one Cursor recommends once Agent is part of the workday, and Ultra is the power-user plan for several agents or automations. Cursor does not publish the included-usage gap between those tiers, and the editor does not change, only the headroom does.

### How much is Cursor for a team?

You pay per active paid seat, Standard and Premium can be mixed, and an unpaid admin does not consume a paid seat. Usage resets each cycle and does not transfer between people. On-demand is enabled by default, and annual billing is the cheaper interval, but you cannot switch that yearly plan back to monthly mid-term. Ten Standard seats are $400 for a month, or $320 a month on the annual rate.

### Does Bugbot cost extra?

Reviews do not require a separate Bugbot seat on current plans. Bugbot reviews pull requests and, for an individual, spends included usage before on-demand, while on Teams, Bugbot bills from on-demand spend. The old seat-based Bugbot price ended in the May 2026 change. A review bot that is not this one belongs with AI for code review.

### Is Cursor cheaper than GitHub Copilot?

At the individual entry, no, because Copilot Pro is $10/mo with 1,500 credits, and completions on paid Copilot plans are unlimited and off that meter. Cursor's individual sticker is higher, and a daily agent user is pushed to Pro+. Ten Copilot Business seats are $190 a month, and ten Cursor Standard seats are $400 a month before token fees. Cursor costs more as a seat and buys a dedicated editor rather than a plugin.

### What happens when Cursor hits a usage limit?

You can wait for the monthly reset, move to a higher plan, or keep going on on-demand at the same API rates, and quality and speed are not stepped down. On Teams, a person who exhausts third-party usage is switched toward the Cursor-models pool, and further third-party use bills on-demand if that switch is left on. One teammate hitting a cap does not reduce anyone else's seat.

Cite this: Devshot, "Cursor Pricing in 2026: Every Plan, Real Costs and Hidden Fees", September 2026.

Devshot covers the tools and bill changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Supabase Pricing in 2026: Every Plan, Real Costs and Hidden Fees

URL: https://devshot.co/blog/supabase-pricing
Type: blog
Published: 2026-09-27
Updated: 2026-09-27
Summary: Supabase pricing in 2026: Pro is $25/month per organization, not per seat. Team is $599/month. The compute credit covers one Micro project.

Guide

## Supabase Pricing in 2026: Every Plan, Real Costs and Hidden Fees

Supabase pricing in 2026: Pro is $25/month per organization, not per seat. Team is $599/month. The compute credit covers one Micro project.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 27, 2026 · 17 min read

Key facts4 plans compared: Free, Pro, Team, Enterprise

- Updated: September 27, 2026
- Prices as of: September 2026

- 4 plans compared: Free, Pro, Team, Enterprise
- Free (best for: A prototype that can sleep): $0
- Pro (best for: One production project on Micro): $25/mo
- Team (best for: SOC 2, longer logs, project roles): $599/mo
- Enterprise (best for: Uptime SLA, private Slack, PrivateLink): Custom quote

Supabase does not charge per seat: Pro is $25/month for the organization, Team is $599/month, and Enterprise publishes no list price. Dashboard members are unlimited, so ten engineers do not multiply the subscription. The bill that moves is compute: one instance per project, and the included credit covers a single Micro instance.

Prices below come from Supabase's pricing page and platform docs, checked September 24, 2026. Each organization holds one plan, and every project is its own Postgres server, with auth, storage, realtime, and edge functions on that same instance, while compute is metered by the hour on top of the plan fee. The product choice against Firestore is in Supabase vs Firebase, and this page is the invoice.

Toolradar data: the [managed databases ranking](https://toolradar.com/best/managed-databases) evaluated 23 tools, updated September 2026. The writeup on this site is the Supabase review, and the directory record is [Supabase on Toolradar](https://toolradar.com/tools/supabase). A catalog-side pass at the same invoice is [Supabase pricing on Toolradar](https://toolradar.com/blog/supabase-pricing-2026).

Plan | Price | Best for | What changes the bill |

Free | $0 | A prototype that can sleep | Pause after a week idle, 2 active projects |

Pro | $25/mo | One production project on Micro | Quotas, then overage if the spend cap is off |

Team | $599/mo | SOC 2, longer logs, project roles | Same usage quotas as Pro, no spend cap |

Enterprise | Custom quote | Uptime SLA, private Slack, PrivateLink | Sales, not checkout |

## Free is a real Postgres box that goes to sleep

Free is a working project, not a demo login. You get unlimited API requests, 50,000 monthly active users, 500 MB of database, 5 GB of egress, 5 GB of cached egress, and 1 GB of file storage. The instance is shared CPU with 500 MB of RAM, smaller than Micro on a paid org, with no card, no daily backups, no email support, and a 50 MB file upload cap.

The leash is inactivity: Supabase pauses a free project after one week without use. You can pause it yourself, and paused projects do not count toward the cap, while two projects may be active. That cap is counted across every Owner and Admin, including people who already used a free-project allowance elsewhere, so a second owner can block a launch.

Exceeding a quota long enough brings fair-use restrictions: the project can pause, the database can go read-only, or API calls can return 402. Paid and free projects cannot share an organization, so a database that must stay awake and a scratch project that should stay free are two orgs until you transfer one. Compute on Free is the Nano size at no hourly charge, and Supabase says those free resources can change.

## Pro is an organization fee, then a project meter

Pro is the self-serve plan, charged up front once a month, and the fee is not a seat and it is not a ceiling. Quotas on the pricing page are 100,000 monthly active users, 8 GB of disk per project, 250 GB of egress, 250 GB of cached egress, and 100 GB of file storage. Email support, 7-day backups, 7-day logs, and unbranded auth email start here, API requests stay unlimited, and uploads can go to 500 GB.

Past those quotas the unit prices stay small until the app is not:

Meter | Included on Pro | After the quota |

Monthly active users | 100,000 | $0.00325 per MAU |

Disk | 8 GB per project | $0.125 per GB |

Egress | 250 GB | $0.09 per GB |

Cached egress | 250 GB | $0.03 per GB |

File storage | 100 GB | $0.0213 per GB |

Edge Function invocations | 2 million | $2 per million |

Realtime messages | 5 million | $2.50 per million |

Realtime peak connections | 500 | $10 per 1,000 |

Cached egress is a storage CDN hit and the cheaper bandwidth line. Uncached egress is one pool for database, auth, functions, realtime, the pooler, and log drains. The spend cap, on by default, blocks overage on those rows instead of billing them, and turning it off is how the unit prices reach the card.

One Micro project stays inside the plan fee, because that fee includes a monthly compute credit equal to one Micro instance. Supabase's own example of two Micro projects lands at $35/month after that credit is applied once. The [billing FAQ](https://supabase.com/docs/guides/platform/billing-faq) puts three Micro projects at $45/month. There is no official annual discount, a credit top-up can cover future months, and it will not pay an invoice that is already overdue.

## Compute is billed per project, by the hour

Every project is a virtual machine running Postgres, and you pay for the size you selected whether or not the database is busy. Paused projects drop off the meter, and a partial hour is billed as a full hour. Hourly rates are in the [compute doc](https://supabase.com/docs/guides/platform/manage-your-usage/compute), and a 730-hour cycle is the month used in Supabase's invoice examples. The monthly list below is the one on that pricing page.

Size | Monthly list | Shape | Direct / pooler connections |

Micro | $10 | Shared, 1 GB RAM | 60 / 200 |

Small | $15 | Shared, 2 GB RAM | 90 / 400 |

Medium | $60 | Shared, 4 GB RAM | 120 / 600 |

Large | $110 | Dedicated, 2 vCPU, 8 GB | 160 / 800 |

XL | $210 | Dedicated, 4 vCPU, 16 GB | 240 / 1,000 |

2XL | $410 | Dedicated, 8 vCPU, 32 GB | 380 / 1,500 |

Sizes continue through 16XL, and anything larger is a sales conversation. Micro through Medium are shared, and Large is the first dedicated size, which is the upgrade when connection caps or noisy neighbors are the problem.

The credit applies to compute only, and it does not roll over, so it will not pay for a read replica, a branch, disk, egress, or IPv4. One project left on Small for a full cycle totals $30 in Supabase's published example, after the credit. Three Small projects in that doc total $60 after the credit, because two of the three instances are uncovered. A mid-cycle resize bills each size for the hours it ran, settled at the next reset, with no charge on the day you click.

Nano is the free-plan size, and on a paid org a leftover Nano bills at the Micro rate. You cannot launch a new Nano once the org is paid, and size changes are not automatic, so move a leftover Nano to Micro when you can take the restart.

Branching, the alternative to a second project, is $0.01344 per branch per hour, the same hourly rate as Micro, and credits do not touch it. At that rate, 730 hours comes to $9.81, on top of a primary whose credit is already spent, and deleting the branch stops the meter.

## Team buys compliance, not a larger quota

Team is the same organization shape at a much higher subscription. The pricing-page comparison gives it the same disk, egress, storage, MAU, function, and realtime quotas as Pro, at the same overage rates, so the quotas do not grow. What changes is SOC 2 and ISO 27001, project-scoped and read-only roles, priority email with an SLA, 14-day backups, 28-day logs, and platform audit logs. Dashboard SSO says to contact them, HIPAA is a paid add-on with no printed fee, PrivateLink is included, and an uptime SLA is not on this plan.

The spend cap does not exist on Team, and that switch is Pro-only, so quota overages bill. Team still gets the same one-project compute credit, and the add-ons below bill either way. Stay on Pro until a security review asks for the report, the role split, or the longer log window. Ten developers still do not multiply the subscription, and ten projects will.

## Enterprise is a quote because the contract is the product

The page lists a designated support manager, an uptime SLA, PrivateLink, round-the-clock support, a private Slack channel, and help with security questionnaires. Disk, egress, MAU, storage, functions, and realtime are marked custom. Point-in-time recovery can pass 28 days, and log retention is 90 days.

Nothing on the public page sits next to the Team subscription except a sales form, so buy this tier for the SLA or a quota Team will not raise. A company that only needs SOC 2 should price Team first. Invoices stay in USD, card is the default, and a larger upfront payment is a support ticket rather than a checkout option.

## The spend cap ignores the lines people opt into

The spend cap is on by default on Pro, which is why a buggy loop does not become an egress invoice. While it is on, crossing a covered quota stops that usage instead of charging it, and past a grace period the fair-use restrictions follow. The billing FAQ says a later overrun can skip a fresh grace period if you already blew the quota once. Covered meters include disk size, egress, edge invocations, logs, MAU, realtime, storage size, and image transformations.

Compute is absent from that list, and so are branching, read-replica compute, custom domains, extra disk IOPS, extra throughput, IPv4, log drains, phone MFA, and point-in-time recovery. Those count as choices you turned on, so a cap that still looks enabled will not absorb a Large instance, a replica, or a branch that ran all month. The rules for that split are in the [cost control doc](https://supabase.com/docs/guides/platform/cost-control).

Add-on | Price | Spend cap |

Point-in-time recovery | $100 per month per 7 days of retention | No |

Custom domain | $10 per domain per month per project | No |

Phone MFA | $75 per month for the first project, then $10 per month per extra project | No |

Log drains | $60 per drain per month, plus $0.20 per million events and $0.09 per GB | No |

IPv4 | $0.0055 per hour, listed as $4 per month | No |

Image transformations | 100 origin images, then $5 per 1,000 | Yes, the overage |

General-purpose disk includes 8 GB, 3,000 IOPS, and 125 MB/s, extra IOPS are $0.024 each, and extra throughput is $0.095 per MB/s. High-performance disk is a different product at $0.195 per GB, with IOPS at $0.119, and it is not the included 8 GB. A read replica mirrors the primary's compute, disk, IOPS, throughput, and IPv4, and replica compute sits outside the credit and the cap.

The [IPv4 doc](https://supabase.com/docs/guides/platform/manage-your-usage/ipv4) shows a full month of the add-on as its own line, and every replica gets an address when the primary has one. SAML for your own users, separate from dashboard SSO, includes 50 monthly active users and then $0.015 per MAU, while social login does not use that rate.

## What ten developers pay next to Firebase, Neon, and PlanetScale

None of these four bills a seat, so the useful question is which meter runs when the database stays on. Prices were checked on each vendor's pages on September 24, 2026. Product fit, apart from this invoice, is the backend-as-a-service list and the Postgres hosting list.

Buyer | Supabase | Firebase | Neon | PlanetScale |

Prototype, no card | Free, pauses after a week | Spark, no-cost quotas | Free, 100 CU-hours per project | No free plan |

One small database, always on | Pro fee, Micro covered by the credit | Blaze, no plan fee, then usage | 0.25 CU for 730 hours is $19.35 before storage | Single-node PS-5 in us-east-1 is $5/mo |

Auth, files, and functions | Included on the org plan | Separate product meters | On the same Neon plans | Database only |

Highly available production | A larger size, or a replica you pay for | Cloud SQL after the trial, from $9.37/mo | Scale, $0.222 per CU-hour | HA PS-5 in us-east-1 is $15/mo |

Ten people, one project | Still one org fee | Still no seat line | Unlimited members, usage only | Still the cluster price |

[Firebase](https://firebase.google.com/pricing) Spark needs no card, and with Identity Platform, monthly active users are no-cost through 50,000. Firestore's standard edition includes 1 GiB stored and 10 GiB of egress a month, then Google Cloud rates the Firebase page does not reprint as one number. Blaze keeps those quotas and adds a billing account. Cloud Functions include 2 million invocations, then $0.40 per million, and phone auth is per SMS, on a rate card separate from the plan table.

SQL Connect, the Postgres path, is a three-month trial and then starts as low as $9.37/month, and that figure is Cloud SQL, not a discount on Firestore. The Firebase review covers the product, and the directory page is [Firebase on Toolradar](https://toolradar.com/tools/firebase).

[Neon](https://neon.com/pricing) Launch and Scale have no monthly minimum, and Launch compute is $0.106 per CU-hour, which suits a database that is allowed to sleep. Neon describes a CU as about 4 GB of RAM, so 0.25 CU is about 1 GB, close to Micro. Left on for 730 hours, that is 182.5 CU-hours, which comes to $19.35 before storage. Paid storage is $0.35 per GB-month, public egress includes 500 GB per project, then $0.10 per GB, and an extra branch past the allowance is $1.50 per branch-month.

Suspended compute is not billed, and invoices under $0.50 are not collected. Free is permanent: 100 CU-hours, 0.5 GB, and 5 GB of egress per project, and hitting compute or egress suspends the database until the next period instead of charging the card. The Neon review is the product page, and the directory record is [Neon on Toolradar](https://toolradar.com/tools/neon).

PlanetScale's [plans doc](https://planetscale.com/docs/planetscale-plans) says the old free plan is gone, and single-node Postgres starts at $5/mo. The [Postgres pricing doc](https://planetscale.com/docs/postgres/pricing) lists that PS-5, in us-east-1, at 1/16 vCPU and 512 MB of RAM, and the highly available PS-5 there is three times the single-node sticker. Network-attached storage includes 10 GB, then $0.125 per GB-month in us-east-1, the same disk rate Supabase charges past 8 GB.

A non-HA PS-5 includes 10 GB of public egress, a production branch includes 100 GB, and further egress in us-east-1 is $0.06 per GB. Each Postgres development branch is billed as its own single-node cluster, and SSO on the Base plan is $199/month. Auth, files, and functions stay on another invoice, and the directory record is [PlanetScale on Toolradar](https://toolradar.com/tools/planetscale).

Project count, instance size, and whether the database may sleep are what move the invoice on all four.

## Fees that never appear on the plan card

The subscription is the line you approved, and compute is the line that arrives anyway. The spend cap skips compute, so a resize, a second project, or a replica changes the invoice while the cap still looks on. The plan fee is charged at the start of the cycle, usage including compute is charged at the end, and a quiet week does not shrink an instance that stayed running.

Disk size sits inside the cap, so a database past 8 GB stalls instead of silently adding gigabytes at the overage rate. Turning the cap off to finish a migration also unblocks egress, MAU, and storage, and those meters move as a set.

Social login sits inside the MAU quota, SAML past 50 users is the $0.015 rate, and phone MFA is the $75 project add-on, outside the cap. Image transformations include 100 origin images, and the per-thousand fee in the table is the rest. Taxes are separate from the plan, and the billing FAQ says sales tax, VAT, and GST apply where Supabase is registered, based on the billing address, with collection scheduled to finish by June 30, 2026. A tax ID is what flips an eligible business invoice to reverse charge.

Charges are in USD and can appear to come from Singapore, the payment entity, so a card that blocks foreign merchants fails renewal. An overdue invoice pauses projects and drops the org to Free until it is paid. Supabase does not refund unused credits, canceling mints credits for unused subscription time, usage you already consumed can still be invoiced, and top-ups cannot pay that outstanding bill.

Docker and the CLI skip the subscription, and Studio ships in that Docker setup, while the machine, disk, and bandwidth are then yours, which is a VPS hosting decision rather than a plan decision. A Next.js app on Vercel is a separate invoice and does not shrink this compute line.

Devshot data: the audience file updated September 20, 2026 counts 28,000 developer subscribers, and the open rate is 30%. They read a backend bill the way they read a cloud bill, so a second project plus a spend cap left on is the month-two surprise. Plan changes for this list land in the Devshot brief.

## Which plan is worth paying for

For one person, start on Free, accept the pause, and move to Pro when the project must stay awake or the database passes 500 MB. Keep Micro until connections or CPU say otherwise, and add a second project only when you can name the invoice line. A branch left up all month costs about a Micro and does not receive the credit, so delete it when the pull request merges.

For a team, Pro is the right subscription until a review names SOC 2, ISO 27001, or a read-only role, and Team does not raise the quotas. Turn the spend cap off only for the meter you meant to exceed, and check the upcoming invoice that week, because the cap is not a per-meter budget. If the database should sleep, price Neon before a larger Supabase size, and if auth lives elsewhere, price PlanetScale before a replica.

If the rest of the stack is still open, the serverless platforms roundup is the wider set, and the Firebase comparison above is the direct one.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We read Supabase's [pricing page](https://supabase.com/pricing), the [billing FAQ](https://supabase.com/docs/guides/platform/billing-faq), the [compute doc](https://supabase.com/docs/guides/platform/manage-your-usage/compute), the [cost control doc](https://supabase.com/docs/guides/platform/cost-control), and the [IPv4 doc](https://supabase.com/docs/guides/platform/manage-your-usage/ipv4) on September 24, 2026.

Firebase figures come from [firebase.google.com/pricing](https://firebase.google.com/pricing), Neon figures come from [neon.com/pricing](https://neon.com/pricing), and PlanetScale figures come from the [plans doc](https://planetscale.com/docs/planetscale-plans) and the [Postgres pricing doc](https://planetscale.com/docs/postgres/pricing), us-east-1 column. The 23-tool count is Toolradar's managed databases ranking that month, and how this site publishes is on the Devshot about page.

## FAQ

### How much does Supabase cost in 2026?

There is no per-seat price: Free is $0, with two active projects and a pause after a week idle, Pro is $25/month per organization, and Team is $599/month per organization. Enterprise publishes no list price, compute is extra per project, a credit covers one Micro instance, and quota overages bill on Team or when the Pro spend cap is off, and invoices are in USD.

### Does Supabase charge per seat?

No. Team members are unlimited on every plan, so ten engineers on one project pay the same subscription as one engineer. What scales is the number of active projects, the compute size of each, replicas, branches, and any quota you cross with the cap off.

### Is Supabase free?

Yes, with limits. Free includes 50,000 monthly active users, 500 MB of database, 5 GB of egress, and 1 GB of file storage, with no card. Projects pause after one week idle, only two may be active, and Owners and Admins share that allowance. Daily backups and email support are paid, and Docker and the CLI avoid the subscription while the server under them does not.

### What is included in Supabase Pro?

The org subscription includes a compute credit equal to one Micro project, 100,000 MAU, 8 GB of disk per project, 250 GB of egress, 250 GB of cached egress, 100 GB of file storage, 2 million edge invocations, email support, and 7-day backups. Unit rates apply when the spend cap is off. A second project, a larger instance, point-in-time recovery, a custom domain, and IPv4 sit outside that subscription.

### Does the Supabase spend cap stop a large bill?

It stops part of the bill: on Pro it is on by default and it blocks extra disk, egress, functions, MAU, realtime, and storage size. It does not cap compute, branches, read replicas, IPv4, point-in-time recovery, custom domains, extra IOPS, log drains, or phone MFA, and Team has no switch. A replica or a size change can still hit the invoice while the cap looks enabled.

### How much is Supabase for two projects?

Supabase's pricing page puts a Pro org with two Micro projects at $35/month after the credit, and three Micro projects are $45/month in the billing FAQ. A branch instead of a second project is $0.01344 per hour, and the credit does not apply, so a branch that stays up all month is close to another Micro. A free project and a paid project cannot live in the same org.

### Is Supabase cheaper than Firebase?

Both have a no-card plan, and Firebase will not pause you for a quiet week. For one always-on database around 1 GB of RAM, Pro is the org fee with Micro covered, while Blaze has no plan fee and then meters Firestore, functions, and bandwidth on their own. Neon at 0.25 CU for a full month is $19.35 before storage, which can undercut Pro when auth and files are not on this invoice. PlanetScale's single-node entry is the us-east-1 sticker in the table, with no auth product.

### Can I pay Supabase yearly?

There is no published annual plan, and the billing FAQ says annual subscriptions are not officially supported. The workaround they describe is a credit top-up so the card is not charged every month. Those credits pay future invoices, they do not pay one that is already outstanding, and unused credits are not refunded.

Cite this: Devshot, "Supabase Pricing in 2026: Every Plan, Real Costs and Hidden Fees", September 2026.

The daily Devshot brief is where plan changes for this stack show up before they show up on an invoice. Subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Vercel Pricing in 2026: Every Plan, Real Costs and Hidden Fees

URL: https://devshot.co/blog/vercel-pricing
Type: blog
Published: 2026-09-27
Updated: 2026-09-27
Summary: Vercel Pro is $20/mo in 2026: one deploying seat and a usage credit of the same amount. Extra deployers, CDN tiers, and function CPU move the invoice.

Guide

## Vercel Pricing in 2026: Every Plan, Real Costs and Hidden Fees

Vercel Pro is $20/mo in 2026: one deploying seat and a usage credit of the same amount. Extra deployers, CDN tiers, and function CPU move the invoice.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 27, 2026 · 16 min read

Key facts3 plans compared: Hobby, Pro, Enterprise

- Updated: September 27, 2026
- Prices as of: September 2026

- 3 plans compared: Hobby, Pro, Enterprise
- Hobby (best for: A personal project that is not commercial): Free
- Pro (best for: One production app, or a team): $20/mo
- Enterprise (best for: An SLA, SCIM, and SSO inside the contract): Custom quote

Vercel Pro is $20/mo, and that sticker is a platform fee, not the hosting bill. It includes one deploying seat and a monthly usage credit of the same amount that expires when the month ends, and extra teammates who deploy are billed per seat. Hobby stays free only for non-commercial projects, and Enterprise is a custom quote.

These prices were read on Vercel's pricing page, Pro plan doc, and CDN docs on September 24, 2026. Vercel is the hosted Frontend Cloud from the Next.js team, so a git push produces a preview URL per branch, and functions bill on active CPU rather than a server you leave running. The product walkthrough is the Vercel review, and this page is only the invoice.

Toolradar data: the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) evaluated 297 tools, updated September 2026. The directory record is [Vercel on Toolradar](https://toolradar.com/tools/vercel).

Plan | Price | Best for | What moves the invoice |

Hobby | Free | A personal project that is not commercial | Hard caps, and overage cannot be purchased |

Pro | $20/mo | One production app, or a team | Deploying seats, then usage past the credit |

Enterprise | Custom quote | An SLA, SCIM, and SSO inside the contract | Sales sets limits. There is no public seat price |

## Hobby is free until the project makes money

Hobby does not ask for a card, and it does not sell extra usage. The included month is 1 million edge requests, 100 GB of fast data transfer, 10 GB of fast origin transfer, 4 hours of active CPU, 360 GB-hours of provisioned memory, and 1 million function invocations. Pass a cap and that meter pauses, usually until 30 days have passed, while web analytics can resume after 7 days.

Commercial use is the real gate, not the traffic cap. Vercel's fair-use rules restrict Hobby to non-commercial, personal projects, so a site that takes payment, advertises a product, runs ads, or exists because someone was paid to build it needs Pro or Enterprise. Affiliate linking as the point of the site counts, and asking for donations does not. A portfolio and a weekend demo can stay on Hobby, while a client site, a SaaS, or a side project with a checkout cannot.

The account limits bite before the traffic caps do. One account can hold a single Hobby team, 200 projects, and 100 deploys a day, with one build at a time, a 45-minute build cap, and functions that stop at 300 seconds. Firewall rules and IP blocks stop at 3, and runtime logs last an hour.

## Pro is a fee, a credit, and then meters

Pro bills three things: the platform fee, extra deploying seats, and on-demand usage after the credit. The fee includes one Owner or Member who can deploy, plus a monthly usage credit of the same amount. The [Pro plan doc](https://vercel.com/docs/plans/pro-plan) says that credit applies to managed infrastructure from the first unit, then Vercel bills the rest on demand, so a quiet app can stay inside the credit and a busy one will not. Unused credit dies at the end of the month and does not roll, and you get a notice at 75% of the credit, then daily and weekly summaries once you are past it.

Faster builds, no build queue, and cold-start prevention are on the Pro grid, and none of them switch a meter off, while email support is on Pro and absent on Hobby. Viewer seats are free and unlimited, so a designer who only comments on previews should not get a deploy role, and a billing role is one seat, separate from the deploying seats you pay for.

Paid Pro teams can claim one first-year domain on an eligible TLD. The offer is closed during a Pro trial, it is one domain per team, and year two renews at the standard rate, which the plan page does not print. Listed prices are USD and exclude VAT and GST, so a taxable invoice is higher than the sticker.

New accounts get a default on-demand budget of $200, and you can change it. At 100% you can pause every project, fire a webhook, or send an SMS. If you leave the hard stop off, the site keeps serving and the invoice keeps growing, which is the difference between a surprise and an outage you chose.

## The pricing grid and the plan doc do not describe the same CDN allowance

Budget the CDN from the plan doc, not from the largest number on the public grid. The [pricing page](https://vercel.com/pricing) lists Pro edge requests as 10 million a month included, then starting at $2 per million, and fast data transfer as 1 TB included, then starting at $0.15 per GB. The Pro plan doc includes Flat Rate CDN's lowest tier at no extra charge: 1 million CDN requests and 1 TB of transfer, and those requests show up as edge requests on the usage chart. Fair-use copy says the guideline ranges are not Pro entitlements, so Pro is billed through the credit, then on-demand rates.

The next Flat Rate tier costs the same as the platform fee and covers 10 million requests and 50 TB. A one-day spike does not move you up mid-cycle, the step-up happens next cycle if the whole month exceeds the tier, and past the top tier you return to per-unit regional rates. Turning Flat Rate off waits until the cycle ends. Bulk file delivery and using the CDN as storage are out of scope, and Vercel can move that project or limit it.

Meter | Hobby included | Pro, after the credit |

CDN / edge requests | 1 million | 1 million on the included Flat Rate tier |

Fast data transfer | 100 GB | 1 TB on that tier, then starting at $0.15 per GB |

Fast origin transfer | 10 GB | Starting at $0.06 per GB |

Active CPU | 4 hours | Starting at $0.128 per hour |

Provisioned memory | 360 GB-hours | Starting at $0.0106 per GB-hour |

Function invocations | 1 million | $0.60 per million |

Function CPU, memory, and invocations have no included hour bucket on Pro, so they spend the credit and then bill at the table rates. Active CPU is time your code is running, which means a function waiting on a model costs less than an always-on box for the same wait. Caching skips the invocation and the CPU, and it does not skip the edge request.

A Next.js page with images, ISR, and middleware can hit three meters on one view. ISR reads start at $0.40 per million and writes at $4 per million, and Hobby includes 1 million reads and 200,000 writes. Image transformations start at $0.05 per 1,000, and image cache uses those same read and write rates. Middleware can bill fast origin transfer twice on one request when it runs on paths that did not need it.

## Ten deployers cost ten fees, and viewers cost nothing

The platform fee already contains the first deploying seat, and each extra Owner or Member is billed at that same monthly rate, so headcount moves the invoice before traffic does. Ten people who can deploy cost $200 a month before any usage past the credit. A contractor who pushes one branch is a paid seat, and a reviewer who only opens preview URLs is not.

Put deploy rights on the people who merge, and leave everyone else as a viewer. A viewer can comment, open analytics, and open deployments, and they cannot change production. A six-person startup that marks every engineer as a Member pays for six platforms when two people ship.

Downgrading removes every member except the original owner, and the account can hold only one Hobby team. If a Hobby team already exists, Vercel makes you delete one team or merge them. Stores and domains have to be moved by hand first, or the downgrade stops with the projects still attached.

Builds are a separate meter from the seat, and new Pro teams default to Elastic machines at $0.0035 per CPU minute, while Hobby's basic machines are included. On Pro, basic is $0.007 per minute, standard is $0.014, enhanced is $0.028, and turbo is $0.105, with duration rounded up to the next minute, so a monorepo on Turbo has chosen a different invoice. On-demand concurrent builds queue after 500 per team.

## Enterprise is the contract that removes the add-on menu

Enterprise publishes no list price, and the public differences that matter are included SAML and SCIM, audit logs, managed WAF rulesets, multi-region compute and failover, a 99.99% SLA, and custom invoicing. SSO-protected previews, secure compute, and running Vercel compute in your own AWS account are Enterprise-only. Pro can buy parts of that menu as add-ons, and it still does not include SCIM, so directory sync belongs on the quote.

SAML single sign-on is $300/mo on Pro and included on Enterprise. A HIPAA BAA is $350/mo on Pro, and Flags Explorer is $250/mo for unlimited overrides after 150 in a month. Preview Deployment Suffix and static IPs are each $100/mo, and static IPs also add private data transfer at the fast-data-transfer rate, so the IP is not a flat fee.

Password protection is a per-project monthly fee equal to the platform fee, included on Enterprise. Speed Insights Plus is $10 per project per month, plus $0.65 per 10,000 events. Every plan also includes 10,000 Speed Insights events per team per rolling 30 days, and that free bucket pauses instead of billing overage. Web Analytics Plus is $10/mo for the team, not per project.

Observability Plus is $1.20 per million events, with 30 days of retention. Without it, Pro logs last a day and Hobby logs last an hour.

Drains are $0.50 per GB and are not on Hobby. Vercel Agent is $0.25 per million tokens plus the provider's token cost, and it is not on Hobby. Pro web analytics has no included events in the docs, the pricing page rate is $3 per 100,000 events, and Hobby includes 50,000, so Pro does not get that free bucket.

## v0 is a second product with its own seats

v0, the prompt-to-app builder, is a paid add-on on the Vercel grid with its own [pricing page](https://v0.app/pricing), marked in USD. Free includes $5 of monthly credits and a limit of 7 messages a day. Plus, the plan new teams buy, is $30/user/mo and includes a monthly credit of the same amount per seat, plus $2 of free daily credits on login.

Business is $100/user/mo, with that same per-seat credit, and training opt-out is the default, while Plus does not include that opt-out, so a team that cannot allow training has to leave Plus. Enterprise is a custom quote, and it is where v0 SAML and role-based access sit.

Credits are not the hosting credit, and generations stop when the balance hits zero. Monthly credits roll to the next cycle and expire after 65 days, while purchased credits last a year and need an active paid plan to spend. On Plus and Business the monthly credits stay per person, and a shared pool, if you buy one, is used only after a person's own credits run out. Premium, the older individual plan, is being sunset and is closed to new users.

A Hobby team does not absorb a new v0 team. If you already have a Hobby team, v0's team docs say a new team at checkout also requires Vercel Pro, and that subscription does not add v0 credits, so Pro is a gate rather than a bundle. Plus seats are builders, so every member can create chats and deployments, and v0 viewer roles are an Enterprise control.

## What ten developers pay next to Netlify, Railway, Render, and Cloudflare

Every figure below was checked on that vendor's pricing page on September 24, 2026. This table is the invoice when the whole team needs to ship, and Vercel is the column that grows with deployers. The product differences sit in the reviews linked under it.

Buyer | Vercel | Netlify | Railway | Render | Cloudflare |

Solo, not commercial | Hobby | Free, 300 credits | Free, $1/mo of usage | Free workspace, free web service | Workers Free, 100,000 requests a day |

One production deployer | Pro | Personal, $9/mo, 1,000 credits | Hobby, $5/mo | Starter instance, $7/mo | Workers Paid, $5/mo |

10 people who deploy | $200 in deploying seats | Pro, $20/mo, unlimited members, 3,000 credits | Pro workspace, seats included | Pro workspace, $25/mo, seats included | Workers Paid, not priced per seat |

Netlify Pro is a workspace, not a seat: unlimited members, 3,000 credits, private organization repos, and at least three concurrent builds. Netlify's USD calculator prices bandwidth at $0.13 per GB and compute at $0.07 per GB-hour at the Pro pack rate. A production deploy costs 15 credits, covered in the Netlify review and [Netlify on Toolradar](https://toolradar.com/tools/netlify).

[Railway](https://railway.com/pricing) Hobby includes a usage credit equal to the subscription. Pro is a workspace fee the same size as Vercel's platform fee, seats included, with a matching usage credit. RAM is $10 per GB per month and egress is $0.05 per GB, per second, so a stopped service stops the compute line. CPU per vCPU matches that platform fee, also per second, and builds are free. The Railway review and [Railway on Toolradar](https://toolradar.com/tools/railway) cover the product.

[Render](https://render.com/pricing) bills a workspace plus an instance, and Pro includes unlimited seats and 25 GB of bandwidth, then the same per-GB rate Vercel charges past included transfer, and Hobby includes 5 GB. A Starter web service is 512 MB and 0.5 CPU, prorated to the second while it is provisioned, which means you pay while the box exists. Scale, where SAML and SCIM live, is $499/mo, and extra pipeline minutes are $5 per 1,000 minutes after 500 on Hobby and 1,000 on Pro. See the Render review and [Render on Toolradar](https://toolradar.com/tools/render).

Cloudflare prices this on [Workers](https://developers.cloudflare.com/workers/platform/pricing/), and that page states the paid minimum in USD. Static asset requests are free and unlimited, and Pages Functions bill as Workers. Paid includes 10 million requests, then $0.30 per extra million, with no egress charge. The record is [Cloudflare on Toolradar](https://toolradar.com/tools/cloudflare).

## Fees that never appear on the plan card

The platform fee is the cover of the invoice. On-demand starts once the credit is gone, and the default budget will not pause deploys unless someone turns that on. A removed viewer costs nothing, and a removed deployer keeps costing money until the seat is actually taken off the team. The plan page does not publish a mid-cycle credit for that removal, so dropping someone mid-month is not described as a refund.

A page that regenerates on every request, or an image component aimed at remote URLs with a short cache, spends ISR writes and image transformations while the seat line stays flat. BotID deep analysis is $1 per 1,000 checks, and firewall rate limiting starts at $0.50 per million allowed requests. An extra microfrontend project costs the same monthly fee as Flags Explorer.

A second Hobby account does not fix a commercial site, and circumventing the caps is a fair-use violation. A long-running worker or a database belongs on a different bill: Heroku alternatives, VPS hosting for developers, the DigitalOcean review, plus Supabase and Firebase, compared in Supabase vs Firebase.

Devshot data: this brief goes to 28,000 developers. The audience file dated September 20, 2026 shows a 30% open rate and a 1.8% click rate. Plan changes for that list run in the Devshot brief.

## Which plan is worth paying for

Stay on Hobby until a cap pauses a deploy or someone pays you to ship. For one production app, Pro is the Vercel answer, and it is the wrong comparison when the site is only static files, which Cloudflare does not bill. For a team, give deploy rights to the people who push, since ten deployers cost more than ten people on Netlify Pro or Railway Pro, where the workspace fee does not grow with headcount.

Turn the hard spend limit on before the first on-demand cycle, and price Enterprise for SCIM, an invoice, or the 99.99% SLA. If you are still choosing the host, start with serverless platforms.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We checked Vercel's [pricing page](https://vercel.com/pricing), the [Pro plan](https://vercel.com/docs/plans/pro-plan), [Flat Rate CDN](https://vercel.com/docs/pricing/flat-rate-cdn), [CDN usage](https://vercel.com/docs/manage-cdn-usage), the [Hobby plan](https://vercel.com/docs/plans/hobby), and [fair use](https://vercel.com/docs/limits/fair-use-guidelines) on September 24, 2026.

Netlify figures come from [netlify.com/pricing](https://www.netlify.com/pricing/). Railway figures come from [railway.com/pricing](https://railway.com/pricing) and the [plans doc](https://docs.railway.com/reference/pricing/plans). Render figures come from [render.com/pricing](https://render.com/pricing). Cloudflare figures come from the Workers pricing page linked above. v0 figures come from [v0.app/pricing](https://v0.app/pricing), the [v0 pricing doc](https://v0.app/docs/pricing), and the [v0 teams doc](https://api2.v0.dev/docs/teams). The 297 tool count is from Toolradar's hosting and deployment ranking that same month. Editorial rules are on the Devshot about page.

## FAQ

### How much does Vercel cost in 2026?

Hobby is free for non-commercial projects, with hard caps and no paid overage. Pro is $20/mo, including one deploying seat and a monthly usage credit of the same amount that expires at month end. Ten people who can deploy are charged as ten seats before any usage, and Enterprise is a custom quote. CDN, function CPU, and add-ons sit on top of the fee.

### Is Vercel free for commercial projects?

No, Hobby is free only for personal, non-commercial use. Taking payment, advertising a product, running ads, being paid to build the site, or making affiliate links the point of the site requires Pro or Enterprise, and donation requests do not. Passing a Hobby cap pauses the meter rather than opening a paid overage line.

### How much is Vercel for a team of 10?

If all ten people can deploy, the bill is $200 a month in seats, because the platform fee includes the first deployer and each extra deployer matches that rate. Viewer seats are free, so ten people are not ten fees unless all ten ship code. Usage past the included credit is extra, and a new team can pause projects when on-demand spend hits the default budget. Netlify Pro and Railway Pro charge a workspace fee rather than a fee per deployer.

### What is included in the Vercel Pro usage credit?

The credit equals the platform fee, applies from the first unit of managed infrastructure, and does not roll over, so a quiet month saves nothing for a busy one. After it, Active CPU starts at $0.128 per hour, provisioned memory at $0.0106 per GB-hour, and invocations at $0.60 per million. The plan doc's included CDN tier is 1 million requests and 1 TB, smaller than the 10 million edge requests on the public grid. Above that tier you pay on-demand rates.

### Is Vercel cheaper than Netlify?

At one deployer, Vercel's platform fee matches Netlify Pro, and Netlify Pro includes unlimited members plus 3,000 credits. At ten deployers, Vercel multiplies by headcount and Netlify stays on that one workspace fee. Netlify's published credit math prices bandwidth at $0.13 per GB. Vercel's fast data transfer, past the included terabyte, starts at $0.15 per GB and is regional. Pick Vercel when Next.js previews and Fluid compute are the product, and pick Netlify when headcount would otherwise multiply the fee.

### Does v0 come with a Vercel plan?

No, v0 has its own plans, and a hosting subscription does not include them. Free includes a monthly credit and 7 messages a day. Plus is $30/user/mo with a per-seat credit of the same amount, and Business is $100/user/mo with training opt-out by default, which Plus does not include. If you already have a Hobby team, creating a new v0 team also requires Vercel Pro, and that Pro subscription does not add v0 credits. Monthly v0 credits roll over and expire after 65 days.

### What do you get on Vercel Enterprise that Pro does not?

Enterprise includes SAML, SCIM, audit logs, managed WAF rulesets, multi-region failover, and a 99.99% SLA, at a custom quote. On Pro, SAML is $300/mo and a HIPAA BAA is $350/mo. SSO-protected previews and compute inside your own AWS account are Enterprise-only. Buy the quote when procurement needs those controls, not to get a cheaper seat.

Cite this: Devshot, "Vercel Pricing in 2026: Every Plan, Real Costs and Hidden Fees", September 2026.

Devshot is the daily brief on the tools and bills developers have to act on. Get it at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Supabase vs Firebase (2026): Which Is Better for Developers?

URL: https://devshot.co/blog/supabase-vs-firebase
Type: blog
Published: 2026-09-26
Updated: 2026-09-26
Summary: Supabase vs Firebase for developers in 2026: Pro at $25 a month versus Firestore on Blaze, and which fits web apps, offline mobile, and SQL.

Guide

## Supabase vs Firebase (2026): Which Is Better for Developers?

Supabase vs Firebase for developers in 2026: Pro at $25 a month versus Firestore on Blaze, and which fits web apps, offline mobile, and SQL.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 26, 2026 · 16 min read

Key facts2 products compared: Supabase, Firebase

- Updated: September 26, 2026
- Prices as of: September 2026

- 2 products compared: Supabase, Firebase
- Supabase (best for: Web apps that want SQL and a way to leave): $25/month on Pro
- Firebase (best for: Offline native apps, or a team already on Google Cloud): No base fee on Blaze

For a new web app in 2026, pick Supabase: Pro is $25 a month, and that fee covers one Micro Postgres instance. Pick Firebase for a native iOS or Android app that must read and write offline, or when you already run on Google Cloud. Spark is free, and it does not include Cloud Functions.

Prices below were verified on each vendor's pricing page in September 2026. A free Supabase project pauses after one week without traffic.

Toolradar data: Toolradar, the software directory we run, evaluated 109 low-code backend tools for its September 2026 ranking, and Supabase is in that set as the Postgres option. Firebase Firestore is tracked beside it among 23 managed databases. The listings are the [low-code backend ranking](https://toolradar.com/best/low-code-backend) and the [managed database ranking](https://toolradar.com/best/managed-databases).

## At a glance

Product | Best for | Starting price | Standout |

[Supabase](https://toolradar.com/tools/supabase) | Web apps that want SQL and a way to leave | $25/month on Pro | Postgres, row-level security, Docker self-host |

[Firebase](https://toolradar.com/tools/firebase) | Offline native apps, or a team already on Google Cloud | No base fee on Blaze | Firestore cache, Crashlytics, Cloud Messaging |

A backend-as-a-service, for a developer, is the hosted database, auth, and file storage you would otherwise stand up before the first feature. Supabase puts Postgres under that bundle, so a team that thinks in tables does not invent a second model. Firebase puts Cloud Firestore under a mobile suite, and SQL is an extra product you turn on with SQL Connect.

The product writeup on this site is the Supabase review. Invoice notes for that vendor sit on [Toolradar's Supabase pricing page](https://toolradar.com/blog/supabase-pricing-2026), and this page is the choice between the two.

## What the sticker price leaves out

Supabase Pro stops at the quota unless you opt into overage. Spend caps are on by default, and you turn the cap off before pay-as-you-grow billing starts. That suits a team that would rather stop than open the invoice. A Google Cloud budget can email you, and the [Firestore pricing page](https://cloud.google.com/firestore/pricing) states that requests still succeed after you pass that budget, so Firebase stays up and then bills you.

The second Supabase project is the invoice people miss, because the plan sits on the organization and each project has its own compute. Supabase's pricing page works an example of two Micro projects on Pro at $35 a month: the plan fee, plus a second instance the included credit does not cover. Budget dev and prod as that pair. Branching, at $0.01344 per branch per hour on Pro, is the other preview path, and a branch you leave open is a running cost.

Free Supabase projects are capped at two active ones, and they pause after a week idle. You can keep as many paused projects as you want, which suits experiments and fails a side project that must answer a webhook on a quiet Tuesday. Firebase publishes no inactivity pause on the Spark plan, so a quiet prototype stays up.

Cloud Functions require Blaze, and the functions docs tell you to upgrade before you deploy, because the pricing table marks invocations as unavailable on Spark. Spark still includes Firestore, non-phone Auth, Hosting, Crashlytics, and Cloud Messaging at no cost. That covers a client-only prototype and stops at a trusted webhook.

SQL Connect, the Postgres path, allows one no-cost Cloud SQL instance per project for three months. If the project is not on Blaze when that trial ends, the instance is archived and deleted after another 90 days, so you lose it without a card. After the trial, Cloud SQL starts as low as $9.37 a month.

## Pricing, plan by plan

Supabase bills the plan monthly, while Firebase Blaze bills usage and Spark needs no payment method. The grids are on the [Supabase pricing page](https://supabase.com/pricing) and the [Firebase pricing page](https://firebase.google.com/pricing).

Plan | What you pay (USD) | What a developer should read |

Supabase Free | $0 | 50,000 monthly active users, 500 MB database, 1 GB files, 5 GB egress, 2 active projects |

Supabase Pro | $25/month | 100,000 monthly active users, then $0.00325 each; 8 GB disk; 250 GB egress; spend cap on |

Supabase Team | $599/month | SOC 2, ISO 27001, 14-day backups, dashboard SSO, HIPAA as a paid add-on |

Supabase compute | Micro $10, Small $15, Medium $60, Large $110 a month | The included credit covers one Micro instance on Pro and Team |

Firebase Spark | No charge, no card | Daily Firestore quotas; Cloud Functions not included |

Firebase Blaze | Usage, Spark quotas included | $300 credit if you are eligible; functions start on this plan |

SQL Connect | 3-month trial, then from $9.37/month | 250,000 operations a month on Blaze, then $0.90 per million |

Both free tiers include 50,000 monthly active users before a paid identity rate. Pro raises Supabase's allowance to 100,000 and then charges per user. At 200,000 monthly active users the overage is $325, before the plan fee, because half of those users sit past the included allowance.

Firebase keeps that same free allowance at no cost on Blaze with Identity Platform, then publishes no list rate of its own past it. SAML on Pro includes 50 monthly active users and then $0.015 each. Firebase's SAML and OIDC allowance is also 50, then Google Cloud pricing.

Firestore in Iowa (us-central1) is cheap until a listener misbehaves. Past the free quota, reads are $0.03 per 100,000 documents and writes are $0.09 for that same batch. A million extra reads cost $0.30 and a million extra writes cost $0.90.

The free quota is daily, resetting around midnight Pacific: 50,000 reads, 20,000 writes, and 20,000 deletes, plus 1 GiB stored and 10 GiB of egress a month. A launch-day spike bills even when the rest of the month is quiet. Named databases do not get that free quota, so a second database pays from the first read.

You pay a read when a listened document is added, updated, or drops out of the result because it changed, so a wide listener costs more than one fetch. Security rules that call exists(), get(), or getAfter() add a read per dependent document, and a budget alert will not cut this off.

Supabase meters egress, disk, and users, not documents, so a join-heavy dashboard does not need a read calculator. A file-heavy app still has to model egress before it treats the plan fee as the bill. Pro includes 8 GB of disk and then $0.125 per extra gigabyte, and 250 GB of egress and then nine cents per gigabyte. Files are 1 GB on Free and 100 GB on Pro, then $0.0213 per extra gigabyte.

Supabase includes 500,000 edge invocations on Free and 2 million on Pro, then $2 per extra million. Blaze includes that same paid allowance, then $0.40 per million, plus 400,000 GB-seconds, 200,000 CPU-seconds, and 5 GB of outbound networking (then $0.12 per GB). Firebase is cheaper per extra invocation once you are on Blaze, and Supabase is the one that runs a function before you add a card.

Point-in-time recovery on Supabase is $100 a month per seven days of retention. Phone multi-factor auth is $75 a month for the first project and $10 a month for each project after that. Firebase phone sign-in is unavailable on Spark and billed per SMS on Blaze, on a separate rate card.

## The database you actually query

When the schema has relationships, Supabase wins, because organizations, members, and invoices are tables and foreign keys. Each project is Postgres you can open with full privileges, under the PostgreSQL license, with REST from PostgREST and GraphQL from pg_graphql. If the host becomes the problem, pg_dump is the documented exit, so pick it when you refuse a one-way door. Other hosts for that engine are in our Postgres hosting roundup.

Row-level security is the authorization model and the footgun, because policies run inside Postgres on every access. Supabase's docs state the failure: a table in an exposed schema without row-level security is readable and writable by any role that holds a grant. Adding a policy does not revoke the default grants on anon and authenticated, so the table stays public until you revoke them. The service_role key bypasses the policies and belongs on a server, never in a browser bundle.

When the data is a document the phone already holds, Firebase wins. Firestore is schemaless, so you model for the screen, duplicate fields so a query does not need a join, and accept that a security rule is application code. That fits a feed, a chat thread, or game state, and it fits poorly the week someone asks for a cohort query across accounts. A GUI on either engine is a separate choice, covered in database clients.

SQL Connect does not turn Firestore into Postgres, so a need for SQL means a second database. It is a separate Cloud SQL database, with a GraphQL schema in .gql files and generated SDKs for Kotlin, iOS, Flutter, and web. Clients call those named operations, and native SQL inside them is a preview with no SLA, which Firebase says can change in incompatible ways. If the GraphQL layer is what you want, compare it with GraphQL tools before the connector becomes your only API.

## Auth, rules, and who can see a row

Both products sign people in with email and social providers on the free tier, and both get expensive at enterprise identity. Supabase Auth is GoTrue, MIT licensed, and row-level security reads auth.uid() from the JWT. Basic multi-factor auth is on every Supabase plan, while leaked-password protection, a single session per user, and session timeouts start on Pro. Dashboard SSO for your teammates is a Team feature, separate from SAML for end users.

Firebase Authentication on Spark covers the non-phone methods, and Identity Platform is the upgrade that applies the included user cap on both plans. If the second factor is SMS, price the messages before you compare database fees, because that line can dwarf either plan. The login line alone, priced for Auth0, Clerk, Supabase Auth and Firebase at 10,000 and 100,000 users, is in Authentication pricing compared.

Supabase wins when you need to explain one query, because the policy is SQL. Firebase wins when the phone must fail closed while offline, because the rules travel with the SDK. Neither replaces a gateway in front of a public webhook, and that comparison is the API gateway guide.

## Functions, realtime, and the offline client

Edge Functions are TypeScript on a Deno-compatible runtime, available on Free within the invocation allowance. A Stripe webhook or a small model call fits, and a long render should move to a background worker. The Supabase CLI serves them locally, and that local stack is not the production self-host.

Cloud Functions run JavaScript, TypeScript, or Python. They can react to Firestore, Auth, and Storage when a document changes. Deploy is Blaze, then the CLI, then Cloud Build, and you pay invocations, CPU, memory, and outbound gigabytes separately.

On a side project, Supabase wins the webhook, because the function exists before you add a card. On a busy mobile app, Firebase wins the document-write trigger, because that trigger is native and the extra invocation is cheaper.

Supabase Realtime, Apache 2 licensed, can broadcast, track presence, or stream Postgres changes. Use broadcast for a cursor or a typing indicator, not a feed of every row change.

Free allows 200 peak connections and 2 million messages a month, with a 256 KB cap. Pro allows 500 connections and 5 million messages, then ten dollars per extra 1,000 connections and $2.50 per extra million messages, with a 3 MB cap. A socket per open tab will leave the connection allowance first.

Firebase Realtime Database, the older JSON tree, allows 100 simultaneous connections on Spark and 200,000 per database on Blaze, then $5 per GB stored past 1 GB. Prefer Firestore listeners, billed as reads, unless you specifically want that connection model.

For a phone that loses its signal, Firebase wins. The Firestore client caches documents the app is using and can read, write, and query that cache with no network.

Android and Apple enable this by default, while the web leaves it off and then only in Chrome, Safari, and Firefox, so a mobile site does not get that cache. Conflicts are last write wins, so skip the cache when two people edit the same record. A form filled in on a subway is a Firestore problem with a documented answer, and a Next.js dashboard that is useless offline is not.

Skip Firebase Hosting if the site already deploys from Vercel. Crashlytics, Cloud Messaging, Analytics, and Performance Monitoring stay no-cost on both plans, which is the mobile bundle people mean by Firebase.

## Leaving, and the compliance cliff

On the way out, Supabase wins. Docker is the documented self-host, and you give up managed branching, point-in-time recovery, managed backups, and the platform API. You patch the host and you take the backups, and support is community, not the email line that starts on Pro.

Storage in that stack is an S3-compatible API with metadata in Postgres. A VM you already know how to run, which the DigitalOcean review covers, is a realistic place for it, and the wider set is self-hosting tools.

Firebase does not offer that Compose file, so leaving means an export, a new auth system, and a rewrite of the rules. Firestore, Cloud Functions, and SQL Connect are Google Cloud services, and the price list is Spark or Blaze. The emulator is how you develop, and it is not a database you can carry to another cloud. Choose Firebase when the offline client, or the rest of Google Cloud, is worth that rewrite.

SOC 2 and ISO 27001 are on Supabase Team, at the $599 monthly fee, with 14-day backups and 28-day logs. Pro stops at seven-day backups and seven-day logs, so if an assessor wants the report, Pro is the wrong plan. Firebase publishes no SOC 2 plan tier, so read the Google Cloud contract rather than inferring one from the Spark column.

Until a log drain is on, dashboard retention is the whole trail. Pair it with log management, or with the Datadog review if that is already the on-call screen.

Devshot sends a short daily brief on the tools developers ship with. Subscribe here.

## Who should pick which

Pick Supabase Pro for a relational web product whose bill should stop at the quota unless someone turns the cap off. One Micro project stays at the plan fee, and two projects should be budgeted at the $35 example on the pricing page. Use Free only for prototypes you can tolerate paused.

Pick Supabase Team when a security review asks for SOC 2 or ISO 27001 on the vendor's plan list. Do not plan to upgrade in the middle of the questionnaire.

Pick Firebase Spark for a client-SDK prototype inside the daily quotas, with no server function. The moment you need a trusted webhook, a schedule, or phone auth, you are on Blaze.

Pick Blaze for a native app that must queue writes offline, or for a team that already pays Google Cloud and wants Crashlytics and Cloud Messaging in the same project. Set a budget, and also an application limit, because the budget will not reject the next read. Use SQL Connect only if you want Cloud SQL and you accept GraphQL as the client API, with the trial's deletion date on the calendar.

Pick self-hosted Supabase when the data cannot sit in a vendor region and someone owns backups, not as a way to dodge the plan fee. Other function hosts are in serverless platforms, and which model that function calls is a separate choice, covered in AI for developers.

## How we compared

We read Supabase's pricing, architecture, row-level security, edge function, self-hosting, and AI guides, plus Firebase's pricing page, Cloud Functions docs, Firestore offline docs, SQL Connect docs, and the Iowa rates on Google Cloud's Firestore pricing page, on 23 September 2026. The two-project total is Supabase's published example. The million-operation totals multiply those Iowa unit prices.

Louis Corneloup, founder of Toolradar and Dupple, edited the verdict. Re-check both pricing pages before you create the production project.

## FAQ

### Which is better for developers in 2026, Supabase or Firebase?

Supabase is the better default for a web app that needs Postgres, joins, and a plan you can cap. Pro is $25 a month and includes one Micro instance. Firebase is the better default for a native iOS or Android app that must use cached data offline, and for a team that wants Crashlytics and Cloud Messaging on the same Google Cloud project. SQL Connect adds Cloud SQL to Firebase, but it does not replace Firestore, and it does not make the two products the same database.

### How much does Supabase cost compared with Firebase?

Supabase Free is $0, pauses after a week idle, and allows two active projects. Team is $599 a month, and two Micro projects on Pro are $35 a month in the vendor's example. At 200,000 monthly active users, Pro adds $325 of overage before the plan fee, at $0.00325 each past the included 100,000.

Spark has no charge and no Cloud Functions, and Blaze has no base fee. In Iowa, a million Firestore reads past the daily quota cost $0.30, and a million writes cost $0.90.

### Does the Firebase free plan include Cloud Functions?

No. Invocations are marked unavailable on Spark, and the Cloud Functions docs say to move to Blaze before you deploy. Blaze includes 2 million invocations a month, then $0.40 per million, plus CPU, memory, and outbound allowances.

Supabase includes 500,000 edge invocations on Free and 2 million on Pro, then $2 per million. A webhook can ship on Supabase before you enter a card, and a Cloud Function cannot ship on Spark.

### Can I use SQL on Firebase the way I use it on Supabase?

Not against Firestore, which is a document database rather than SQL. SQL Connect is Cloud SQL for PostgreSQL, called through GraphQL operations, with SDKs for Kotlin, iOS, Flutter, and web. Native SQL in those operations is a preview without an SLA, and Firebase says it can change in incompatible ways.

The first instance is a three-month trial, one per project, then from $9.37 a month, and it is removed if you never move to Blaze: archived when the trial ends, deleted 90 days later. Supabase gives you SQL on the primary database from the first project, including pg_dump and row-level security.

### Which one works offline on a phone?

Firebase does this when the phone has no signal. Firestore caches the documents the app is using, allows reads and writes against that cache, and syncs later. Persistence is on by default for Android and Apple, off by default on the web, and limited on the web to Chrome, Safari, and Firefox.

A listener that reconnects after more than 30 minutes, with persistence enabled, is billed as a new query. Supabase keeps the login session on the device and streams database changes over Realtime while the client is connected. Firestore is the product with a documented offline query cache.

### Can I self-host Supabase or Firebase?

You can self-host Supabase with Docker: you run one project, you own patching and backups, and you do not get managed point-in-time recovery, managed backups, or branching. Support for that install is community, and the local CLI stack is for development only. Firebase's docs describe Spark and Blaze on Google Cloud, and the emulator is a development tool, not a database you can move to your own servers.

### Is there a free backend I can leave in production?

Supabase Free stays up only while something uses it, so after a week idle the project pauses, and only two projects can be active. Spark can keep serving Firestore, Hosting, and Auth without a card, inside the daily quotas, and it will not run Cloud Functions. Neither free plan includes a SOC 2 report from the vendor, and on Supabase that report starts on Team. If the app must answer traffic while you are asleep, and it must run a server function, budget for a paid plan before you call the prototype done.

Cite this: Devshot, "Supabase vs Firebase (2026): Which Is Better for Developers?", September 2026.

Devshot covers the tools and changes developers have to act on, in one daily brief. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Vercel vs Netlify (2026): Which Is Better for Developers?

URL: https://devshot.co/blog/vercel-vs-netlify
Type: blog
Published: 2026-09-26
Updated: 2026-09-26
Summary: Vercel Pro is $20/mo in 2026 with one deploying seat. Netlify Pro matches that fee, includes unlimited members, and meters traffic in credits.

Guide

## Vercel vs Netlify (2026): Which Is Better for Developers?

Vercel Pro is $20/mo in 2026 with one deploying seat. Netlify Pro matches that fee, includes unlimited members, and meters traffic in credits.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 26, 2026 · 13 min read

Key facts2 products compared: Vercel, Netlify

- Updated: September 26, 2026
- Prices as of: September 2026

- 2 products compared: Vercel, Netlify
- Vercel (best for: A commercial Next.js app with few people who deploy): $20/mo on Pro
- Netlify (best for: A team, a static site, or forms that should not meter): Personal $9/mo; Pro matches the fee above

Pick Vercel Pro at $20/mo for a Next.js app you will charge for in 2026. The fee covers one deployer and a matching usage credit, which fits a small team that ships.

Pick Netlify when several people deploy on one subscription, or when form posts should not touch the meter. Netlify Pro starts at that fee with 3,000 credits.

Prices below were read on 24 September 2026 from each vendor's pricing page, Vercel's Pro plan doc and fair-use guidelines, and Netlify's credit-plan docs. A Netlify account opened on or after 4 September 2025 bills in credits, so an older team may still sit on a legacy plan, and moving to credits cannot be undone.

Toolradar data: the [hosting and deployment ranking](https://toolradar.com/best/hosting-deployment) evaluated 297 tools, updated September 2026, and 230 of them offer a free or freemium plan. [Vercel](https://toolradar.com/tools/vercel) and Netlify are both in that set.

## At a glance

Product | Best for | Starting price | Standout |

[Vercel](https://toolradar.com/tools/vercel) | A commercial Next.js app with few people who deploy | $20/mo on Pro | Skew protection on Pro, Fluid compute, included Flat Rate CDN tier |

Netlify | A team, a static site, or forms that should not meter | Personal $9/mo; Pro matches the fee above | Unlimited members on Pro, preview deploys at zero credits |

A frontend host, for a developer, is the git-connected build, the CDN, and the function that runs the framework's server code, and both products are that host. The split is who may use the free tier, how a push is billed, and which Next.js behaviors you turn on yourself.

The product writeups on this site are the Vercel review and the Netlify review. The seat-and-meter invoice for Vercel alone is Vercel pricing, and this page is the choice between the two.

## What the sticker hides

Vercel Hobby is free, and it is not a commercial host, so a project that pays anyone who helped build it has to leave. Fair-use rules treat payment, product ads, display ads, a paid build, and affiliate links as the point of the site as commercial.

A client brochure fails that test even when it sells nothing, while donation requests stay allowed.

Netlify's credit-plan docs do not add that ban on Free, so a small paid site can start there. Free holds 300 credits, then stops, and you cannot buy more, which keeps the plan to a smaller experiment.

At a zero balance every project pauses, and visitors get a "Site not available" page until the next cycle or an upgrade.

Personal, at $9/mo, is the first plan that can buy credits, and it is one owner, so a second person who must push has already outgrown it.

The two Pro fees are not the same allowance. Vercel's Pro doc includes the lowest Flat Rate CDN tier at no extra charge: 1 million CDN requests and 1 TB of transfer a month, and that is the number to budget.

The public grid lists 10 million edge requests included, then $2 per million, and fast data transfer as 1 TB, then $0.15 per GB. Fair-use copy says those ranges are not Pro entitlements.

Netlify has no included terabyte, so bandwidth spends credits from the first byte. Bandwidth is 20 credits per GB, web requests are 2 credits per 10,000, and a successful production deploy is 15 credits, while previews, branch deploys, and failed builds cost nothing.

At the Pro pack rate the pricing page prints, that is $0.13 per GB, $0.01 per 10,000 requests, and $0.10 per production deploy.

## Pricing, plan by plan

Vercel bills the fee, then seats, then usage after the credit. Netlify bills a credit allotment, then packs if you enable them, so people move one invoice and traffic moves the other. Rates are on the [Vercel pricing page](https://vercel.com/pricing) and the [Netlify pricing page](https://www.netlify.com/pricing/).

Plan | What you pay (USD) | What a developer should read |

Vercel Hobby | Free | Non-commercial only. 1 million edge requests, 100 GB fast data transfer, 4 hours active CPU, 1 million function invocations. No paid overage |

Vercel Pro | $20/mo | One deploying seat and a usage credit of the same size. Credit expires at month end. Viewers are free |

Vercel Enterprise | Custom quote | SAML, SCIM, and a 99.99% uptime SLA in the contract |

Netlify Free | Free, 300 credits | Hard stop. Previews do not spend credits. Forms do not either |

Netlify Personal | $9/mo, 1,000 credits | One team owner. Auto recharge is 500 credits for $5, off unless you enable it |

Netlify Pro | Same fee at 3,000 credits; $33/mo at 5,000; $63/mo at 10,000; $95/mo at 15,000; $126/mo at 20,000 | Unlimited members. Rollover starts at the 5,000-credit tier. Pack is 1,500 credits for $10 |

Netlify Enterprise | Custom quote | 99.99% SLA, SSO, and SCIM on the pricing page |

Hobby pauses the meter you exceeded and will not sell the overage, so the personal project stalls. Netlify Free pauses the whole team, so one exhausted project takes the others offline.

Vercel Pro can keep serving: new teams default to an on-demand budget of $200, and at 100% you either pause every project or leave the invoice open.

Five people who can deploy on Vercel cost $100 a month before traffic, because the fee already contains the first seat and four more match it. A reviewer who only opens previews should be a viewer, and viewers are not billed.

Netlify Pro does not sell that deployer line, while Personal still has one owner and no developer role, so a second person who must push is already on Pro.

Vercel credit dies with the cycle, and Netlify monthly credits reset too, except on Pro at 5,000 credits or above, where leftovers last one extra month. Pack credits do not expire that way, and auto recharge stays off until an owner enables it for the whole team.

An extra concurrent build on Netlify is $40/mo, and Pro includes three at once, while Free and Personal include one, so two repos queue. Vercel bills the machine: new Pro teams default to Elastic at $0.0035 per CPU minute, and builds wait only after 500 concurrent per team.

## Next.js, native versus the adapter

For a Next.js app that should keep a live session on the deployment that started it, Vercel wins, because skew protection is included on Pro and absent on Hobby. Netlify supports the same control only after NETLIFY_NEXT_SKEW_PROTECTION is true, and releases before 14.1.4 also need the deployment-id flags.

A client fetch does not carry the id unless you send x-deployment-id, or you use a Next.js 15.4 experiment Netlify still marks as not production-ready.

Netlify's [Next.js overview](https://docs.netlify.com/build/frameworks/framework-setup-guides/nextjs/overview) maps the App Router, SSR, ISR, Server Actions, images, and middleware onto OpenNext from version 13.5. Edge-runtime pages still run in the functions region on Node.js, so that runtime does not put the page at the edge.

Rewrites cannot target public, and headers run after middleware, so a rule from the other host can miss the file.

ISR and images are separate Vercel meters, so images can bill after the CDN allowance is fine. Hobby includes 1 million ISR reads and 200,000 writes, while on Pro reads start at $0.40 per million and writes at $4 per million.

Transformations start at $0.05 per 1,000, and on Netlify next/image counts as bandwidth plus web requests, not as its own line.

Netlify wins a content site that needs a form, because submissions are free on every credit plan. The Next.js guide says that framework needs extra code, so a React form is not the HTML form the platform detects alone.

Vercel does not list a forms product on the pricing grid, so a contact form there is your route or another vendor.

Edge middleware on Vercel is capped at 50ms of average CPU, and network waits do not count, so a redirect is cheap and an API call in that file is not. Netlify edge functions are web requests at 2 credits per 10,000, and they skip GB-hours, which makes a rewrite simpler to price on Netlify.

Real work in that file belongs on Vercel's CPU line.

## Pushes, previews, and the build queue

On preview traffic, Netlify wins, because a pull-request deploy and a branch deploy spend nothing, and only the production publish costs 15 credits. A rollback to an older production deploy does not spend credits either.

Twenty production publishes are 300 credits, our multiplication of that rate, which on Free is the whole month if nothing else is metered, and on Pro leaves 2,700 credits.

Vercel lists unlimited deployments on Hobby and Pro, so the cost moves to the machine and the request rather than the publish count. Hobby includes basic machines, while Pro charges them, from Elastic through Turbo at $0.105 per minute, rounded up.

Webhook builds are Pro only, and a pipeline past the default branch belongs with CI/CD tools, not inside either host's queue.

## Bandwidth, requests, and a quiet month

For a commercial content site under a few dozen gigabytes, Netlify Personal is the smaller subscription, because Hobby is unavailable once the site is commercial. Personal's 1,000 credits are 50 GB if spent only on bandwidth, at 20 credits per GB, or about 66 production deploys if spent only on publishes, and a real month mixes both.

One million requests are 200 credits, and the calculator prices them at $0.01 per 10,000.

For a busy Next.js app inside the included tier, Vercel wins the transfer line. At Netlify's published $0.13 per GB, 1,000 GB is $130 before the subscription, while Vercel charges the platform fee and then on-demand only after the included transfer.

Fast origin transfer, between the CDN and functions, is separate: 10 GB on Hobby, then $0.06 per GB on Pro, and Netlify does not split that out of the 20-credit GB.

Compute is shaped differently, and a waiting function shows that. Netlify charges 10 credits per GB-hour, memory multiplied by time, printed as $0.07 per GB-hour at the pack rate, with no per-invocation line.

Vercel bills active CPU starting at $0.128 per hour, provisioned memory starting at $0.0106 per GB-hour, and invocations starting at $0.60 per million.

A database wait costs less CPU than wall-clock and still counts an invocation, and a cached response skips that CPU, not the edge request. HTML, assets, and edge functions on Netlify all count as web requests.

## Seats, locks, and the contract

For a team that all need a deploy button, Netlify wins. Pro includes unlimited collaborators, shared environment variables, private organization repos, password protection, and audit logs kept 7 days. Analytics cover 30 days on Pro, 24 hours on Personal, and a day on Free, and every plan stops at 500 projects.

Vercel sells the wider security menu and the tighter seat, and viewers stay free. SAML is $300/mo on Pro and included on Enterprise, and a HIPAA BAA is $350/mo on Pro.

Password protection is a per-project add-on equal to the platform fee, included on Enterprise, and SCIM is Enterprise only, so directory sync is a quote.

The [Enterprise SLA](https://vercel.com/legal/sla) commits to 99.99% availability, excluding excused downtime, with credits if you file within 30 days and attach logs. Netlify's pricing page lists that same uptime commitment, an enterprise network tier, and log drains on the custom plan.

Neither choice prints an Enterprise price, so both contracts are a quote.

Hobby and a Pro trial let Vercel train models on your content, and paid Pro leaves that off unless you opt in. Netlify's credit docs do not state a matching clause.

Runtime logs last an hour on Hobby and a day on Pro, so an incident that must outlive the dashboard needs monitoring tools or log management.

Devshot sends one daily note on what developers need to ship. Subscribe here.

## Who should pick which

Pick Vercel Hobby for a personal project that can survive a paused meter. Donations fit the fair-use note, a paid engagement does not, and a downgrade leaves only the original owner on the one Hobby team the account may hold.

Pick Vercel Pro for a Next.js product you sell, with one or two deployers and everyone else as a viewer, and buy Enterprise when you need SCIM or the uptime commitment.

Pick Netlify Free for a prototype inside the Free credit cap, and leave it before a real audience arrives. Pick Personal for a single owner on a public repo, inside that plan's credit allotment, with auto recharge left off.

Pick Netlify Pro when the repo is private, more than one person deploys, or you want password protection and 30-day analytics in the subscription. The $33 tier is the first with rollover, so a lumpy launch does not erase the quiet weeks around it.

Use the forms product for HTML, and budget extra code if the form is inside Next.js.

Pick neither for a long-running process, a database you size, or a VM you SSH into. Start with the Render review, the Railway review, and Heroku alternatives.

For a machine you administer, read VPS hosting for developers and the DigitalOcean review. The wider set is serverless platforms, and a backend next to the frontend is Supabase vs Firebase.

## How we compared

We read Vercel's pricing page, Pro plan doc, fair-use guidelines, Enterprise SLA, and terms, plus Netlify's pricing page, credit-plan reference, how-credits page, and Next.js overview, on 24 September 2026.

The 20-deploy, 50 GB, and five-deployer totals multiply those published rates, and the $130 bandwidth figure multiplies the pricing page's $0.13 per GB by 1,000.

Louis Corneloup, founder of Toolradar and Dupple, edited the verdict. Re-check both pricing pages, because a Netlify account from before 4 September 2025 may still be on legacy rates.

## FAQ

### Which is better for developers in 2026, Vercel or Netlify?

Vercel is the better default for a commercial Next.js app, because Pro is $20/mo, with one deploying seat, a matching usage credit, and skew protection included. Netlify is the better default when deployers should not be a per-seat line, when previews should be free, or when an HTML form should cost zero credits.

Its Pro plan starts at the same monthly fee with 3,000 credits and unlimited members.

### How much does Vercel cost compared with Netlify for a team that ships?

Hobby is free for non-commercial use and cannot buy overage, while Pro adds seats after the credit and five deployers are $100 a month before usage. Enterprise remains a custom quote rather than a price on the grid.

Netlify Free includes 300 credits and then pauses, while Personal is $9/mo for 1,000 credits. Pro steps from the base allotment to 5,000 credits at $33/mo, 10,000 at $63/mo, 15,000 at $95/mo, and 20,000 at $126/mo. Extra bandwidth at the pack rate is $0.13 per GB.

### Is either host free for a commercial site you intend to keep online?

Vercel Hobby is not free for a commercial site, because a site a paid developer built requires Pro or Enterprise, while donation requests sit outside that definition. Netlify's credit docs do not restrict Free to non-commercial work, but they stop Free at a hard credit cap with no pack.

A zero balance replaces the site until the next cycle or an upgrade, and Personal is the cheapest Netlify plan that can buy credits.

### Does Netlify run Next.js as completely as Vercel for a production app?

Netlify documents App Router, SSR, ISR, Server Actions, images, and middleware through OpenNext, from Next.js 13.5 up. Edge-runtime pages still run in the functions region on Node.js, so that choice does not place the page at the edge.

Skew protection is an environment variable, not included the way it is on Vercel Pro. Rewrites cannot point at public files, and headers run after middleware, while image transformations on Vercel start at $0.05 per 1,000.

### What happens on each host when included usage runs out?

A zero Netlify balance pauses every project, and visitors see a "Site not available" page. Free cannot buy credits, while Personal and Pro can, via auto recharge or a pack: $5 for 500 credits, or $10 for 1,500.

Hobby pauses the meter you passed and will not sell overage. Vercel Pro's default on-demand budget is $200, and you choose whether that ceiling pauses projects or extends the invoice.

### Which host is cheaper when everyone on the team deploys?

Netlify Pro costs less once more than one person must deploy, because members sit inside the subscription rather than on a per-seat line. Vercel bills each extra Owner or Member like the platform fee, so five deployers are $100 a month before traffic, and viewers stay free.

With a single deployer on Next.js, compare the included transfer tier instead of the seat line.

Cite this: Devshot, "Vercel vs Netlify (2026): Which Is Better for Developers?", September 2026.

Devshot is the daily brief on tools and changes developers act on. Join at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# AI-Assisted Coding in 2026: The Workflow That Actually Works

URL: https://devshot.co/blog/ai-assisted-coding
Type: blog
Published: 2026-07-21
Updated: 2026-09-25
Summary: The disciplined AI-assisted coding workflow for 2026: scope, context, small chunks, review every line, test, then commit. Real tools, honest limits, no slop.

Guide

## AI-Assisted Coding in 2026: The Workflow That Actually Works

The disciplined AI-assisted coding workflow for 2026: scope, context, small chunks, review every line, test, then commit. Real tools, honest limits, no slop.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 13 min read

Key facts6 tools compared: GitHub Copilot, Cursor, Claude Code, OpenAI Codex, Windsurf (now Devin…

- Updated: September 25, 2026
- Prices as of: September 2026

- 6 tools compared: GitHub Copilot, Cursor, Claude Code, OpenAI Codex, Windsurf (now Devin Desktop), Continue
- GitHub Copilot (best for: In-editor completions and chat across most IDEs): Free tier; Pro $10/mo; Pro+ $39/mo; Business $19 and Enterprise $39 per seat/mo
- Cursor (best for: AI-first editor with agent mode across a whole repo): Free (Hobby); Pro $20/mo ($16 billed yearly); Teams $40/user/mo ($32 yearly); Enterprise custom
- Claude Code (best for: Terminal agent for multi-file changes and test loops): Bundled with Claude Pro (about $17 to $20/mo) and Max (from $100/mo)
- OpenAI Codex (best for: Agentic CLI plus IDE extension tied to ChatGPT): Included in ChatGPT Plus ($20/mo) and higher tiers

You have reviewed the pull request. Four hundred lines, generated in twenty minutes, and nobody on the team can explain why the retry loop swallows the exception it was supposed to surface. The tests pass because the AI also wrote the tests, and they assert the wrong behavior. This is the failure mode nobody warns you about when they sell you on AI-assisted coding: it does not usually produce code that is obviously broken. It produces code that looks finished, compiles, and quietly encodes a decision no human made.

The tools are genuinely good now. The problem is that most developers use them like a vending machine. Type a request, take whatever falls out, paste it in. That is how you ship slop. AI-assisted coding works when you treat the model as a fast junior who never gets tired and never says no, which means it also never pushes back on a bad idea. The discipline is yours to supply. This is the workflow that keeps the speed and drops the slop.

## Why AI-assisted coding works (and where it goes wrong)

The reason AI pair programming helps at all is narrow and specific: large models are excellent at pattern completion. Boilerplate, glue code, a test that mirrors five other tests, a regex, a SQL query against a schema you paste in. These are places where the "right" answer is mostly a matter of following an established shape, and the model has seen the shape a million times. Here it is faster than you and rarely wrong.

It goes wrong the moment the task stops being pattern completion and starts being judgment. Should this function retry or fail loudly? Is this the right abstraction or a premature one? Does this change break an assumption six files away? The model has no idea, because it cannot see your system, cannot run your code, and has no stake in whether the thing works next week. It will still answer with total confidence. That mismatch, high confidence and zero context, is the entire risk surface.

Task type | How reliable the model is | What you still own |

Boilerplate, config, scaffolding | High | A quick read for wrong defaults |

Regex, SQL against a pasted schema | High | Rejection cases and the query plan |

Tests for code you specify | Medium | Whether the assertions are correct |

Refactors with fixed behavior | Medium | Diffing that behavior really held |

Business logic and API design | Low | The design decision itself |

Auth, payments, migrations, deletes | Low | Line-by-line review, always |

The workflow below exists to keep you in the high-reliability lane as much as possible, and to force a hard stop before anything from the low-reliability lane reaches main.

## Step 1: Scope the task before you type a prompt

The single biggest predictor of good output is a small, well-defined task. "Build the payments integration" gets you a wall of plausible code you cannot review. "Write a function that maps a [Stripe](https://docs.stripe.com/payments/payment-intents) PaymentIntent webhook to our internal Order status, handling only the succeeded and payment_failed events" gets you something you can check in ninety seconds.

Before you open the chat or the agent, decide what one unit of work you want. If you cannot describe it in a sentence or two, it is too big. Break it down first, on paper or in a comment block, then feed the model one piece at a time. Scoping is not overhead. It is the part of the job the model cannot do for you, and it is where senior engineers pull ahead of the vending-machine crowd.

## Step 2: Give the model the context it cannot see

The model has never seen your repository. It does not know your types, your naming conventions, the helper that already does the thing you are about to ask for, or the version of the library you actually run. Every one of those gaps is a place it will invent something. Your job is to close the gaps before you ask.

Paste the real function signatures, the schema, the relevant existing code, and the exact library version. If you are using an agent tool that indexes the repo, point it at the specific files rather than trusting it to find them. A prompt that includes the surrounding code and one concrete input/output example is the difference between a usable answer and a confident guess. When you are working in the chat interface rather than an agent, our ChatGPT for coding guide has copy-ready prompts that build this context in.

## Step 3: Generate in small chunks

Ask for one function, one component, one migration at a time. There are two reasons, and both are about your ability to verify.

First, a small generation is reviewable. You can hold a forty-line function in your head and spot the retry loop that swallows errors. You cannot do that with four hundred lines. Second, small chunks keep the model's context clean. When you dump an entire feature request on it, it loses track of constraints you stated earlier and starts contradicting itself. Tight scope in, tight output out.

If you are using an agent that edits multiple files in a loop, this still applies. Give it one coherent task, let it finish, review the diff, then give it the next. Chaining ten tasks into one prompt and walking away is how you come back to a diff nobody can safely merge.

## Step 4: Review every line as if a stranger wrote it

This is the step people skip, and skipping it is the whole problem. AI-generated code that you did not read is not yours, and you cannot vouch for it in review. Read it like you would a pull request from a contractor you have never met: what does each line do, why is it there, and what happens on the unhappy path?

Watch specifically for the failures the model is prone to. Error handling that catches and discards. Off-by-one boundaries. An abstraction that is fancier than the problem needs. A call to a helper that does not exist in your codebase. Hardcoded values that should be config. If you find yourself thinking "that looks about right," stop and actually trace it. "About right" is exactly where the expensive bug hides. When the change is a review of someone else's code rather than your own, a second model can help as a first pass, which we cover in best AI for code review.

## Step 5: Test it, and check the tests

Run the code. Not "read the code and feel good about it," run it. The model cannot execute anything in your environment, so until you do, correctness is a hypothesis. If it wrote tests, read those tests before you trust the green checkmark. AI-written tests have a habit of asserting whatever the code currently does, which turns a passing suite into proof that the bug is stable.

Write or ask for the cases that actually break things: empty input, the boundary value, the malformed payload, the concurrent call. If a test only exercises the happy path, it is decoration. For anything that fails at runtime rather than compile time, a model is a decent debugging partner once you feed it the full trace and the surrounding code, which best AI for debugging walks through in detail.

## Step 6: Commit small, with a message you actually wrote

Commit the reviewed, tested chunk on its own. Small commits are not a style preference here, they are a safety mechanism. When something breaks next week, a [bisect](https://git-scm.com/docs/git-bisect) through small commits finds the AI-generated change in minutes. A single mega-commit labeled "add feature" hides it forever.

Write the commit message yourself, or at least rewrite whatever the tool suggests, because the message is where you record the decision the model could not make. Why does this retry three times and not five? Why fail loudly instead of degrading? That reasoning is the thing your future teammates need, and it is the one artifact AI-assisted coding cannot produce, because it never existed in the model, only in your head.

Devshot sends one email a day on the AI tools and workflows developers actually use in production. Read it here.

## Tools to use

These are the real tools engineers reach for in 2026. Two ownership changes worth knowing before you pick: [Windsurf](https://windsurf.com) was renamed [Devin Desktop](https://devin.ai) under Cognition, and the [Continue](https://continue.dev) team joined [Cursor](https://cursor.com), though the open-source Continue project is still available. Prices move, so confirm the current number on each vendor's pricing page before you buy.

Tool | Best for | Pricing model |

GitHub Copilot | In-editor completions and chat across most IDEs | Free tier; Pro $10/mo; Pro+ $39/mo; Business $19 and Enterprise $39 per seat/mo |

Cursor | AI-first editor with agent mode across a whole repo | Free (Hobby); Pro $20/mo ($16 billed yearly); Teams $40/user/mo ($32 yearly); Enterprise custom |

Claude Code | Terminal agent for multi-file changes and test loops | Bundled with Claude Pro (about $17 to $20/mo) and Max (from $100/mo) |

OpenAI Codex | Agentic CLI plus IDE extension tied to ChatGPT | Included in ChatGPT Plus ($20/mo) and higher tiers |

Windsurf (now Devin Desktop) | IDE plus a command center for many agents | Free tier; Pro $20/mo; Max $200/mo; Teams from $80/mo |

Continue | Open-source, bring-your-own-model extension for VS Code and JetBrains | Free and open source; run any hosted or local model |

The split that matters is completions versus agents. [Copilot](https://github.com/features/copilot) and Continue live mostly in the editor, suggesting as you type. [Claude Code](https://claude.com/product/claude-code), [Codex](https://openai.com/codex/), Cursor's agent mode, and Devin Desktop are agents: they index the repo, edit multiple files, and run build and test commands in a loop. Conversation-sized work does not need an agent. Multi-file, in-repo changes are where agents earn their keep. Most engineers run a completion tool for daily typing and reach for an agent on the bigger jobs. For a deeper side-by-side, see our best AI coding assistants breakdown.

## Pitfalls

Over-trust. The costliest habit is accepting code because it looks right. Looks-right code that ships a subtle logic error is more expensive than code that fails to compile, because the compiler catches the second one and only production catches the first. The fix is boring and non-negotiable: read every line you did not write.

Hallucinated APIs. Ask a model to write against a library from memory and it will occasionally invent a method that does not exist or swap two argument positions. This is worse for smaller, faster-moving packages the model saw little of during training. Paste the current docs or the real README whenever the API surface matters, and never assume a method exists because the code that calls it runs in your head.

Security defaults. Ask for something generic like "a login handler" and you tend to get the simple version, not the hardened one. String-built SQL instead of [parameterized queries](https://cheatsheetseries.owasp.org/cheatsheets/SQL_Injection_Prevention_Cheat_Sheet.html). A missing authorization check. A secret logged in plaintext. Models optimize for a working example, not a safe one. Ask explicitly for parameterized queries, input validation, and secure defaults, and treat anything touching auth, payments, or user data as needing a real review, not a glance.

Context rot. In a long agent session or a sprawling chat, the model gradually loses the constraints you set early on. It forgets the naming convention, reintroduces a pattern you told it to avoid, or contradicts a decision from twenty messages back. When output quality drifts, do not argue with it across ten more turns. Start a fresh session, restate the constraints cleanly, and paste back only the code that matters. A clean context beats a long one.

## FAQ

### What is AI-assisted coding?

It is writing software with a large language model as an active partner: generating functions, tests, and boilerplate, explaining unfamiliar code, and editing across files, while a human scopes the work, reviews the output, and owns what ships. It sits between writing everything by hand and full autonomy. You stay the engineer of record; the model is a fast assistant that cannot see your system or run your code.

### How is AI pair programming different from a human pair?

A human pair pushes back, remembers last week's design decision, and shares accountability for what merges. A model does none of that. It never disagrees with a bad plan, has no memory of your codebase beyond what you paste in, and carries zero responsibility for the outcome. It is faster at producing code and useless at judgment. Use it for volume and speed, not for a second opinion on whether the approach is sound.

### Will AI-assisted coding make me a worse engineer?

Only if you stop reading the code. Developers who accept output blind lose the ability to reason about their own systems. Developers who scope tightly, review every line, and understand what they merge get faster without going soft. The skill shifts from typing syntax to specifying problems and verifying solutions, which is the more valuable half of the job anyway.

### How do I stop AI from hallucinating APIs and libraries?

Give it the real interface instead of asking it to recall one. Paste the current documentation, the actual function signatures, and the exact library version. If you use an agent that indexes your repo, point it at the specific files. Hallucination is mostly a symptom of missing context, so the fix is to supply the context the model cannot see rather than to prompt more cleverly.

### Which AI coding tool should I start with?

If you live in VS Code and want low friction, start with GitHub Copilot's free tier or the open-source Continue with a model you already pay for. If you want an agent that changes many files at once, Claude Code, Cursor's agent mode, or OpenAI Codex are the common picks. Try one completion tool and one agent, see which fits your work, and confirm current pricing before you commit.

### Is it safe to paste proprietary code into these tools?

Check each tool's data policy first. Some consumer tiers may use your inputs to improve models unless you opt out; business and enterprise tiers generally exclude your data by default. Redact secrets and customer data regardless of plan, and clear anything under an NDA with your company's policy before pasting. The convenience is not worth leaking a credential or breaching a contract.

### Can an AI agent handle a whole feature end to end?

Sometimes, for well-defined and pattern-heavy work, but you still review the result. Even when an agent edits a dozen files and gets the tests green, it made design choices you did not see it make, and it may have written the tests to match its own bugs. Let it do the mechanical work, then read the diff as if a new hire wrote it. The more the code touches auth, payments, or data, the less you delegate.

### Do I still need to write tests if the AI writes them?

You need to read them, which is most of the work. AI-written tests tend to assert whatever the code currently does, so a passing suite can just mean the bug is consistent. Check that the assertions describe the behavior you actually want, and add the cases the model skips: empty input, boundaries, malformed data, and concurrency. A test you did not verify is not coverage, it is a green light you have not earned.

Some offers on this page may be paid placements or contain affiliate links.

---

# 9 Best AI Tools for Python Coding in 2026 (Tested)

URL: https://devshot.co/blog/best-ai-for-python-coding
Type: blog
Published: 2026-07-21
Updated: 2026-09-25
Summary: The 9 best AI tools for Python coding in 2026, from Cursor and Copilot to Claude Code and Codex, for notebooks, type hints, and real projects.

Guide

## 9 Best AI Tools for Python Coding in 2026 (Tested)

The 9 best AI tools for Python coding in 2026, from Cursor and Copilot to Claude Code and Codex, for notebooks, type hints, and real projects.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 13 min read

Key facts9 tools compared: Cursor, GitHub Copilot, Claude Code, OpenAI Codex, JetBrains AI (PyCharm)…

- Updated: September 25, 2026
- Prices as of: September 2026

- 9 tools compared: Cursor, GitHub Copilot, Claude Code, OpenAI Codex, JetBrains AI (PyCharm), Devin Desktop (formerly Windsurf), Tabnine, Continue, Google Colab AI
- Cursor (best for: Whole-project edits plus fast autocomplete): Free; Pro $20/mo
- GitHub Copilot (best for: Everyday completion in VS Code or PyCharm): Free; Pro $10/mo
- Claude Code (best for: Terminal-driven multi-file refactors): Claude Pro $20/mo ($17/mo annual)
- OpenAI Codex (best for: Delegating tasks to a cloud agent): ChatGPT Plus $20/mo

Python punishes AI assistants in ways other languages do not. The type system is optional, so a model that guesses at a function signature can be confidently wrong for pages at a time. Half your real work lives in [Jupyter notebooks](https://jupyter.org) where state is out of order and a variable defined in cell 12 shapes the output of cell 3. Dependency and virtual-environment problems are their own category of pain: an assistant that suggests import polars means nothing if polars is not in the venv the interpreter is actually pointing at. And a lot of Python is data and ML code, where "it runs" and "it is correct" are very different claims.

So the useful question is not "which AI writes Python," because they all do. It is which tools understand Python's specific failure modes: notebook context, type hints, import resolution, and the gap between a script that executes and one that produces the right numbers. We compared the current field against those criteria, using vendor documentation, changelogs and what developers report in public reviews. Below are nine tools worth your time, what each is actually good at for Python, real prices, and where each one falls short. For the wider picture across languages, our AI for developers hub covers the general tooling, and best AI coding assistants ranks them beyond Python alone.

## Quick comparison

Tool | Best for Python | Starting price |

Cursor | Whole-project edits plus fast autocomplete | Free; Pro $20/mo |

GitHub Copilot | Everyday completion in VS Code or PyCharm | Free; Pro $10/mo |

Claude Code | Terminal-driven multi-file refactors | Claude Pro $20/mo ($17/mo annual) |

OpenAI Codex | Delegating tasks to a cloud agent | ChatGPT Plus $20/mo |

JetBrains AI (PyCharm) | Type-aware help inside PyCharm | Free tier; paid AI from ~$10/mo (check current) |

Devin Desktop (formerly Windsurf) | Agent-first editor, Cursor alternative | Free tier; Pro $20/mo |

Tabnine | Private, on-prem or air-gapped teams | $39/user/mo |

Continue | Open-source, bring-your-own-model | Free (open source) |

Google Colab AI | Notebooks and data-science EDA | Free tier; Pro ~$9.99/mo (check current) |

The five stages below are where these tools compete. Almost no tool wins all five for Python, which is why most working developers end up running two.

## 1. Cursor

[Cursor](https://cursor.com) is a fork of VS Code with AI built into the core rather than bolted on. For Python that matters because Cursor indexes your whole project, so when you ask it to add a field to a Pydantic model it can trace the change through the schema, the routes, and the tests in one pass. Tab completion is fast and predicts multi-line edits, and the agent can run your test suite, read the failures, and patch until pytest is green.

Its Python weak spot is notebooks. Cursor opens .ipynb files and can edit cells, but its notebook experience is thinner than a dedicated data-science IDE, and it can lose track of out-of-order execution state.

Pricing: a free Hobby tier with limited agent requests, Pro at $20/month, and Teams at $40/user/month. Best for developers who want one editor to write, refactor, and ship whole Python services. For the head-to-head against Copilot, see Cursor vs GitHub Copilot.

## 2. GitHub Copilot

[Copilot](https://github.com/features/copilot) is the most widely used AI assistant and the safest default for Python. It runs inside VS Code, PyCharm, Neovim, and Visual Studio, so you keep your existing setup. Its inline completions are strong on idiomatic Python (comprehensions, context managers, dataclass boilerplate), and its VS Code notebook support is genuinely good, which is not true of every tool here. Copilot also has an agent mode for multi-file edits and a CLI.

The weakness is depth on large refactors. Copilot is excellent at the next 10 lines and merely fine at reasoning across 20 files, where Cursor and Claude Code pull ahead.

Pricing: a free tier with a monthly completion cap, Pro at $10/month with unlimited completions and a premium request allowance, Pro+ at $39/month for more premium model access, plus Business and Enterprise. Best for developers who want reliable Python autocomplete inside the editor they already use.

## 3. Claude Code

[Claude Code](https://claude.com/product/claude-code) is a terminal agent, not an editor plugin, and that framing suits certain Python jobs well. You point it at a repo and describe the change. It reads files, edits across the project, runs commands, and iterates. For a task like "migrate this Flask app to FastAPI and update the tests," a capable agent working across the tree beats line-by-line completion. It handles imports, type hints, and test runs as part of the loop rather than as an afterthought.

The cost is that there is no inline tab completion. If you want an autocomplete-first flow, this is not that tool. It also burns tokens on big tasks, so watch usage.

Pricing: included in Claude Pro at $20/month ($17/month billed annually) and Claude Max from $100/month, with higher usage limits at Max. API pay-as-you-go is also available. Best for refactors, migrations, and multi-file work you would rather delegate than type.

Devshot sends one practical AI-for-developers workflow to your inbox every morning, free to join.

## 4. OpenAI Codex (ChatGPT)

[Codex](https://openai.com/codex/) is OpenAI's coding agent, reachable through the web, a CLI, and an IDE extension, and it runs on the same ChatGPT subscription you may already pay for. Its distinctive move is asynchronous cloud tasks: you hand off a Python job, it works in a sandbox, and you review a diff later. That fits well-scoped tickets like "add pagination to these three endpoints and write the tests." It is competent with common Python stacks and reads existing project conventions before writing.

The weakness is that the cloud-agent flow adds latency and review overhead compared with an inline assistant, and it is less suited to tight edit-and-run loops.

Pricing: ChatGPT Free has basic access, Plus is $20/month, Pro starts at $100/month with higher limits, and Business is $25/user/month ($20 billed annually). Best for developers who live in ChatGPT and want to delegate discrete Python tasks. Our guide to ChatGPT for coding goes deeper on prompt patterns.

## 5. JetBrains AI (PyCharm)

If PyCharm is already your Python IDE, [JetBrains AI](https://www.jetbrains.com/ai/) is the assistant with the least friction and the most language awareness. It builds on PyCharm's static analysis, so its suggestions respect your type hints, resolve imports against the correct interpreter, and understand your project structure instead of guessing from text. It handles completion, chat, test generation, and inline edits, and PyCharm's notebook and scientific-mode support makes it stronger than most on data work.

The trade-off is lock-in: the deepest features assume you stay inside JetBrains IDEs.

Pricing: PyCharm now ships with a free AI tier, and paid AI plans start around $10/month for AI Pro with roughly $20 and up for AI Ultimate. JetBrains changes these tiers often, so check current pricing before you commit. Best for Python developers who are committed to PyCharm and want type-aware help.

How each tool's Python strengths compare. Capabilities shift with every release.

## 6. Devin Desktop (formerly Windsurf)

[Devin Desktop](https://devin.ai/desktop), the editor sold as Windsurf until Cognition [renamed it on June 2, 2026](https://devin.ai/blog/windsurf-is-now-devin-desktop), is an agent-first editor and the closest rival to Cursor. Its local agent (Cascade, now replaced by Devin Local) keeps a running model of your project and edits across files while explaining its steps, which works well on Python services with clear module boundaries. Completions are solid and the agent is comfortable running commands and reacting to test output.

Two honesty notes. Cognition, the team behind the [Devin](https://devin.ai) cloud agent, has changed the product's name, agent and plans within a year, so older Windsurf reviews and pricing no longer apply. And like Cursor, its notebook support trails a dedicated data IDE.

Pricing: a free tier with unlimited Tab completions, then Pro at the same monthly price as Cursor Pro and $200/month for Max, with Teams at $80/month plus $40 per full seat ([Devin pricing](https://devin.ai/pricing), checked September 2026). The same plans include Devin's cloud agents if you want a fully asynchronous agent instead. Best for developers who want a Cursor alternative built around an autonomous agent.

## 7. Tabnine

[Tabnine](https://www.tabnine.com) is the pick for teams that cannot send proprietary Python to a third-party cloud. It supports self-hosted, VPC, on-premises, and air-gapped deployment, promises zero code retention, and offers license-safe models trained on permissively licensed code. That combination matters in finance, healthcare, and defense, where the constraint is not model quality but where your code is allowed to travel. It does completion, IDE chat, and agentic work, with a CLI on the higher tier.

The weakness is that its raw suggestion quality on complex Python trails the top frontier-model tools, which is often an acceptable trade for the privacy guarantees.

Pricing: the Code Assistant plan is $39/user/month and the Agentic plan is $59/user/month, both billed annually, with enterprise quotes available. Best for regulated or privacy-sensitive Python teams.

## 8. Continue

[Continue](https://continue.dev) is the open-source option, licensed under Apache 2.0, available as a VS Code extension, a JetBrains plugin, and a CLI. Its value for Python is control: you choose the model, including local models through [Ollama](https://ollama.com), so you can run a private setup with no per-seat SaaS fee and no code leaving your machine. You can wire it to your own context sources and tailor it to a specific Python codebase.

The trade-offs are real. You assemble and maintain the setup yourself, quality depends entirely on the model you pick, and the project has been reorganizing its repositories, so confirm the current supported packages before you standardize a team on it.

Pricing: free and open source when you bring your own model or API key, with a hosted hub that has paid tiers for teams. Best for developers who want model choice, local inference, or an assistant they can self-host.

## 9. Google Colab AI

Most Python AI tools treat notebooks as second-class. [Colab](https://colab.research.google.com) is built around them. Its Gemini-powered assistant generates cells from natural language, explains errors in context, and understands the notebook's actual execution state, which is exactly the context that trips up general-purpose editors. For exploratory data analysis, quick model prototyping, and teaching, it removes the environment setup entirely because the runtime is hosted.

The limits are obvious: it is a notebook tool, not a place to build and ship a package or a service, and heavier compute or higher AI limits sit behind a paid tier.

Pricing: the free tier includes AI assistance and a modest GPU, and Colab Pro runs about $9.99/month with Pro+ higher, though prices change, so check current pricing. Best for data scientists and anyone whose Python lives mostly in notebooks.

## How to choose for Python

Match the tool to the job, not to the hype. The five stages in the diagram above rarely have a single best tool, so pick by the work that dominates your week.

If your Python work is mostly... | Start with | Why |

Building and shipping services | Cursor or Devin Desktop | Whole-project edits and test loops |

Everyday coding in your current editor | GitHub Copilot | Broad IDE support, reliable completion |

Large refactors and migrations | Claude Code | Terminal agent works across many files |

Data science and notebooks | Google Colab AI or PyCharm | Real notebook-state awareness |

Private or regulated codebases | Tabnine or Continue | On-prem, air-gapped, or local models |

A few Python-specific tie-breakers. If type safety matters to you, favor tools that read your annotations and interpreter (PyCharm-based JetBrains AI and Cursor lead here). If your day is half notebooks, do not force a service-oriented editor to do notebook work; pair a general tool with Colab or PyCharm. If your code cannot leave the building, that constraint outranks everything else, and the choice narrows to Tabnine or a local Continue setup. Most productive Python developers we know run two tools: one for inline completion and one agent for bigger changes. For the mental model behind that split, read AI-assisted coding.

## FAQ

### What is the best AI tool for Python coding?

There is no single winner, because the tools optimize for different jobs. For building and shipping Python services, Cursor is the strongest all-rounder. For everyday completion in the editor you already use, GitHub Copilot is the safe default. For large refactors, Claude Code leads. For notebook-heavy data science, Colab or PyCharm win. Pick by the work that fills most of your week.

### Is there a free AI tool for Python?

Yes. GitHub Copilot, Cursor, and Devin Desktop (formerly Windsurf) all have free tiers, ChatGPT Free includes basic Codex access, PyCharm ships a free AI tier, and Continue is open source and free when you bring your own model or API key. The free plans usually cap monthly requests or completions, which is enough to test each tool before you pay.

### Which AI is best for Python data science and Jupyter notebooks?

Google Colab AI and JetBrains AI in PyCharm are the two that treat notebooks as first-class. Both understand notebook execution state, which general-purpose editors often get wrong when cells run out of order. GitHub Copilot's VS Code notebook support is also good. Editors like Cursor and Devin Desktop can open .ipynb files but are weaker on notebook-specific context.

### Can AI understand Python type hints and catch type errors?

Partly. Tools built on a static analyzer, such as JetBrains AI in PyCharm, use your type annotations and resolved interpreter to make type-aware suggestions and flag mismatches. Editor tools like Cursor and Copilot read your annotations from context and usually respect them, but they can still suggest type-inconsistent code. Keep running a real type checker like [mypy](https://mypy-lang.org) or [pyright](https://microsoft.github.io/pyright/); AI complements it, it does not replace it.

### Is Cursor or GitHub Copilot better for Python?

Copilot is better if you want reliable inline completion inside your existing editor with the least disruption. Cursor is better for whole-project reasoning, multi-file edits, and agent-driven changes to a Python codebase. Many developers use Copilot for daily typing and reach for Cursor or an agent when a change spans many files.

### Can AI help manage Python virtual environments and dependencies?

Agent tools that run commands, such as Claude Code, Cursor's agent, and Codex, can create a venv, install packages, and pin versions in requirements.txt or a pyproject.toml, then verify by running your tests. Inline completion tools only suggest imports and cannot fix a broken environment. Always confirm the assistant is targeting the interpreter your project actually uses.

### Are AI coding tools safe for private or proprietary Python code?

It depends on the tool. Tabnine offers on-premises, VPC, and air-gapped deployment with zero code retention, and Continue can run entirely with local models, so no code leaves your machine. Cloud tools like Copilot, Cursor, and Codex send code to their servers, though business tiers usually add data-handling and no-training guarantees. If your code cannot leave the building, that requirement should decide your choice first.

### Do these tools work with frameworks like Django, FastAPI, and PyTorch?

Yes. Every tool here handles popular Python frameworks, since they are heavily represented in training data. The frontier-model tools (Cursor, Copilot, Claude Code, Codex) are strongest at framework-specific patterns like FastAPI dependency injection or Django model relationships. For PyTorch and other ML code, an assistant helps you write faster but cannot verify that your model logic is numerically correct, so keep your own tests and evaluation in place.

Some offers on this page may be paid placements or contain affiliate links.

---

# Best Newsletters for Developers in 2026: 8 Picks Compared

URL: https://devshot.co/blog/best-developer-newsletters
Type: blog
Published: 2026-09-25
Updated: 2026-09-25
Summary: The best newsletters for developers in 2026, compared by cadence, price and focus: TLDR, The Pragmatic Engineer, ByteByteGo, JavaScript Weekly and 4 more.

Guide

## Best Newsletters for Developers in 2026: 8 Picks Compared

The best newsletters for developers in 2026, compared by cadence, price and focus: TLDR, The Pragmatic Engineer, ByteByteGo, JavaScript Weekly and 4 more.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 10 min read

Key facts8 newsletters compared: TLDR, The Pragmatic Engineer, ByteByteGo, JavaScript Weekly, Bytes…

- Updated: September 25, 2026
- Prices as of: September 2026

- 8 newsletters compared: TLDR, The Pragmatic Engineer, ByteByteGo, JavaScript Weekly, Bytes, Hacker Newsletter, Programming Digest, Devshot
- TLDR (best for: Broad tech and programming news in five minutes): Free
- The Pragmatic Engineer (best for: Engineering leaders and senior ICs at Big Tech scale): Free tier; paid $15/mo or $150/yr
- ByteByteGo (best for: System design and distributed systems depth): Free tier; premium $14.99/mo or $149.99/yr
- JavaScript Weekly (best for: JS and TypeScript ecosystem news and libraries): Free

The best newsletters for developers in 2026 are TLDR, The Pragmatic Engineer, ByteByteGo, JavaScript Weekly, Bytes, Hacker Newsletter, Programming Digest, and Devshot. Six of the eight cost nothing to read. The Pragmatic Engineer and ByteByteGo run a free tier plus a paid deep-dive tier, at $15/month (or $150/year) and $14.99/month (or $149.99/year), verified on each vendor's own site in September 2026. TLDR and Bytes are the strongest daily and twice-weekly picks for a working engineer who wants signal without a second inbox to manage.

## The best developer newsletters at a glance

Newsletter | Publisher | Cadence | Price | Best for |

TLDR | TLDR Media | Daily | Free | Broad tech and programming news in five minutes |

The Pragmatic Engineer | Gergely Orosz | Multiple issues a week | Free tier; paid $15/mo or $150/yr | Engineering leaders and senior ICs at Big Tech scale |

ByteByteGo | Alex Xu, Sahn Lam | Weekly, plus Wednesday on paid | Free tier; premium $14.99/mo or $149.99/yr | System design and distributed systems depth |

JavaScript Weekly | Cooperpress | Weekly | Free | JS and TypeScript ecosystem news and libraries |

Bytes | Fireship | Twice a week | Free | JavaScript and React, with jokes |

Hacker Newsletter | Kale Davis | Weekly | Free | Best of Hacker News without the doomscroll |

Programming Digest | Bonobo Press | Weekly | Free | Cross-language engineering best practices |

Devshot | Dupple | Daily | Free | Coding news and AI-dev workflows in 5 minutes |

Devshot data: our data shows the daily list has grown to 27,887 subscribers with a 30.3% open rate, as of September 2026, per Dupple's Beehiiv audience figures. That readership is the vantage point behind this ranking, including our own newsletter, disclosed openly below.

## 1. TLDR, the fastest daily scan of the whole industry

[TLDR](https://tldr.tech/) is a free daily digest covering AI, startups, programming, cybersecurity, and hardware in short, skimmable blurbs, read by more than 8,000,000 subscribers according to the vendor's own homepage, including people at Anthropic and OpenAI.

Its standout is breadth per minute: a five-minute read touches more of the industry than any other pick here, which is why it works as a first newsletter rather than a specialist one. Watch out: the programming-specific content is one section among several, not the whole issue, so a reader who only cares about code gets diluted signal on days the AI or startup news runs long.

## 2. The Pragmatic Engineer, the pick for engineering leaders

[The Pragmatic Engineer](https://newsletter.pragmaticengineer.com/about), written by Gergely Orosz (previously Uber, Skype and Microsoft), is free for a Tuesday half-deepdive and a Wednesday podcast summary. The paid tier unlocks the full Tuesday deepdive, a Thursday "Pulse" on industry trends, and templates for engineering managers, priced at $15 a month or $150 a year, per the vendor's own pricing page. It has passed 1,100,000 subscribers.

The standout is insider reporting on how Big Tech and high-growth startups actually run engineering, months before mainstream outlets cover the same shift. Watch out: most of the value sits behind the paid tier, so a reader who stays free only sees half of each Tuesday piece and the Wednesday podcast recap, not the Thursday Pulse.

## 3. ByteByteGo, the deepest system-design newsletter

[ByteByteGo](https://blog.bytebytego.com/about), from "System Design Interview" authors Alex Xu and Sahn Lam, sends a free issue every Saturday to more than 1,000,000 readers. Premium members add a second deep dive every Wednesday and get to suggest topics, at $14.99 a month or $149.99 a year, verified on the vendor's own About page today.

The standout is turning genuinely hard distributed-systems topics, like designing for high availability or scaling Netflix, into diagrams a reader can actually follow. Watch out: the free Saturday issue alone is thinner than the paid archive; if system design interviews or architecture reviews are part of your job, the Wednesday deep dive is the reason to pay.

## 4. JavaScript Weekly, the long-running JS ecosystem tracker

[JavaScript Weekly](https://javascriptweekly.com/) from Cooperpress has published every week since November 2010, past 800 issues, and remains completely free with no paid tier. Each issue rounds up JavaScript and TypeScript news, library releases, and articles worth a working frontend or full-stack developer's time.

Its standout is longevity and consistency: more than fifteen years without a missed week is a track record none of the newer entrants here can match. Watch out: it does not cover much outside the JS and TS ecosystem, so a backend engineer working in Go or Python will find most issues irrelevant; our AI for Python coding roundup is a better fit for that audience.

## 5. Bytes, the funniest way to keep up with JavaScript

[Bytes](https://bytes.dev/), made by Fireship, is a free newsletter delivered twice a week to more than 200,000 developers by the vendor's own count, covering JavaScript, React, and the wider frontend ecosystem.

The standout is tone: a running "spot the bug" section and a comedic voice make it the pick readers say they actually finish, not just open. Watch out: the humor-first style is not for everyone, and like JavaScript Weekly, it stays inside the JS and React world rather than covering backend or infrastructure topics.

## 6. Hacker Newsletter, the best of Hacker News without the scroll

[Hacker Newsletter](https://hackernewsletter.com/) is a free weekly curation of the best links from Hacker News, sent on Fridays to more than 60,000 tech professionals per the vendor's own site.

Its standout is filtering: instead of checking Hacker News daily, a reader gets the stories that actually held up over the week, chosen by a human editor rather than an algorithm. Watch out: because it draws only from Hacker News, it misses stories that never trended there, so it is a companion to a broader source, not a full news feed on its own.

## 7. Programming Digest, cross-language best practices

[Programming Digest](https://programmingdigest.net/) is a free weekly newsletter with more than 20,717 subscribers, read by engineers at Microsoft, Amazon, Google, Netflix, Dropbox, and Shopify by the vendor's own count. Each issue hand-picks articles on software engineering practice with a short summary of why it made the list.

The standout is language-agnostic curation: rather than chasing one framework's news cycle, it surfaces API design, architecture, and process writing that applies whatever stack you run. Watch out: because it is broader than any single language, issues can feel less immediately actionable than a JS-specific pick like Bytes if your whole job is one framework.

## 8. Devshot, five minutes on the coding workflows that changed this week

Devshot is Dupple's own free daily newsletter, disclosed here and ranked on the same criteria as every competitor above. It sends each weekday morning, trusted by developers at companies including GitHub, Atlassian, Vercel, Stripe, GitLab, and AWS by the vendor's own count, and as of September 2026 it reaches 27,887 subscribers with a 30.3% open rate.

The standout is focus on how AI is actually changing day-to-day development work, cross-referenced against our own AI for developers and AI-assisted coding coverage so a reader can go deeper the same day. Watch out: at under 28,000 subscribers it is smaller than TLDR or Bytes, so it does not yet have the community and job-board network effects those larger lists have built up.

## How to pick a developer newsletter for your job

Match the newsletter to what you actually need on a given day. A generalist who wants one daily habit should start with TLDR or Devshot, since both fit into five minutes and cover the news that shapes what you build with next. A specialist chasing deeper technical content, system design interviews, or a management track should add The Pragmatic Engineer or ByteByteGo, where the paid tier's Wednesday and Thursday issues are the actual product, not a light upsell.

Do not subscribe to two newsletters that overlap heavily. TLDR and Hacker Newsletter both recap broad tech and programming news, so pick one for that job and add a specialist for the gap it leaves: Bytes or JavaScript Weekly for frontend, Programming Digest for cross-language practice pieces. If your team is deciding how deep to go with tools like Claude Code, Cursor versus GitHub Copilot, or ChatGPT for coding, a newsletter is where you will see the next workflow shift before it hits your team's Slack. The same logic applies to prompt-to-app builders: a newsletter tells you a new one shipped weeks before a roundup like ours gets around to testing it.

Free is the default here, and that is not an accident: developer audiences are large enough that ad and sponsorship revenue covers most of these newsletters without a paywall. The two paid options, The Pragmatic Engineer and ByteByteGo, both keep a genuinely useful free tier, so try the free version for a month before deciding the deep-dive content is worth the subscription. For the wider question of whether AI changes what a developer newsletter should even cover, our will coding be replaced by AI piece is a useful next read.

For a broader, non-developer-specific view of how AI news sources compare across roles, Dupple's [best AI news sources](https://dupple.com/learn/best-ai-news-sources) roundup on our sister site is worth a look.

## Methodology

We checked each newsletter's own site or Substack About page in September 2026 for cadence, price, and subscriber count, and read a sample of recent issues rather than relying on marketing copy alone. Devshot's own figures come from Dupple's live Beehiiv audience data, the same source behind our AI coding statistics page. We did not accept a third-party subscriber estimate for any pick; where a vendor states only a rounded or approximate figure, this page repeats it as stated rather than inventing precision. Inclusion is by editorial judgment of fit for a developer audience.

## FAQ

### What is the best newsletter for developers overall in 2026?

There is no single best pick because the newsletters serve different jobs. For a daily five-minute scan across the industry, TLDR (free, more than 8,000,000 subscribers) or Devshot (free, 27,887 subscribers) fit best. For deeper technical or leadership content, The Pragmatic Engineer and ByteByteGo are the strongest paid picks, each with a real free tier to try first.

### Are developer newsletters free?

Mostly, yes. Of the eight compared here, six are entirely free: TLDR, JavaScript Weekly, Bytes, Hacker Newsletter, Programming Digest, and Devshot. The Pragmatic Engineer and ByteByteGo both run a free tier plus a paid deep-dive tier, priced as shown in the table above.

### What is the best free newsletter for JavaScript developers?

Bytes and JavaScript Weekly are the two picks built specifically for the JS and TypeScript ecosystem. Bytes sends twice a week with a lighter, funnier tone; JavaScript Weekly sends once a week and has run continuously since November 2010, which makes it the more consistent long-term source of the two.

### Is The Pragmatic Engineer worth paying for?

It depends on how much of the content you use. The free tier gets a half deepdive on Tuesdays and a Wednesday podcast summary; the paid tier at $15/month or $150/year unlocks the full Tuesday deepdive plus a Thursday industry-trends piece called The Pulse. If you read every issue and lean on the management and career content, the paid tier is where most of the value sits.

### How often should a developer read a coding newsletter?

Match cadence to the role. Someone who wants a daily pulse on the industry benefits from TLDR or Devshot, both built around a five-minute daily read. A senior engineer or engineering manager usually gets more value from a weekly or twice-weekly deeper read, like The Pragmatic Engineer, ByteByteGo, or Bytes, that filters out day-to-day noise.

### Does Hacker Newsletter replace checking Hacker News directly?

For most people, yes, as a weekly filter. It curates the stories that held up over the week from Hacker News and sends them every Friday to more than 60,000 subscribers, which saves the daily scroll. It will not surface anything that never trended on Hacker News itself, so pair it with a broader daily source if that gap matters to you.

### Why is Devshot on this list if you publish it?

Because a "best developer newsletters" page that leaves out a real, independently rankable option to avoid the appearance of bias would be less useful, not more honest. Devshot is disclosed above as Dupple's own newsletter and judged on the same cadence, price, and audience criteria as every other entry, including a real weakness: its subscriber base is smaller than the largest picks here.

Cite this: Devshot, "Best Newsletters for Developers in 2026," Dupple, September 2026.

Devshot sends the coding news and AI-dev workflows that actually matter for a working developer, in five minutes, every weekday morning, free.

Some offers on this page may be paid placements or contain affiliate links.

---

# ChatGPT for Coding in 2026: What It Nails, What It Fumbles, What to Stop Asking

URL: https://devshot.co/blog/chatgpt-for-coding
Type: blog
Published: 2026-07-17
Updated: 2026-09-25
Summary: Twelve coding tasks sorted by how reliable ChatGPT is: six it nails, four that need checking, two to stop asking, plus chat vs agent and current plan pricing.

Guide

## ChatGPT for Coding in 2026: What It Nails, What It Fumbles, What to Stop Asking

Twelve coding tasks sorted by how reliable ChatGPT is: six it nails, four that need checking, two to stop asking, plus chat vs agent and current plan pricing.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 15 min read

The question worth asking in 2026 is not whether the model can write code: it is whether the job belongs in a chat window at all, or in an agent that has your repository checked out and can run your tests. Those are two products with two failure modes, and picking wrong is why developers conclude that "AI coding" either changed everything or does nothing.

A chat thread is a colleague with an excellent memory of the public internet, no login to your VPN, and no terminal. An agent is that same colleague sitting at your machine with write access. The first is cheap, fast, and safe to be wrong. The second is faster on real work and much more expensive when it is confidently wrong across nine files.

So the useful framing is a reliability ranking, not a list of tricks. Below, twelve coding tasks in three tiers: six where a chat window is close to always right, four where it is right roughly half the time and you carry the verification cost, and two where you should stop asking and go somewhere else.

## Chat window vs coding agent

"ChatGPT" here means the chat interface, the desktop app, or the browser extension: a box you paste code into. It does not read your repository, does not know your branch, and cannot run anything unless you paste the output back in. Every strength and limitation below follows from that one fact.

Here is what OpenAI charges, per [ChatGPT's pricing page](https://chatgpt.com/pricing) as of September 2026:

- Free ($0/month): GPT-5.6 Luna, 27K context window, tight limits on uploads and tools. Fine for a one-off question, not a working session.
- Go ($8/month): more messages, a bigger 54K window, still built for casual chat, not dev workloads.
- Plus ($20/month): GPT-5.6 Sol, GPT-6 models in Codex, 256K reasoning context, expanded Codex access, projects. The tier most developers should buy.
- Pro ($100/month for 5x usage, $200 for 20x): 400K context, maximum Codex tasks. New $200 sign-ups [paused since September 10, 2026](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers). Worth it only if you hit Plus's limits weekly.
- Business (2-seat minimum): admin controls. Buy it for compliance, not coding power.

The context window is the number that decides whether a chat thread is viable for your task. A 27K window on Free will not hold a real file plus its tests plus a stack trace. Plus's 256K reasoning context will hold several files at once, which is roughly the point where pasting stops feeling like a workaround.

OpenAI's own answer to the missing-repo problem is Codex, a separate agentic tool with a CLI and an IDE extension, limited on Free and Go and expanded from Plus up. Codex, Claude Code, Cursor's agent mode, and Copilot's agent mode all compete in the same category: they index your repo, edit multiple files, and run build and test commands in a loop. Claude Code ships inside [Claude Pro ($20) and Max (from $100)](https://claude.com/pricing), and our Claude Code pricing guide covers the limits. Cursor Pro is [$20 a month, and Teams $40 per user, or $16 and $32 billed yearly](https://cursor.com/pricing); our Cursor pricing breakdown has the upper tiers. Copilot is [$10 Pro, $39 Pro+](https://github.com/features/copilot/plans).

Pick the chat window when the task fits in a conversation and you want to understand the answer: explain this, write this function, why does this trace look like that, turn this schema into a query. Pick an agent when the task is defined by the repo rather than by the snippet: rename a concept across twelve files, make the failing suite green, migrate a config format everywhere it appears. Most engineers keep both open.

Settle one thing before you paste: personal accounts can feed model training until you switch it off under Data Controls, while Business and Enterprise are excluded by default. Whatever the tier, strip keys, credentials and client-covered code first. Redacting a variable name costs five seconds; explaining a leak costs considerably more.

## Six things it gets right almost every time

These are the tasks where the model needs no knowledge of your system, only the text you gave it. Verification is cheap, failure is loud, and the hit rate is high enough that you can work fast.

### Explain code you didn't write

The most common use, and the safest. Good for onboarding onto a legacy module, or for reading a dependency's internals before you patch around it.
Explain what this function does, line by line where it's not obvious. Then tell me: what are the edge cases it handles, what happens if `items` is empty or None, and is there anything here that looks like a bug rather than intentional behavior? def reconcile(items, ledger): seen = {} for i in items: key = (i.account_id, i.posted_at.date()) seen.setdefault(key, []).append(i) ...
Paste the whole function, not a snippet. Half a function with no context is how you get a plausible-sounding explanation of the wrong thing. The follow-up that pays for itself: ask which lines it is least sure about, and why.

### Write a function from a tight spec

Give it the exact signature, the types, and one concrete example. A vague description gets a vague function back, and you spend the time you saved rewriting it.
Write a Python function `chunk_by_weight(items: list[dict], max_weight: float) -> list[list[dict]]` that splits a list of items (each with a "weight" key) into sublists where no sublist's total weight exceeds max_weight. Preserve original order. If a single item's weight exceeds max_weight on its own, put it alone in its own sublist rather than raising. Include a docstring and one usage example in the docstring.
Name the edge case you actually care about (here, the oversized single item) or you get a naive implementation that breaks on real data. The rule of thumb: whatever you leave unspecified, you are delegating to a guess.

### Generate boilerplate

The highest-volume, lowest-risk category. A Dockerfile, a CI workflow, a CRUD router skeleton, a Terraform block: pattern-heavy, well represented in public code, and cheap to verify.
Write a multi-stage Dockerfile for a Node 20 TypeScript app: install deps and build in a builder stage, copy only the compiled dist and production node_modules into a slim runtime stage, run as a non-root user, and expose port 3000. Add a HEALTHCHECK that curls /health.
Boilerplate is where a subscription pays for itself fastest: a wrong Dockerfile fails loudly in CI rather than shipping a silent bug. Failure mode matters more than raw accuracy when you decide how much to trust output.

### Build a regex you don't want to write by hand

Regex is pure pattern matching. The model needs to know nothing about your system to get it right, which is exactly the condition under which it performs best.
Write a regex that matches valid semantic version strings (e.g. 1.2.3, 2.0.0-beta.1, 1.4.10+build.5) and captures major, minor, patch, and an optional prerelease/build suffix as named groups. Show the pattern, name each group, and give me 5 example strings it should match and 3 it should reject.
Always ask for the rejection examples too. A pattern that matches everything you want but also swallows garbage you never tested is the standard way this goes wrong, and the rejection list turns a review into a thirty-second check.

### Write SQL from a schema you describe

Paste the schema, not a description of the tables. Guessed column names produce queries that read beautifully and fail against your actual database.
Given this schema, write a PostgreSQL query that returns each customer's total spend in the last 90 days, only for customers with at least 3 orders in that window, sorted by spend descending. orders(id, customer_id, total_cents, created_at) customers(id, email, created_at) Use created_at for the date filter, total_cents for spend (convert to dollars in the output), and explain the query plan implications of the HAVING clause you use.
Run it against staging first, and read the plan yourself on large tables. It does not know which indexes exist, so its performance commentary is generic advice, not analysis of your database. Correctness of the logic is the strong part; cost is your problem.

### Learn a new API from docs you paste in

Do not ask it to recall a library's usage from memory. Paste the current README section or docs page and constrain the answer to that text.
Here's the README section for a library I haven't used before. Based only on what's in this text, show me a minimal working example that does [specific thing], and tell me if anything I'm asking for isn't covered by what you can see here. [paste the README section or docs page]
That last clause is the whole trick. "Tell me if what I'm asking for isn't in here" converts a generator into a reader and gives it an explicit way to say no instead of filling the gap.

## Four it gets right about half the time

These tasks depend on things the chat window cannot see: your runtime, your data, your team's history. The output is often useful and occasionally wrong in ways that look fine. Budget verification time, or do not start.

### Debug from a stack trace

Paste the full trace, not just the last line. It needs the call chain to reason about where the bug lives versus where it merely surfaced.
I'm getting this error and I don't understand why, since the function above it checks for None. Walk me through what's likely happening, and give me two hypotheses ranked by likelihood. Traceback (most recent call last): File "worker.py", line 88, in process_batch result = handler(item) File "worker.py", line 41, in handler return item.metadata["source"]["id"] TypeError: 'NoneType' object is not subscriptable # handler(): def handler(item): if item is None: return None return item.metadata["source"]["id"]
It spots the mismatch well: the guard covers item, not item.metadata["source"]. What it cannot tell you is why source is None in production, because that lives in your data. Two ranked hypotheses instead of one answer is the difference between a lead and a confident dead end.

### Write tests

Name the framework and the cases that matter. Left unguided, it writes happy-path tests and calls the job done.
Write pytest unit tests for this function. I want cases for: normal input, an empty list, a single item over max_weight, all items exactly at max_weight, and negative weights (should raise ValueError). Use parametrize where it reduces duplication. Don't test implementation details, test behavior. def chunk_by_weight(items, max_weight): ...
"Write tests for this" on its own gets you three trivial assertions that miss the case that breaks in production. And the model cannot run what it wrote: a test that passes in its head may not even import in your project. Half the value here is that enumerating the cases yourself forces you to think about the boundaries.

### Refactor without changing behavior

State the constraint explicitly: same inputs, same outputs, different internals. Otherwise you get a rewrite that "improves" behavior something else depended on.
Refactor this function to reduce nesting and improve readability. Do not change any behavior, including error messages and the order operations happen in. Explain each change in one line so I can verify nothing shifted. def validate_order(order): if order: if order.get("items"): if len(order["items"]) > 0: for item in order["items"]: if not item.get("sku"): return False, "missing sku" return True, None
Diff the output line by line. A tidy early return can silently change which error message a caller sees first, and that drift passes review because the new code looks better. The one-line justification per change is what makes the diff reviewable.

### Second opinion on a diff

Paste the actual diff, not a description of it. Ask for named categories of feedback, or you get a summary of your own change back.
Review this diff as if you were a strict senior engineer. Flag: 1) anything that could break in production but isn't covered by a test, 2) unclear naming, 3) any place error handling swallows a real failure, 4) anything that duplicates logic that probably exists elsewhere in a codebase like this. Be specific about line numbers, don't just summarize. [paste the diff]
Treat it as a pre-review pass that catches the easy things before a human spends attention on them. It cannot know whether the change matches last week's design decision or whether the duplicated helper it flagged exists on purpose, and it has no accountability for what merges, which is most of what a review is.

## Two things to stop asking it

Not "be careful with these." Stop. The expected value is negative because both failures are silent.

### Current library versions and API surfaces from memory

Every model has a training cutoff, and none of them know about last week's release. Ask for usage of a fast-moving library from memory and you can get a method that never existed, an argument order from two majors ago, or an approach deprecated with a loud warning you will not see until runtime. It is worse for smaller libraries with less public code around them, which are exactly the ones you were most likely to need help with.

There is no prompt that fixes recall. There is only a different question:
I'm pasting the current docs for [library] version [X.Y]. Using only this text, show me how to do [task]. If the docs I pasted don't cover it, say so instead of filling in the gap.
Same need, different mechanism: reading instead of remembering. If the docs are too long to paste, read them yourself rather than trusting recall.

### A security review of code that ships

A chat window can point out a string-concatenated query or a missing input check, and it will do that reliably enough to be worth a pass on a hobby project. It cannot do a security review, because a review is about your threat model, your trust boundaries, your auth flow, and what an attacker can reach from outside, none of which is in the snippet you pasted.

The specific danger is that generic requests get generic answers. Ask for "a login function" and you get the simple pattern that appears most in public code, not a hardened one: plausible session handling, defaults nobody hardened, error paths that leak which half of the credentials was wrong. It compiles, it passes your tests, and it looks like code a reasonable person wrote. That is precisely why it survives review.

Use it for the cheap pass (ask explicitly for parameterised queries, output encoding, and secure defaults), then send anything touching authentication, payments, permissions, or data deletion to a person, a scanner, and where the stakes justify it, an audit. The costly failure in this whole category is never visibly broken code. It is code that looks correct, compiles, and fails on the case you did not test.

If you want to keep track of which of these tools actually changed month to month, Devshot writes up AI and developer tooling every day.

## What these tools actually cost

We price every tool we review, so this is measured rather than estimated. Across 429 tools, 293 publish a price and 33% offer a free tier. Among developer tools, the median entry plan is $24.50 a month, which runs above the $24 median across every category we price.

The spread matters more than the median. Half of the developer tools sit between $10 and $49, and the range runs from $2.49 to $299. A quoted "starting at" price near the bottom of that range usually means per-seat add-ons land on top of it.

Price point | Developer tools | All tools |

Cheapest paid plan | $2.49 | $1 |

Lower quartile | $10 | $10 |

Median | $24.50 | $24 |

Upper quartile | $49 | $49 |

Most expensive | $299 | $990 |

Tools measured | 18 | 293 |

Median advertised entry price/mo. Source: Dupple pricing index, 293 tools with public pricing out of 429 reviewed, 2026-08-19.

## FAQ

### Should I be using ChatGPT or a coding agent like Codex, Claude Code, or Cursor?

Both, for different shapes of work. The chat window is for conversation-sized tasks where you want to understand the answer: explain this, draft this function, turn this schema into a query. An agent is for repo-sized tasks where you want the work done: multi-file changes, making a failing suite pass, mechanical migrations. If you find yourself pasting five files into a chat thread to give it context, that is the signal to switch.

### Which plan do I actually need as a developer?

Plus at $20/month, for almost everyone (our ChatGPT vs Claude for coding comparison covers the Claude side). It gets you the current model, a 256K reasoning context that holds real files rather than fragments, and expanded Codex access. Pro ($100 a month; the $200 tier is closed to new sign-ups) only pays off if you hit Plus's limits weekly. Free's 27K window will not survive a real file plus its stack trace, so treat it as a trial rather than a tier.

### Why does it invent methods and arguments that don't exist?

Because it is producing the most plausible continuation of your prompt, and for a library it saw rarely during training, a method that should exist looks a lot like one that does. The rate goes up for small, new, or fast-moving packages. The fix is mechanical rather than clever: paste the current docs and constrain the answer to what is in them, and explicitly give it permission to say the docs do not cover your case.

### Can I paste proprietary code into it?

Check your workspace policy first, then the account settings: personal accounts can be used for model improvement until you switch that off under Data Controls, and Business and Enterprise are excluded by default. Strip keys, tokens and customer names before pasting. If the code is under an NDA, that is a question for whoever signed it, not for a settings toggle.

### Can it run my test suite?

No. It can execute Python in a sandboxed session for data analysis, but that sandbox has no access to your repository, your dependencies, your database, or your CI. It writes tests you run. That gap is also why "it says this passes" means nothing on its own, and why anything that depends on runtime state (flaky CI, a race condition, a slow query under real load) is a guess dressed up as an answer.

### Is a ChatGPT pass enough to skip human code review?

No. It is a good pre-review filter: run it before you request review and you waste less of a colleague's attention on naming, missing tests, and swallowed errors. What it cannot supply is the actual function of review, which is a person with context on your system agreeing to be accountable for what merges. Use it to arrive at review with a tighter diff, not to arrive with none.

Some offers on this page may be paid placements or contain affiliate links.

---

# ChatGPT vs Claude for Coding in 2026: Which Is Better?

URL: https://devshot.co/blog/chatgpt-vs-claude-for-coding
Type: blog
Published: 2026-07-21
Updated: 2026-09-25
Summary: ChatGPT vs Claude for coding in 2026: a detailed breakdown of code generation, debugging, agents (Codex vs Claude Code), long context, and price.

Guide

## ChatGPT vs Claude for Coding in 2026: Which Is Better?

ChatGPT vs Claude for coding in 2026: a detailed breakdown of code generation, debugging, agents (Codex vs Claude Code), long context, and price.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 11 min read

Ask which one is better and you get a benchmark screenshot. That is the wrong lens. On a single well-specified function, [ChatGPT](https://chatgpt.com) and [Claude](https://claude.com) now produce code that is close to interchangeable, and whichever one topped the leaderboard last month will not be the one on top when you read this. The benchmark race stopped being the interesting question sometime in 2025.

The contrarian read after living in both daily: the model you pick barely decides the outcome anymore. The agent wrapped around it does. OpenAI's [Codex](https://openai.com/codex/) and Anthropic's [Claude Code](https://claude.com/product/claude-code) behave very differently inside a long agentic loop, and that difference (how eagerly each one edits, how it handles a large repo, how it reacts when it is unsure) matters far more to your day than whose one-shot function is 3% cleaner. So "which is better" resolves to "better for which workflow." This piece answers that, round by round, with current 2026 models and honest limits, and no invented scores.

## Quick verdict

Dimension | ChatGPT (OpenAI) | Claude (Anthropic) |

Flagship coding model | GPT-6 Astra | Claude Opus 5.5 |

Balanced daily model | GPT-5.6 Sol in ChatGPT, GPT-6 Sol in the API | Claude Sonnet 5 |

One-shot code generation | Excellent, close to a tie | Excellent, close to a tie |

Debugging a stack trace | Strong, direct | Strong, tends to explain more |

Large codebase / long context | Very good | Very good, both offer 1M-token context |

Agentic CLI | Codex (eager, fast, wide edits) | Claude Code (cautious, tight diffs) |

IDE integration | Codex extension, deep ChatGPT reach | Claude Code in VS Code and JetBrains |

Refactor discipline | Good, sometimes over-reaches | Good, tends to ask before large changes |

Entry price for devs | ChatGPT Plus about $20/mo | Claude Pro about $17 to $20/mo |

## Round by round

### Code generation

For writing one function from a clear spec, this is a coin flip, and anyone telling you otherwise is selling a leaderboard. Both GPT-6 Astra and Claude Opus 5.5 handle types, edge cases you name explicitly, and idiomatic style in the major languages without trouble. The mid tiers, GPT-5.6 Sol (ChatGPT's everyday Plus model) and Claude Sonnet 5, are good enough for most day-to-day generation that you will rarely reach for the flagship on a single function.

Small tendencies show up with repetition rather than in any one output. Claude tends to write more defensive code by default, adding the null check and the early return you did not ask for, which is welcome until it is noise. ChatGPT tends to be more literal, giving you exactly the shape you specified and less scaffolding around it. Neither is a durable advantage. Name your edge cases in the prompt and both converge on the same answer.

### Debugging

Paste a full stack trace and both models reason well about where the bug actually lives versus where it merely surfaced. The difference is bedside manner. Claude tends to walk you through the call chain and explain why, which is useful when you are learning an unfamiliar module and slower when you already know and just want the fix. ChatGPT tends to be more direct, leading with the likely cause and the patch.

The shared, hard limit is the same for both: neither can see your running system. Ask why source is null in production and you get a plausible hypothesis, not an answer, because neither has your data, your environment, or your logs unless you paste them in. For a deeper split of local debug assistants versus production observability AI, see the best AI for debugging guide.

### Large codebase and long context

Both flagship tiers now offer a [1M-token context window](https://docs.claude.com/en/docs/build-with-claude/context-windows), which on paper is roughly 555,000 words or a mid-sized service in one prompt. In practice, raw window size is not the story. What matters is whether the model stays coherent deep into that window, and both [degrade the same way](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) every long-context model does: instructions from early in the prompt get quietly dropped, and a detail you pinned at the top loses to something louder near the bottom.

The realistic takeaway is that neither wins by having more context, because you should not be dumping a whole monorepo into either one and hoping. The tool that reads only the files it needs beats the tool that swallows everything, which is exactly why the agent layer matters more than the raw window. That is the next round.

(Devshot breaks down the AI-for-developers news in five minutes each morning.)

### Agentic CLI: Claude Code vs Codex

This is where the two stop being interchangeable. Both ship a terminal agent that reads your repo, edits across many files, and runs build and test commands in a loop. The default temperament is where they diverge, and it is consistent enough to plan around.

Codex (running GPT-6 Astra from the Plus plan up alongside the GPT-5.6 models, per OpenAI's [model guide](https://help.openai.com/en/articles/20001354-gpt-56-and-gpt-6-pro-in-chatgpt), with a CLI, an IDE extension, a desktop app, and a cloud mode for parallel runs) tends to move fast and edit wide. Give it a task and it will often touch more files than you expected, get further before checking in, and lean toward doing rather than asking. That is a gift when the task is well-scoped and a liability when it is vague, because an eager agent given an ambiguous instruction produces a large diff you now have to unwind.

Claude Code (running Claude Opus 5.5 or Sonnet 5, in the terminal and in VS Code and JetBrains, with support for many parallel subagents) tends to be more cautious. It is more likely to stop and confirm before a large change, keep diffs tighter, and stage the work. That is a gift when you want to stay in the loop and a liability when you wanted it to just finish while you got coffee.

Neither temperament is correct in the abstract. The match is to your risk tolerance on a given task. Scoped, mechanical work that you trust the tool to grind through rewards the eager agent. Sensitive, multi-file changes where a wrong turn is expensive reward the cautious one. Many engineers keep both installed and choose per task. For the wider category including [Cursor](https://cursor.com) and Copilot, see best AI coding assistants.

### IDE integration

Both have moved past copy-paste. Codex offers an IDE extension plus the pull of the broader ChatGPT product, so the same subscription that answers your general questions also drives the agent. Claude Code [integrates natively](https://code.claude.com/docs/en/overview) into VS Code (including Cursor and Devin Desktop) and JetBrains IDEs, and connects to Git, GitHub, GitLab, and MCP servers.

If you want one vendor for coding and everything else in a single tab, ChatGPT plus Codex is the tighter bundle. If you want a coding-first agent that lives in your editor and terminal, Claude Code is the more focused tool. This is a preference, not a winner.

### Refactoring

Refactoring exposes the temperament split more than any other task. Ask for a behavior-preserving cleanup and Claude tends to hold the constraint, asking before it changes an interface and keeping the diff auditable. ChatGPT, especially through the eager Codex agent, sometimes improves behavior you were relying on, renaming or restructuring past the line you drew.

The practical rule applies to both: state the constraint out loud (same inputs, same outputs, same error messages, same operation order) and diff the result yourself. A cleaned-up early return can silently change which error a caller sees first, and no model reliably catches that for you. The best AI for code review guide covers the PR-review bots that add a second automated pass on top.

### Price

Pricing moves, so confirm the live pages before you commit, but here is the shape as of September 2026.

| ChatGPT / OpenAI | Claude / Anthropic |

Entry dev subscription | Plus, about $20/mo, includes Codex access | Pro, about $17/mo annual or $20 monthly, includes Claude Code |

Higher tier | Pro, from about $100/mo | Max, from about $100/mo (5x or 20x usage) |

Flagship API, per 1M tokens | GPT-6 Astra, [$10 in / $50 out](https://developers.openai.com/api/docs/pricing) | Opus 5.5, [$4 in / $20 out](https://platform.claude.com/docs/en/about-claude/models/overview) |

Balanced API, per 1M tokens | GPT-6 Sol, $2 in / $10 out | Sonnet 5, $2 in / $10 out |

For most solo developers the entry subscriptions are close enough that price is not the deciding factor. Both bundle a capable agent into the entry plan. The API rates matter only if you are building on top of the models rather than coding with the assistant, and there the balanced tiers (GPT-6 Sol, Sonnet 5, priced identically) are the sensible default before you reach for a flagship.

## Who should pick what

Pick Claude if you value a cautious agent that keeps you in the loop, you do most work through a terminal and editor, and you want the tool that asks before it rewrites an interface. Opus 5.5 is Anthropic's recommended starting point for long-running agentic coding, per its [models overview](https://platform.claude.com/docs/en/about-claude/models/overview), and Claude Code is the more coding-focused product.

Pick ChatGPT if you want one subscription for coding and everything else, you prefer an agent that gets further on its own before checking in, and you already live in the ChatGPT product. Codex on GPT-6 Astra is the eager, wide-reaching agent, and Plus bundles it at the same rough price as a coding-only plan.

Pick both if you ship real work daily, which is what many engineers actually do. Use the cautious agent for sensitive multi-file changes and the eager one for scoped grind work, and stop treating this as a loyalty test. The combined cost is roughly two $20 subscriptions, cheap against an hour of your time.

The one honest caveat: any specific claim here about which model tops a benchmark this quarter will age fast. The temperament differences and the workflow fit hold up far better than the leaderboard, which is exactly why you should choose on how you work, not on last month's score. For the full map of AI across the development lifecycle, start with the AI for developers hub, and for raw-ChatGPT prompts and their failure modes see ChatGPT for coding.

## FAQ

### Is Claude or ChatGPT better for coding in 2026?

Neither wins outright. On a single function they are close to interchangeable. The real difference is the agent: Claude Code tends to be cautious and diff-tight, Codex tends to be eager and wide-reaching. Choose on which behavior fits your work, not on a benchmark that changes monthly.

### Claude Code vs Codex: what is the actual difference?

Both are terminal agents that edit across files and run your tests in a loop. Codex (on GPT-6 Astra) leans toward doing more autonomously before it checks in and touching more files. Claude Code (on Opus 5.5 or Sonnet 5) leans toward confirming before large changes and keeping diffs smaller. Same category, opposite defaults.

### Which handles a large monorepo better?

Both flagship tiers offer a 1M-token context window, and both degrade the usual way when you fill it. The winner is whichever agent reads only the files it needs rather than swallowing the repo. That is an agent-behavior question more than a model question, so test both on your actual codebase before deciding.

### Is Claude actually better at coding, or is that just reputation?

Reputation runs ahead of the current gap. Claude earned a strong coding name over several model generations, and Anthropic positions Opus 5.5 for long-running agentic coding, but ChatGPT's flagship closed most of the one-shot distance. On raw generation today it is close. The durable difference is temperament, not a quality gulf.

### Which is cheaper for a solo developer?

Roughly equal. ChatGPT Plus is about $20 a month and includes Codex. Claude Pro is about $17 annual or $20 monthly and includes Claude Code. Both higher tiers start near $100. Price is rarely the deciding factor at the individual level, so verify the current pages and pick on fit.

### Which hallucinates less on unfamiliar libraries?

Both invent plausible-but-fake APIs when asked to code against a library from memory, and both do it more with smaller, faster-moving packages. There is no reliable winner here. The fix is identical for both: paste the current docs instead of trusting recalled API knowledge, since neither has a live changelog.

### Do I need the flagship model, or is the mid tier enough for coding?

For most day-to-day coding the balanced tiers, GPT-5.6 Sol (or GPT-6 Sol in the API) and Claude Sonnet 5, are enough and cheaper. Reach for GPT-6 Astra or Claude Opus 5.5 on genuinely hard, multi-step agentic tasks where the extra reasoning earns its cost. Starting on the mid tier and escalating only when you hit a wall is the sane default.

### Can I use both ChatGPT and Claude together?

Yes, and many engineers do. A common split is the cautious agent for risky multi-file changes and the eager agent for well-scoped grind work, or one vendor for coding and the other as a second opinion on a diff. Both entry plans are cheap enough that running them in parallel is a rounding error against your time.

Some offers on this page may be paid placements or contain affiliate links.

---

# Claude Code Pricing in 2026: Plans, Usage Limits and API Costs

URL: https://devshot.co/blog/claude-code-pricing
Type: blog
Published: 2026-09-25
Updated: 2026-09-25
Summary: Claude Code pricing in 2026: it rides on a Claude subscription starting at Pro, with Max, Team and Enterprise tiers above…

Guide

## Claude Code Pricing in 2026: Plans, Usage Limits and API Costs

Claude Code pricing in 2026: it rides on a Claude subscription starting at Pro, with Max, Team and Enterprise tiers above it, plus a separate per-token API rate.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 13 min read

Key facts8 plans compared: Free, Pro, Max 5x, Max 20x, Team Standard, Team Premium, Enterprise, Console…

- Updated: September 25, 2026
- Prices as of: September 2026

- 8 plans compared: Free, Pro, Max 5x, Max 20x, Team Standard, Team Premium, Enterprise, Console (API)
- Free (best for: Chat only, no Claude Code): $0
- Pro (best for: Short sessions on a small codebase): $20/mo, or $17/mo annual
- Max 5x (best for: Daily terminal use): $100/mo
- Max 20x (best for: All-day agent work): $200/mo

Claude Code has no price tag of its own. It rides on a Claude subscription, starting at Pro for $20/mo, and there is no free tier that unlocks it. Heavier terminal use pushes people up to Max, and a Team seat, priced per member, adds SSO and admin controls on top.

These figures were read on Anthropic's [pricing page](https://claude.com/pricing) and the [Claude Code product page](https://claude.com/product/claude-code) on September 25, 2026. Claude Code is Anthropic's terminal coding agent, and the confusing part is not the plan grid, it is that chat and code draw from one shared usage pool, so a long debugging session can eat into the allowance you meant to spend talking to Claude later that day.

Toolradar data: the [AI coding ranking](https://toolradar.com/best/ai-coding) evaluated 153 tools, updated September 2026. The setup guide on this site is how to use Claude Code, the product writeup is the Claude Code review, and the directory record is [Claude Code on Toolradar](https://toolradar.com/tools/claude-code).

Plan | Price | Best for | What changes the bill |

Free | $0 | Chat only, no Claude Code | Does not include Claude Code at all |

Pro | $20/mo, or $17/mo annual | Short sessions on a small codebase | Shares one pool with chat, weekly cap |

Max 5x | $100/mo | Daily terminal use | 5x the Pro allowance, still poolwide |

Max 20x | $200/mo | All-day agent work | 20x the Pro allowance |

Team Standard | $25/user/mo, or $20/user annual | A team that wants seats, not API keys | 1.25x Pro per seat, 2-seat minimum |

Team Premium | $125/user/mo, or $100/user annual | Heavy users on the same team | 6.25x Pro per seat |

Enterprise | From $20/seat/mo annual, plus usage | Governance, SSO, audit logs | Newer seat plans meter usage like the API |

Console (API) | Pay per token | Building on Claude, not chatting | Priced per model, per million tokens |

## There is no free way into Claude Code

The Free plan covers chat on the web, desktop and mobile, web search, file creation, code execution and up to five projects, and it costs nothing. It does not include Claude Code. Anthropic's own product page draws the line plainly: Claude Code is included starting on Pro, not on Free.

That is the opposite of how Cursor or GitHub Copilot onboard people. Cursor's Hobby tier and Copilot's Free tier both let a developer try the agent for nothing before paying. With Claude Code, the cheapest way to open a terminal session is a paid subscription, so there is no zero-cost trial of the actual product, only of the chat app around it.

## Pro shares its pool with chat

Pro is billed monthly, or $17/mo if the year is paid upfront. That price is not new: it is the same tier Anthropic has sold for chat since before Claude Code existed, now extended to cover the terminal agent too.

The one thing to understand before buying Pro for coding is the shared pool. Claude Code, Claude chat, Claude Design and Claude Science all draw from the same rolling five-hour session limit and the same weekly cap. Anthropic does not publish an exact prompt count for Pro, because usage is metered by tokens processed rather than messages sent, and a large file read burns the allowance faster than a short question does. A person who chats with Claude during the day and then opens Claude Code at night is spending from one shared meter, not two separate ones.

Pro suits short coding sprints in a small repo, the phrasing Anthropic itself uses on the product page. A single bug fix, a small script, or an occasional refactor fits inside that allowance most weeks. A daily multi-file agent habit usually does not, and that gap is the signal to look at Max.

## Max 5x and Max 20x are multiples, not new features

Max comes in two sizes, and both buy the same product as Pro, just more of it. Max 5x is $100/mo and gives five times the Pro plan's per-session allowance. Max 20x is $200/mo and gives twenty times that allowance. Neither tier unlocks a capability Pro lacks; the terminal agent, the models and the tool access stay identical across all three.

The step from Pro to Max 5x is a five-times price jump for a five-times usage jump, so the per-token value does not change, only the size of the tank. Anthropic recommends the smaller Max tier for people who work with Claude across a variety of tasks most days, and the larger one for daily users who lean on Claude for most of their work, including long agent runs that would otherwise hit the Pro ceiling by lunchtime.

Skip Max until you can name the week Pro ran out, not the week you assume it will. Claude Code does not warn you in advance how close you are to a limit, so the honest test is whether you have actually been interrupted mid-task, not whether a heavier plan sounds safer.

## Team seats add SSO, still metered per person

Team Standard is $25 per member per month billed monthly, cheaper on the annual rate shown above, with a two-seat minimum and room to grow to 150 seats before Enterprise becomes the only option. Team Premium is $125 per member per month monthly, or $100 annually, for members with a heavier workload.

Each seat carries its own allowance rather than a shared team pool: Standard gives 1.25x the Pro plan's per-session usage, and Premium gives 6.25x. A quiet teammate's leftover allowance does not top up a busy one's, so a five-person team where two live in Claude Code all day and three barely touch it needs a mix of seat tiers, not five identical Standard seats. Team plans add SSO, an admin console, enterprise search and centralized billing on top of the usage, none of which personal Pro or Max accounts include.

Claude Code ships on every Team seat, so a team does not pay extra to turn it on. The seat price is the whole story until someone wants more than the seat's included allowance, which is where usage credits come in.

## Enterprise now has two different shapes

Enterprise pricing on claude.com/pricing lists a base seat rate billed annually, with usage on top metered at API rates. That is a newer, seat-based Enterprise plan, and it reads almost like a cheap Team seat with usage layered over it rather than folded into the price.

Anthropic's support documentation describes a second, older shape too: a usage-based Enterprise plan, including self-serve Enterprise, with no per-seat usage limits at all, because every token is billed at API rates from the start with no included allowance to run out of. Which shape an organization is on changes the whole budgeting conversation. A seat-based Enterprise customer watches a per-seat ceiling the way a Team customer does, while a usage-based Enterprise customer watches a token meter the way an API customer does. Neither shape has a single public number that applies to every buyer, so Anthropic's guidance for both is the same: talk to an account manager.

## The API is a separate meter, priced per model

Outside a subscription, Claude Code can run against the Claude API through the Claude Console, and that path bills by the token, not by the seat. Prices below are per million tokens, read on Anthropic's [API pricing docs](https://platform.claude.com/docs/en/about-claude/pricing) on September 25, 2026.

Model | Input | Output | 5-min cache write | Cache read |

Claude Sonnet 5 | $2/MTok | $10/MTok | $2.50/MTok | $0.20/MTok |

Claude Opus 5.5 | $4/MTok | $20/MTok | $5/MTok | $0.20/MTok |

Claude Opus 5.5, fast mode | $8/MTok | $40/MTok | Stacks on top of fast pricing | Stacks on top of fast pricing |

Claude Haiku 4.5 | $1/MTok | $5/MTok | $1.25/MTok | $0.10/MTok |

Sonnet 5 is the model Anthropic's own docs point developers toward for most coding work, with Opus reserved for harder architectural decisions. The API has no monthly ceiling and no shared pool with chat, so a solo developer running Claude Code against an API key pays only for the tokens a session actually burns, and a session that reads a large file and rewrites it across several turns can add up fast. Anthropic's own cost-management guidance is the best first-party anchor here: across enterprise deployments, the average developer spends around $13 on an active day and $150 to $250 a month, and 90% of users stay under $30 on their heaviest days.

Batch processing, for non-interactive jobs, cuts both input and output token prices in half on every model, but it does not apply to interactive Claude Code sessions, only to asynchronous API calls. For the per-token rates behind the Claude API itself, see best LLM APIs in 2026.

## What ten developers pay next to Cursor, Copilot and Devin

Every figure below was checked on that vendor's own pricing page on September 25, 2026. At the individual entry tier, Claude Pro, [Cursor](https://cursor.com/pricing) Pro and [Devin](https://devin.ai/pricing) Pro all land on the same sticker, and [GitHub Copilot](https://docs.github.com/en/copilot/concepts/billing/individual-plans) Pro undercuts all three at $10/mo with 1,500 credits. At the heaviest individual tier, Claude's larger Max plan, Cursor's Ultra and Devin's Max all match that same sticker too, well above anything Copilot sells solo, since Copilot's top individual plan is Pro+ at $39/mo with 7,000 credits.

Seat math is where the four stop looking interchangeable. This table is only the invoice; the product choice, editor versus terminal agent, is covered in the Cursor review and Cursor vs GitHub Copilot.

Buyer | Claude Code | Cursor | GitHub Copilot | Devin |

10 people, monthly seats | Team Standard, $250 | Teams Standard, $400 | Business, $190 | Teams, $80 base + $400 in seats |

10 people, cheaper interval | Team Standard, $2,400/yr | Teams Standard, $320/mo billed annually | No published annual rate | No published annual rate |

Cursor's Teams Standard seat runs well above Claude's Team Standard seat, which is the trade-off between an editor built around Cursor's own model router and a terminal agent that shares Claude's subscription. Copilot's Business seat, at $19/user/mo, is the cheapest seat of the four, and plain code completions stay unlimited and off its credit meter on every paid tier, something neither Claude Code nor Cursor offers at any price. Devin's team plan adds an $80/mo platform fee on top of $40/mo per full seat, a line item Claude Code's Team plan does not charge. To see these plan prices against Cursor, including a 10-seat team cost, read Claude Code vs Cursor 2026.

## Fees that never show up on the plan card

The subscription price is the seat, not the whole month. Usage credits let a Pro, Max, Team or Enterprise member keep working past their included allowance, billed at standard API rates once turned on, and a member with no spend limit set can run that meter with no ceiling. A team that never checks the spend report can be paying full seat prices and a growing usage-credit bill at the same time without anyone noticing until the invoice lands.

A long session that sits open for hours can burn far more of the weekly allowance than the actual work suggests, because Claude Code resends the whole conversation with every tool call, and a break longer than the prompt cache's lifetime forces a full, more expensive reprocess on the next message. Data residency, when an organization pins inference to the US, adds a 1.1x multiplier on every token category on the API, a fee that never shows up on a subscription plan card because it only applies to Console and cloud-platform billing.

Devshot data: this brief goes to 28,000 developers, and the audience file, updated September 2026, shows a 30% open rate and a 1.8% click rate. Plan and pricing changes for that list run in the Devshot brief.

## Which plan is worth paying for

For one person, start on Pro and move to the smaller Max tier only after a real week has run out, not because a bigger number sounds safer. The larger Max tier is for people who already know their agent runs span most of the day; buying it up front on a hunch pays for headroom nobody is using yet.

For a team, Team Standard is the cheaper way in than API keys for most groups, and Premium seats belong only on the people who will actually burn 6.25x the allowance, not on the whole roster by default. If your team already knows its token spend from running the API directly, price that against Team seats before committing, since the [API pricing docs](https://platform.claude.com/docs/en/about-claude/pricing) let you model exact costs per model rather than estimating against a pooled seat allowance.

If you are still choosing the tool rather than the plan, how to use Claude Code covers the setup, and best AI coding assistants and AI for developers cover where it sits against the rest of the category. For the prompt-to-app end of the market instead of a terminal agent, best vibe coding tools is the right list, and if OpenAI's agent is the other option on your shortlist, see the Codex review. If you are weighing Claude Code against the free, open-source OpenCode agent, OpenCode vs Claude Code 2026 prices one month of agent work both ways.

## How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We checked Anthropic's [pricing page](https://claude.com/pricing), the [Claude Code product page](https://claude.com/product/claude-code), the [API pricing docs](https://platform.claude.com/docs/en/about-claude/pricing), the [Claude Code cost management guide](https://code.claude.com/docs/en/costs), the [Max plan article](https://support.claude.com/en/articles/11049741-what-is-the-max-plan), the [Team plan article](https://support.claude.com/en/articles/9266767-what-is-the-team-plan) and the [Team and Enterprise Claude Code article](https://support.claude.com/en/articles/11845131-use-claude-code-with-your-team-or-enterprise-plan) on September 25, 2026.

Cursor figures come from [cursor.com/pricing](https://cursor.com/pricing). Copilot figures come from GitHub's [individual plans](https://docs.github.com/en/copilot/concepts/billing/individual-plans) docs. Devin figures come from [devin.ai/pricing](https://devin.ai/pricing). The 153 tool count is from Toolradar's live AI coding ranking that same month. More on how we publish is on the Devshot about page.

## FAQ

### How much does Claude Code cost in 2026?

There is no standalone Claude Code price. It rides on Pro, and above that on Max 5x or Max 20x. Team Standard seats are $25/user/mo, and Team Premium seats are $125/user/mo. Building against the API instead of a subscription bills per token with no monthly ceiling.

### Is Claude Code free?

No. The Free Claude plan does not include Claude Code at all, only chat, web search and a handful of other features. The cheapest way to open a Claude Code terminal session is the Pro plan. That is different from Cursor and GitHub Copilot, which both offer a genuinely free tier of their own agent.

### What is the difference between Max 5x and Max 20x?

Both include the exact same Claude Code features as Pro; the only difference is the size of the usage allowance. Max 5x, at $100/mo, gives five times Pro's per-session limit, and Max 20x, at $200/mo, gives twenty times that limit. Anthropic points the smaller tier at people who use Claude across varied tasks most days, and the larger one at daily users running long agent sessions.

### How much does the Claude Code API cost?

Claude Sonnet 5, the model Anthropic points developers toward for most coding work, is $2/MTok for input and $10/MTok for output on the pricing docs. There is no monthly cap on the API. Anthropic's own cost-management guidance puts the average enterprise developer around $13 on an active day and $150 to $250 a month, with 90% of users staying under $30 on their heaviest days.

### How much is Claude Code for a team of 10?

Ten Team Standard seats are $250/mo billed monthly, cheaper than ten Cursor Teams Standard seats at $400/mo and close to ten GitHub Copilot Business seats at $190/mo. Paying annually brings the Claude figure down to $2,400 for the year. Each seat carries its own allowance rather than a pooled team budget, so a mix of seat tiers fits a team where usage is uneven better than putting everyone on the same one.

### Is Claude Code cheaper than Cursor?

At the entry tier, no, both individual plans land on the same monthly sticker. One step up, Claude Code's smaller Max tier costs more than Cursor's matching step, Pro+ at $60/mo, though both top out at the same price on their heaviest individual tier. On teams, Claude's Team Standard seat undercuts Cursor's Teams seat. Claude Code's price also buys chat and Claude Code from one pool, while Cursor's price is for the editor alone.

Cite this: Devshot, "Claude Code Pricing in 2026: Plans, Usage Limits and API Costs", September 2026.

Devshot is the daily brief on the tools and bills developers have to act on. Get it at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cursor vs GitHub Copilot in 2026: Which AI Coding Tool Wins?

URL: https://devshot.co/blog/cursor-vs-github-copilot
Type: blog
Published: 2026-07-21
Updated: 2026-09-25
Summary: Cursor vs GitHub Copilot in 2026, from a developer's chair: the AI-native IDE versus the plugin. Autocomplete, agent mode, models, context, and price.

Guide

## Cursor vs GitHub Copilot in 2026: Which AI Coding Tool Wins?

Cursor vs GitHub Copilot in 2026, from a developer's chair: the AI-native IDE versus the plugin. Autocomplete, agent mode, models, context, and price.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated September 25, 2026 · 11 min read

Here is the contrarian part up front: these two tools are not really competing for the same slot on your machine. [GitHub Copilot](https://github.com/features/copilot) is a plugin that makes the editor you already use smarter. [Cursor](https://cursor.com) is a whole editor that assumes the AI is driving. Most "which is better" arguments miss because they compare a feature list instead of that fork in the road.

So the honest verdict has two answers. If you want to keep VS Code or a JetBrains IDE exactly as it is and bolt on strong completion, chat, and a GitHub-native agent, Copilot wins on fit and on price. If you are willing to switch editors so the AI can restructure files, run multi-step tasks, and hold your whole repo in context by default, Cursor wins on depth. Everything below is the round-by-round that gets you to your own answer, checked against current pricing and features rather than last year's takes.

## Quick verdict

Dimension | Cursor | GitHub Copilot |

What it is | A standalone editor (a fork of VS Code) built around AI | A plugin layered onto your existing IDE |

Best for | Developers who want the AI to lead multi-file work | Developers who want their current setup, unchanged |

Inline autocomplete | Its own fast Tab model, strong multi-line edits | Mature inline completion across many languages |

Agent mode | Deep, multi-file, runs commands, edits across the repo | Agent mode in-IDE plus a cloud coding agent on GitHub |

Codebase context | Indexes and retrieves the whole repo by default | Context you attach plus repo awareness on github.com |

Model choice | Own Composer model plus frontier models via a picker | Model picker across Anthropic, OpenAI, Google |

Editor | You move into Cursor | Stay in VS Code, JetBrains, Neovim, and more |

Paid entry price | $20/month (Pro) | $10/month (Pro) |

Free tier | Yes, limited agent and Tab usage | Yes, capped completions and chat |

The whole choice reduces to one question about your editor, not a feature spreadsheet.

## Round by round

### Autocomplete

Copilot invented the category of ghost-text completion and it is still excellent. It predicts the next line or block as you type, handles boilerplate across dozens of languages, and rarely gets in your way. For most developers this is 80 percent of the daily value, and Copilot delivers it in the editor you already trust.

Cursor's answer is its Tab model, which is tuned less for single-line completion and more for predicting your next edit, including edits that jump around a file or apply a diff across several lines. When you rename a variable or change a function signature, Tab often proposes the follow-up changes before you reach for them. It feels less like autocomplete and more like the editor finishing your refactor.

Winner: close, with an edge to Cursor for multi-line and refactor-style edits, and an edge to Copilot for breadth and staying inside your current IDE.

### Chat and agent mode

Both tools have moved well past chat into agents that plan, edit multiple files, and run commands. The difference is where the agent lives.

Cursor's agent is built into the editor and treats your project as its workspace. It can read across files, make coordinated edits, run terminal commands, and iterate on failures without you copy-pasting anything. Because the editor is the product, the agent has fewer walls between it and your code, which is why Cursor feels stronger on large, multi-file changes.

Copilot now offers agent mode inside the IDE and, separately, a [cloud coding agent](https://docs.github.com/en/copilot) you assign work to from github.com or an issue. You hand it a task, it opens a pull request, and you review the diff like any other contribution. That is a genuinely different workflow: async, GitHub-native, and easy to fold into an existing review process. Copilot also brings automated pull request review and Copilot Spaces for shared context.

Winner: Cursor for hands-on, in-editor multi-file agent work; Copilot for async, PR-shaped tasks that live in your GitHub workflow.

### Codebase context

This is where the "IDE vs plugin" split shows most. Cursor [indexes your repository](https://cursor.com/docs) and retrieves relevant files automatically, so the agent starts most tasks already knowing your project's shape. You can also point it at docs and specific files, but the default is whole-repo awareness.

Copilot leans on the context you attach plus repository awareness, and its strongest whole-repo understanding shows up on github.com rather than always in the editor. It is very good, but you tend to be more deliberate about what you feed it.

Winner: Cursor, for out-of-the-box codebase retrieval, especially on larger projects.

Devshot sends one email a day on the AI tools that actually change how you ship code. If you want the useful ones without the hype, subscribe here.

### Model choice

Neither tool locks you to one model, and both roster change often, so check the current model list before you commit. As of mid-2026 the shape looks like this.

Model access | Cursor | GitHub Copilot |

Own model | Composer (Cursor's agent model) plus its Tab model | None; routes to partner models |

Anthropic Claude | Yes, up to Claude Opus 5.5 and Fable 5.1 | Yes, including Claude Opus 5.5 and Sonnet 5 |

OpenAI GPT | Yes, up to the GPT-5.6 models | Yes, including GPT-6 Sol |

Google Gemini | Yes | Yes |

xAI Grok | Yes, up to Grok 4.7 | Yes, Grok 4.7 |

Premium gating | Higher usage on Pro and higher-usage tiers | Premium models gated to Pro+, Max, and org plans |

The practical takeaway: Copilot gives you a clean model picker with premium models behind higher tiers, which suits teams that want governance. Cursor adds its own Composer model, which is tuned for its agent and is often fast and cheap for routine edits, and lets you escalate to frontier models when a task needs the horsepower. If you care about model behavior specifically, our ChatGPT vs Claude for coding breakdown covers how the underlying models differ on real programming work.

### IDE and workflow

Copilot's biggest advantage is that it meets you where you are. It runs in VS Code, JetBrains IDEs, Neovim, Visual Studio, and the CLI, and it plugs into the GitHub flow you already use for issues, pull requests, and reviews. Nothing about your setup changes; the assistant just appears inside it.

Cursor asks you to move in. It is a fork of VS Code, so your extensions, keybindings, and themes mostly carry over and the learning curve is short, but it is still a new application to adopt. In exchange, everything is wired for AI: the agent, the context indexing, and the Tab model are first-class rather than bolted on. Teams standardized on JetBrains or on a heavily customized editor will feel this cost most.

Winner: Copilot for fitting into existing workflows; Cursor for a workflow rebuilt around the agent.

### Price

Copilot is cheaper at the entry point and priced for organizations. There is a free tier with capped completions and chat, [Pro at $10/month](https://github.com/features/copilot/plans), Pro+ at $39/month for premium models and a larger credit allowance, and a Max tier around $100/month for heavy individual use. For teams, Business is about $19 per seat/month and Enterprise about $39 per seat/month. For the same per-seat question across the two Git hosts, with seat totals at 5, 25 and 100 developers, see GitHub vs GitLab 2026.

Cursor has a free Hobby tier with limited agent and Tab usage, Pro at $20/month, and Teams at about $40 per user/month, with an Enterprise tier on custom pricing. Higher-usage individual tiers exist for people who run the agent constantly, so [check current pricing](https://cursor.com/pricing) if you are a power user. Both tools meter heavy frontier-model usage, so a very active agent user on either platform can exceed the base allowance. Cursor is also compared with Claude Code head to head, with plan prices and a 10-seat team cost, in Claude Code vs Cursor 2026.

Winner: Copilot on raw price, especially the $10 Pro tier and per-seat team plans.

Filled dot marks the tool that leads on each task. The split tracks the IDE-versus-plugin design, not a quality gap.

## Who should pick what

Pick Cursor if the AI is central to how you want to work. If you spend real time on multi-file refactors, want an agent that holds your whole repo in context, and do not mind adopting a new editor to get it, Cursor is the stronger tool. Indie hackers shipping fast, developers on greenfield projects, and anyone who wants to lean hard into agent-driven coding will get the most from it. The $20 Pro tier is the honest test drive.

Pick GitHub Copilot if your workflow is already good and you want to make it sharper. If your team lives in JetBrains, you have a customized editor you will not abandon, or you want the assistant wired into GitHub issues, pull requests, and reviews, Copilot fits without friction. The $10 Pro tier and per-seat team plans also make it the easier sell to a budget owner, and the cloud coding agent gives you async, PR-shaped work that Cursor does not match.

And the unsatisfying-but-true option: run both for a month. Copilot in your daily IDE for completion and quick chat, Cursor open for the gnarly multi-file jobs. They cost little together and they teach you which model of AI coding actually matches your brain. If you are still mapping the wider field, our guide to the best AI coding assistants and the primer on AI-assisted coding put both tools in context.

## FAQ

### Is Cursor better than GitHub Copilot in 2026?

Neither is strictly better; they solve the problem differently. Cursor is the stronger tool for deep, multi-file, agent-led work because the AI is built into the editor. Copilot is the stronger fit if you want to keep your current IDE and workflow, and it costs less to start. Match the tool to how you work rather than chasing a single winner.

### Can I use Cursor and GitHub Copilot together?

Yes, and many developers do. You can keep Copilot in your everyday IDE for completion and quick chat, and open Cursor for larger refactors or agent runs. The two do not conflict since they run as separate applications, though you are paying for both if you go past the free tiers.

### Does Cursor use the same models as Copilot?

They overlap. Both route to frontier models from Anthropic, OpenAI, and Google, so you can run Claude Opus 5.5 or an OpenAI model in either, per Cursor's [model list](https://cursor.com/docs/models) and GitHub's [supported models](https://docs.github.com/en/copilot/reference/ai-models/supported-models). The difference is that Cursor also ships its own Composer and Tab models tuned for its agent, while Copilot gates its premium models behind higher tiers. Model rosters change often, so check each tool's current list.

### Is GitHub Copilot cheaper than Cursor?

At the entry point, yes. Copilot Pro is $10/month against Cursor Pro at $20/month, and Copilot's per-seat team plans start around $19. Both meter heavy frontier-model usage, so a very active agent user can run up costs on either platform beyond the base price.

### Does Cursor work in JetBrains, or only VS Code?

Cursor is its own editor, a fork of VS Code, so you work inside Cursor rather than inside JetBrains. If your team is committed to IntelliJ, PyCharm, or another JetBrains IDE, Copilot is the better fit because it runs as a plugin inside those editors. This is often the deciding factor for JetBrains shops.

### What is agent mode, and how does it differ between the two?

Agent mode lets the AI plan a task, edit multiple files, and run commands instead of just suggesting one snippet. Cursor's agent runs in the editor with full project context and is strong on hands-on, multi-file changes. Copilot offers agent mode in the IDE plus a separate cloud coding agent that you assign work to from GitHub, which opens a pull request you review like any other contribution.

### Which is better for large monorepos?

Cursor has the edge for big codebases because it indexes and retrieves across the whole repository by default, so the agent starts with more awareness of your project. Copilot handles large repos too, but you are usually more deliberate about attaching context in the editor, with its deepest repo understanding surfacing on github.com. Test both on your actual monorepo before deciding, since indexing behavior varies with repo size.

### Is my code used to train the models on either tool?

Both offer privacy modes and business or enterprise plans where your code is not retained or used for training. Cursor has a privacy mode you can enforce team-wide, and Copilot's Business and Enterprise plans exclude your content from training by default. Read the current data-handling terms for the specific plan you buy, since defaults differ between free and paid tiers. For more on working AI into a team codebase safely, see our guide to AI for developers.

Some offers on this page may be paid placements or contain affiliate links.

---

# How to Use Claude Code: A Working Setup, Not a Feature Tour

URL: https://devshot.co/blog/how-to-use-claude-code
Type: blog
Published: 2026-08-29
Updated: 2026-08-29
Summary: How to use Claude Code productively: the CLAUDE.md file that decides output quality, plan mode, permissions, and the review discipline that keeps agent-written…

Guide

## How to Use Claude Code: A Working Setup, Not a Feature Tour

How to use Claude Code productively: the CLAUDE.md file that decides output quality, plan mode, permissions, and the review discipline that keeps agent-written code maintainable.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated August 29, 2026 · 6 min read

The first hour with an agentic coding tool goes one of two ways. Either it edits four files you did not expect and you turn it off, or it does something genuinely useful and you spend the next week wondering why the results are so inconsistent.

Both outcomes have the same cause: the agent is working without the context a new colleague would get on day one. It cannot see your conventions, your deploy path, or the three things that will break production. This guide is about supplying that, because it is the difference between a novelty and a tool you keep.

## Start with the project file

Claude Code [reads a CLAUDE.md at the root of your repository on every session](https://code.claude.com/docs/en/memory). This single file does more for output quality than any prompting technique.

Write it as onboarding notes for a competent developer who has never seen the codebase. Short, specific, and about things that are not inferable from the code.

Include | Skip |

Build, test and deploy commands | Anything the agent can read from package.json |

Conventions that are not obvious from the code | Restating your framework's documented defaults |

Architecture decisions and why | A file-by-file directory listing |

Things that will break production if touched | Aspirational style rules nobody follows |

Concretely: "run npm run check before committing, it runs types and lint together", "never edit dist/, it is generated", "the payments module is legacy, changes need a review from the finance team". That kind of thing.

The test of a good CLAUDE.md is whether a competent stranger could ship a small change without asking anyone. If they could, the agent can too.

## Plan before it edits

For anything beyond a one-line fix, ask for the approach before the code. Reviewing a plan takes a minute; reviewing a wrong implementation across six files takes an afternoon and you will miss something.

The pattern that works: describe the outcome, ask for a plan, correct the plan, then let it run. Most bad sessions are a bad plan that nobody looked at.

## Permissions, deliberately

The agent [asks before actions that change things](https://code.claude.com/docs/en/permissions). The temptation is to approve everything to stop the interruptions, which removes the only checkpoint you have.

A workable middle ground:

Action | Setting |

Reading files, searching, running tests | Allow freely |

Editing files in the repo | Allow, because you review the diff |

git commit | Allow once you trust the message quality |

git push, deploys, migrations, anything destructive | Always ask |

The distinction that matters is reversibility. A bad edit is a git checkout. A bad push or a dropped table is a much longer afternoon.

## Read the diff

This is the discipline that separates useful output from accumulating debt. The code will compile. The tests may pass, particularly if the agent also wrote them. Neither tells you the logic is right.

Read for the things that look fine and are not: an error swallowed in a catch, a condition inverted, a retry that hides a failure it should surface, a test asserting the behaviour that was implemented rather than the behaviour you wanted.

If the diff is too large to read carefully, that is the signal the task was too large. Ask for it in pieces.

## Tools

Claude Code is a terminal tool, so it composes with what you already run: git, your test runner, your linter. Point it at your existing npm run check rather than inventing a new workflow.

For where it sits against the wider category, see our comparisons of the best AI coding assistants, AI for code review and vibe coding tools if you want the prompt-to-app end of the market instead.

## Pitfalls

Letting one session run for hours. Context degrades and so does output. Finish a task, commit, start fresh.

Asking for a feature instead of a change. "Add authentication" produces something unusable. "Add a login route that checks the session cookie against the existing users table, matching how the admin route already does it" produces something reviewable.

Never updating CLAUDE.md. If you correct the same thing twice, that correction belongs in the file. Otherwise you will correct it forever.

Trusting tests the agent wrote without reading them. A test that asserts current behaviour passes and proves nothing.

## What these tools actually cost

We price every tool we review, so this is measured rather than estimated. Across 429 tools, 293 publish a price and 33% offer a free tier. Among developer tools, the median entry plan is $24.50 a month, which runs above the $24 median across every category we price.

The spread matters more than the median. Half of the developer tools sit between $10 and $49, and the range runs from $2.49 to $299. A quoted "starting at" price near the bottom of that range usually means per-seat add-ons land on top of it. Once your setup works, OpenCode vs Claude Code 2026 shows what the same agent work costs on a metered model.

Price point | Developer tools | All tools |

Cheapest paid plan | $2.49 | $1 |

Lower quartile | $10 | $10 |

Median | $24.50 | $24 |

Upper quartile | $49 | $49 |

Most expensive | $299 | $990 |

Tools measured | 18 | 293 |

Median advertised entry price/mo. Source: Dupple pricing index, 293 tools with public pricing out of 429 reviewed, 2026-08-19.

## FAQ

### What is Claude Code?

It is Anthropic's [agentic coding tool that runs in your terminal](https://code.claude.com/docs/en/overview) and works directly on your repository: reading files, editing them, running commands and tests, and committing. The difference from a chat window is that it acts on the real codebase rather than returning a snippet you paste, which is what makes both the speed and the review discipline matter.

### How is it different from an editor extension?

Editor assistants are strongest at completion and local edits inside the file you are looking at. A terminal agent works across the whole repository and can run your toolchain, so it suits multi-file changes, refactors and anything that needs the tests run. Many developers use both, for genuinely different jobs.

### Do I still need to review the code?

Yes, and it is the whole discipline. Generated code compiles and looks finished, which is precisely why it slips through. Read every line of the diff, particularly error handling and anything touching auth, payments or data deletion. If the diff is too big to read, the task was too big.

### How do I stop it making changes I did not ask for?

Two things. Ask for a plan before implementation on anything non-trivial, and keep the permission prompts on for irreversible actions like pushing, deploying or migrating. Scope also helps more than instructions: one clear change per session, committed before you start the next.

Some offers on this page may be paid placements or contain affiliate links.

---

# 9 Best AI Code Generators in 2026 (From Prompt to Working Code)

URL: https://devshot.co/blog/best-ai-code-generators
Type: blog
Published: 2026-07-21
Updated: 2026-07-21
Summary: The best AI code generators in 2026: prompt-to-app builders like v0, Bolt and Lovable, plus in-IDE agents like Cursor and Claude Code.

Guide

## 9 Best AI Code Generators in 2026 (From Prompt to Working Code)

The best AI code generators in 2026: prompt-to-app builders like v0, Bolt and Lovable, plus in-IDE agents like Cursor and Claude Code. Verified pricing, real limits.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated July 21, 2026 · 14 min read

Key facts9 tools compared: v0 (Vercel), Bolt.new, Lovable, Replit Agent, Base44 (Wix), Cursor, Claude…

- Updated: July 21, 2026
- Prices as of: July 2026

- 9 tools compared: v0 (Vercel), Bolt.new, Lovable, Replit Agent, Base44 (Wix), Cursor, Claude Code, OpenAI Codex, GitHub Copilot
- v0 (Vercel) (best for: React and Next.js UIs, full-stack, deploy to Vercel): Free tier; Plus $30/user/mo
- Bolt.new (best for: In-browser full-stack apps with hosting): Free tier; Pro $25/mo
- Lovable (best for: Full-stack web apps on a Supabase backend): Free tier; paid from ~$25/mo (check current)
- Replit Agent (best for: Prompt to deploy with database and hosting in one place): Free tier; Core $20/mo

A year ago, an "AI code generator" mostly meant autocomplete that finished the line you were already typing. In 2026 the phrase covers something bigger: you write a sentence and get a running application, with a database and login attached, deployed to a live URL. That shift is the whole story of the past year.

Two things changed at once. The prompt-to-app builders, [v0](https://v0.app), [Bolt](https://bolt.new), [Lovable](https://lovable.dev), [Replit Agent](https://replit.com), [Base44](https://base44.com), went from demo toys to tools that ship real full-stack apps with a backend wired in, not just a static screen. And the in-IDE agents, [Cursor](https://cursor.com), [Claude Code](https://claude.com/product/claude-code), [Codex](https://openai.com/codex/), [Copilot](https://github.com/features/copilot), stopped suggesting single lines and started writing whole features across files, running the build, reading the error, and fixing themselves in a loop. Both now generate code from a prompt. They just generate it in very different places.

It helps to separate a generator from an assistant. A coding assistant rides along while you type and suggests the next few lines. A code generator produces something that did not exist before: a new component, a new endpoint, or an entire app from a description. This guide is about the second job, prompt to working code, and the split that decides which tool fits: app builders that hand you a deployed app, versus in-IDE agents that generate inside a repo you already own.

## Quick comparison

Prices below were checked against each vendor's pricing page in July 2026. Credit and token allowances move often, so confirm the current number before you buy.

Tool | Type | Best for | Paid entry price |

v0 (Vercel) | App builder | React and Next.js UIs, full-stack, deploy to Vercel | Free tier; Plus $30/user/mo |

Bolt.new | App builder | In-browser full-stack apps with hosting | Free tier; Pro $25/mo |

Lovable | App builder | Full-stack web apps on a Supabase backend | Free tier; paid from ~$25/mo (check current) |

Replit Agent | App builder | Prompt to deploy with database and hosting in one place | Free tier; Core $20/mo |

Base44 (Wix) | App builder | Full-stack apps with built-in database, auth, payments | Free tier; Starter $16/mo |

Cursor | In-IDE agent | AI-first editor with agent mode across the repo | Free tier; Pro $20/mo |

Claude Code | Terminal agent | Multi-file changes and test loops from the CLI | Bundled with Claude Pro ($17 to $20/mo) |

OpenAI Codex | CLI + IDE agent | Agentic generation tied to a ChatGPT plan | Included with ChatGPT Plus ($20/mo) |

GitHub Copilot | In-IDE agent | Completions plus agent mode across editors | Free tier; Pro $10/mo |

## App builders: prompt to a running app

These five turn a description into a deployed app. Frontend, and in most cases a real backend, come out the other end.

### 1. v0 by Vercel

v0 is the strongest tool if your output is React and Next.js. You describe a screen or a flow, and it generates clean component code with Tailwind, wires up state, and lets you refine in a visual Design Mode before you push. Because it comes from Vercel, deployment is one click and GitHub sync is built in, so the generated project is a normal repo you keep working in.

Pricing starts free with $5 in monthly credits and a 7-message daily cap. The Plus plan is $30 per user per month with $30 in credits, and Business is $100 per user per month with training opt-out on by default. It uses a credit model tied to four underlying model tiers, so heavy generation burns credits faster. v0 is at its best on UI and full-stack Next.js work and weakest if your stack is not JavaScript.

### 2. Bolt.new

Bolt runs an entire dev environment in the browser using [WebContainers](https://webcontainers.io), which means it can install packages, run a real Node server, and show you a live preview without anything on your machine. You prompt a full-stack app, watch it build, edit files directly, and deploy with hosting and a custom domain from the same tab. It supports a wide range of JavaScript frameworks rather than locking you to one.

The free tier gives 300K daily tokens and 1M monthly. Pro is $25 per month for 10M monthly tokens with no daily limit, Teams is $30 per member per month, and Enterprise is custom. Tokens are the currency here, and a complex app can eat a month's allowance quickly, so scope your prompts. Bolt shines for spinning up a working prototype in one sitting.

### 3. Lovable

Lovable markets itself as an "AI software engineer" for non-developers and developers alike. You chat, it builds a full-stack web app with a React frontend, and it provisions a [Supabase](https://supabase.com) backend for the database, auth, and storage. The pitch is that someone with no stack knowledge can ship a real app, and for CRUD-shaped products that mostly holds.

The free plan grants 5 build credits a day, up to 30 a month. Paid Pro and Business plans add a monthly credit balance and cheaper top-ups ($0.30 per credit on Pro, $0.60 on Business), with the base monthly price in the low-to-mid $20s range at last check. Confirm the current number on the pricing page, since Lovable has changed its credit model more than once. The Supabase dependency is worth knowing up front: your data and auth live there, which is convenient but is a decision made for you.

### 4. Replit Agent

Replit Agent is the most end-to-end of the group because Replit already is a cloud IDE, database, and host. You describe an app, the agent scaffolds it, writes the code, sets up the database, and can deploy it, all inside one environment with no local setup. It works through checkpoints you approve as it goes, and newer plans run several agents in parallel.

The Starter tier is free with daily agent credits and one published project. Core is $20 per month (billed annually) with $25 in monthly credits and up to two parallel agents. Pro is $95 per month with $100 in credits, up to ten parallel agents, access to the strongest models, and 28-day database rollbacks. If you want prompt to deployed URL without touching your own machine, this is the shortest route.

### 5. Base44

Base44, now owned by Wix, generates the full stack from a plain-language description and includes the parts the others often bolt on separately: database, user authentication, payments, and integrations with tools like Gmail, Slack, and Salesforce. You own the code and data, and a two-way GitHub sync lets you export the source to your own repo, which softens the usual lock-in worry with app builders.

Pricing on annual billing runs Free (25 message credits), Starter $16 per month (100 credits), Builder $40 (250), Pro $80 (500), and Elite $160 (1,200). Integration credits scale with each tier. Base44 is a good pick when you want auth and payments handled from the first prompt rather than added later.

## In-IDE agents: generate inside your codebase

These four generate code where the code already lives. They read your repo, write across files, and run commands, which makes them the right tool once a project exists rather than when you are starting from zero.

### 6. Cursor

Cursor is a fork of [VS Code](https://code.visualstudio.com) built around AI. Its agent mode can take a task, plan a change, edit multiple files, run the terminal, and iterate, while its Tab feature predicts and applies multi-line edits as you work. Because it is a full editor, generated code lands in context with your types and conventions, not in a separate chat window you copy from.

Hobby is free with limited agent requests and completions. Pro is $20 per month, with higher Pro+ and Ultra tiers for heavy users, and Teams is $40 per user per month with SSO and admin controls. Every plan includes a set amount of model usage with overage billed on demand. Cursor is the common choice for developers who want an agent that generates and an editor that autocompletes in one place. For a direct face-off with the incumbent, see Cursor vs GitHub Copilot.

### 7. Claude Code

Claude Code is a terminal agent. You give it a task in your project directory, and it reads files, makes multi-file changes, runs your tests, reads the output, and keeps going until the task is done or it needs you. It has no editor of its own, which is the point: it drops into whatever workflow you already have and generates against your real codebase.

It is bundled into Claude subscriptions rather than sold on its own. Claude Pro is $17 per month on annual billing (or $20 monthly) and includes Claude Code, and Claude Max starts at $100 per month with 5x or 20x the usage for people who run it all day. It is strong on larger, multi-step changes and test-driven loops. The tradeoff is that terminal-only, agentic generation rewards developers who can review a diff, and punishes those who cannot.

### 8. OpenAI Codex

Codex is OpenAI's agentic generator, available as a CLI and an IDE extension, and tied to a ChatGPT plan rather than priced separately. You point it at a task and it edits files, runs commands, and can work on longer jobs in the background or in the cloud. It generates against your repo the way the other agents do, with the ChatGPT ecosystem behind it.

Access comes with paid ChatGPT plans, starting at Plus ($20 per month), with higher usage on Pro and additional API-based usage available. Because the exact Codex allowances shift with ChatGPT's plan structure, check the current terms before you lean on it for heavy work. It is a natural fit if you already pay for ChatGPT and want an agent that generates without a second subscription.

### 9. GitHub Copilot

Copilot is the most widely installed of the group and now does far more than completions. It offers unlimited code completions on paid plans plus an agent mode that plans and generates across files, and it runs inside VS Code, Visual Studio, JetBrains, Eclipse, and Xcode. It can also route work to third-party agents like Claude Code and Codex from within GitHub.

The free plan includes 2,000 completions a month. Pro is $10 per month with $15 in monthly credits, Pro+ is $39 with premium models including Opus and roughly 4x the usage, and there are Business and Enterprise tiers for organizations. If your team lives in GitHub and you want generation with the least friction, Copilot is the default. For the wider set of day-to-day tools, our AI coding assistants roundup goes deeper.

Devshot sends one email a day on the AI tools and workflows developers actually use in production. Read it here.

## Which type do you need

The mistake is picking a tool before deciding which job you are doing. App builders are unbeatable for going from nothing to a live app fast, especially for a prototype, an internal tool, or a founder validating an idea. In-IDE agents are the right call when a real codebase already exists and you need generated changes to respect its structure, tests, and conventions.

If you want to... | Reach for | Why |

Ship a prototype or MVP with a backend, no local setup | Replit Agent, Bolt, Base44 | They scaffold, generate, and deploy in one environment |

Generate polished React and Next.js UI you keep coding on | v0 | Clean component output plus one-click Vercel deploy |

Build a full app as a non-developer | Lovable, Base44 | Chat-driven, with database and auth provisioned for you |

Add features to an existing repo | Cursor, Claude Code, Codex | They read and edit your code in place |

Get generation with the least change to your setup | GitHub Copilot | Works inside the editor you already use |

A common real-world pattern is to use both families in sequence: an app builder to stand the project up, then export the code and continue in an in-IDE agent once the app is real enough that structure matters. Base44 and v0 both make that export clean. Whichever you start with, the disciplined workflow of scoping tightly and reviewing every generated line is what keeps the speed from turning into a mess.

## Limits and gotchas

Generated code that runs is not the same as generated code you can maintain, and the gap is where teams get hurt.

Hallucinated dependencies and APIs. Every one of these tools will sometimes import a package that does not exist, call a method with the wrong arguments, or pin a version that was never released. It happens more with fast-moving libraries the model saw little of. Read the imports, and do not assume a function exists because the code that calls it looks confident. If a build fails on a missing module, that is often the tell.

Security defaults are weak. Ask for "a login" or "a form that saves to the database" and you tend to get the working version, not the safe one: [string-built SQL](https://owasp.org/www-community/attacks/SQL_Injection) instead of parameterized queries, a missing authorization check, secrets in plaintext, an open storage bucket. App builders that provision auth and a backend for you can hide these decisions entirely. Treat anything touching auth, payments, or user data as needing a real review, not a glance.

Maintainability degrades as the app grows. The first prompt produces something clean. The twentieth prompt, layered onto code nobody read, produces a tangle: duplicated logic, inconsistent patterns, components that repeat instead of reuse. Prompt-to-app tools are fastest at the start and slowest to untangle later. The defense is to export early, review what was generated, and refactor before the codebase gets too big to hold in your head.

Lock-in and cost creep. Credit and token models make heavy generation quietly expensive, and a full app on a builder's hosting is not always trivial to move. Check the export path before you commit real work, and keep an eye on the meter. If you plan to run Python-heavy backends, our best AI for Python coding guide covers which tools handle that stack well.

## FAQ

### What is an AI code generator?

It is a tool that produces new code from a natural-language description, rather than just completing what you already typed. That spans two families in 2026: app builders like v0, Bolt, and Lovable that turn a prompt into a full running application, and in-IDE agents like Cursor, Claude Code, and Codex that generate code across files inside an existing repository. The common thread is that you specify intent in plain language and get working code back.

### What is the difference between an AI code generator and a coding assistant?

A coding assistant helps while you type, suggesting the next few lines in context. A code generator creates something that did not exist yet, a new component, endpoint, or whole app, from a description. In practice the line blurs, since tools like Cursor and Copilot do both, but the useful distinction is scope: assistants speed up code you are already writing, generators produce code you have not written.

### Can AI actually build a full app from a prompt?

Yes, for a real class of apps: CRUD tools, dashboards, prototypes, and standard web apps with auth and a database. Replit Agent, Lovable, Bolt, and Base44 do this today, backend included. Where it breaks down is anything with unusual business logic, tight performance needs, or heavy existing systems to integrate with. The generated app is a strong starting point, not a finished product you skip reviewing.

### Which AI code generator is best for a non-developer?

Lovable and Base44 are built for people who cannot write the code themselves. Both are chat-driven and provision the backend, database, and auth for you, so you describe the app and get a working version. Replit Agent is also approachable and gives you a deploy button in the same place. The catch for all three is that you still own decisions about data and security that you may not know to ask about, so bring in a developer before anything handles real users or payments.

### Do I own the code these tools generate?

Usually yes, but check each vendor's terms. v0, Bolt, Base44, and the in-IDE agents let you export or sync to a real Git repository that is yours to keep. The thing to verify is not just ownership of the code but the portability of the whole app: hosting, database, and auth can be tied to the platform even when the source is exportable. Confirm the export path before you build something you cannot afford to lose.

### Is AI-generated code safe to ship to production?

Not without review. Generated code tends to default to the simplest working version, which often skips input validation, authorization checks, and safe data handling. It can also carry hallucinated dependencies or subtle logic errors that pass a quick read. Review every generated line the way you would a pull request from a stranger, run the code against real edge cases, and give anything touching auth, payments, or user data a proper security pass before it ships.

### App builder or in-IDE agent, which should I choose?

Start with the job. If you are going from nothing to a live app and want speed, pick an app builder like Bolt, Replit Agent, or v0. If a codebase already exists and you need generated changes to fit its structure and tests, pick an in-IDE agent like Cursor, Claude Code, or Codex. Many developers use both in sequence: build with one, export, then maintain with the other.

Some offers on this page may be paid placements or contain affiliate links.

---

# Will Coding Be Replaced by AI? What Abstraction Did to the Job Last Time

URL: https://devshot.co/blog/will-coding-be-replaced-by-ai
Type: blog
Published: 2026-07-21
Updated: 2026-07-21
Summary: Compilers, garbage collection and frameworks each deleted huge amounts of coding without deleting coders. Where AI is different, what it still cannot…

Guide

## Will Coding Be Replaced by AI? What Abstraction Did to the Job Last Time

Compilers, garbage collection and frameworks each deleted huge amounts of coding without deleting coders. Where AI is different, what it still cannot do, and what it means for juniors.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated July 21, 2026 · 17 min read

The job already survived the compiler, which retired hand-written assembly for almost everyone; garbage collection, which deleted an entire family of memory bugs along with the code that managed them; managed runtimes, which took away platform-specific plumbing; and the frameworks that now generate the routing, the ORM layer, the auth scaffolding, and most of the boilerplate in a typical application before anyone writes a line of business logic. Each of those removed an enormous share of the typing, and none of them reduced the number of people employed to build software. The argument that this time is different is that those abstractions removed mechanical work while AI removes the reasoning, and it is worth taking seriously rather than waving away. Having taken it seriously: no, coding is not being replaced, but the activity people used to call "writing code" is being pulled apart into a part that is becoming cheap and a part that is becoming the whole job.

The single most load-bearing number here is not from a vendor. The US Bureau of Labor Statistics projects software developer employment to grow 15 percent from 2024 to 2034, much faster than the average occupation, and that projection was made with AI coding tools already in wide use ([BLS Occupational Outlook Handbook](https://www.bls.gov/ooh/computer-and-information-technology/software-developers-quality-assurance-analysts-and-testers.htm)). If the technology were quietly deleting the profession, the people whose entire function is counting jobs would be the first to notice. So the useful question is not whether programmers survive. It is which hours of your week are about to move, and what you are left holding.

## What actually got automated, and what that historically did to the job

Look at what each previous abstraction actually did, because the pattern is consistent and it is not the pattern people expect.

The compiler did not make programmers redundant; it made a programmer's output larger and shifted the scarce skill from register allocation to algorithm and structure. Garbage collection did not remove the need to understand memory; it removed a category of routine, repetitive, error-prone work and left the harder version, reasoning about lifetimes, leaks, and pressure, to the people who needed it. Frameworks did not eliminate web developers, despite generating far more of a typical application than most people admit. They made building an application cheap enough that vastly more organizations built one, and the work moved up into integration, product, and the parts a framework has no opinion about.

There is a consistent mechanism underneath. When the cost of producing software falls, the amount of software people want does not stay fixed. It rises, usually by more than the cost fell, because there was always a queue of things nobody could justify building. What changed each time was the composition of the day: less mechanical production, more design, more integration, more of the work that decides whether the thing was worth building at all.

The case that AI breaks the pattern rests on one claim: previous abstractions raised the level you worked at while still requiring you to specify behavior precisely, and a model appears to accept an imprecise specification and produce working software from it. If that were fully true, the abstraction would consume the specifying skill itself, which is the actual job, rather than the typing.

It is partly true, and that is the honest version. Models do accept vague input and produce plausible output. What they cannot do is make a vague specification correct, because the ambiguity is not in the phrasing, it is in the world: nobody decided what happens to the half-finished order, or which of the two conflicting stakeholder requests wins, or whether this edge case is worth the complexity. A model resolves those silently and confidently, and the resolution is a guess. Someone still has to notice that a decision was made, and decide whether it was the right one. That is the specifying skill, and it did not go anywhere. It just stopped being expressed in syntax.

Coding work | What AI does | Who owns the outcome |

Autocomplete and boilerplate | Writes it inline as you type | You accept or reject each line |

A new function or component | Generates a working first draft | You verify it fits the real codebase |

Unit tests | Drafts cases straight from the code | You decide what actually matters to test |

Refactor or rename | Applies the change across files | You confirm behavior did not change |

Debugging a known error | Suggests the likely fixes | You reproduce and confirm the real cause |

Pull request review | Flags smells, bugs, and risks | A human approves the merge |

System design | Sketches a few options | An engineer picks and is accountable |

What to build and why | Almost no help | Product and engineering own it entirely |

## What the models are genuinely good at now

No forecasting in this section. These workflows are running in real repositories today, and they are why the question feels urgent rather than academic.

Autocomplete and boilerplate. Multi-line completion finishes the obvious next few lines correctly most of the time. The glue you used to type from muscle memory, a loop over a response, a null check, a mapper between two shapes, arrives before you finish the thought. This is the least interesting capability and the one that saves the most cumulative minutes.

Whole-function and component generation. Describe a component or a request handler and you get a usable draft in seconds. For well-trodden patterns the draft often runs on the first attempt. The catch is that it runs in isolation, which is not the same as running inside your architecture, with your error handling conventions, against your actual data.

Test drafting. Point a model at a function and it produces a table of cases, including edge cases you would have skipped at the end of a long day. It cannot tell you which behaviors are load-bearing for the product, so the coverage it produces is broad and undirected, but it clears the mechanical part fast.

Refactors. Rename a concept across forty files, extract a chunk into a module, migrate a callsite pattern. Agents handle the tedious spread of a mechanical change well, and a careful diff review catches the misses. This is arguably where agentic tooling has the best ratio of value to risk, because the intended behavior change is zero and any behavior change is therefore a bug you can look for.

Debugging assistance. Paste a stack trace and you get a ranked set of likely causes and a proposed fix. On common, well-documented errors this is a genuine time saver. On anything novel it guesses with the same confidence, which is the failure mode covered below.

Pull request review. Review bots read a diff and flag null-safety gaps, missing error handling, and obvious security holes before a person looks. They are noisy. They also catch the boring mistakes that slip past a tired reviewer at 6pm, which is most of the mistakes that reach production.

Developers noticed. In the [2025 Stack Overflow Developer Survey](https://survey.stackoverflow.co/2025/ai), 84 percent of respondents now use or plan to use AI tools, up from 76 percent the year before, and 51 percent of professional developers use them daily. Adoption is settled. Trust is the interesting variable.

Source: [Stack Overflow 2025 Developer Survey](https://survey.stackoverflow.co/2025/ai). Near-universal use, near-zero blind trust. That gap is exactly where the human job lives.

Only 3.1 percent of developers say they highly trust the accuracy of AI output. Meanwhile 66 percent report their top frustration is "AI solutions that are almost right, but not quite," and 45 percent say debugging AI-generated code takes them more time, not less. That is the signature of a tool that accelerates a competent engineer and quietly sabotages an inattentive one, which is a very different thing from a tool that replaces either.

## Three failure modes that still need an engineer

Not five, not a list of vague human virtues. Three specific things go wrong, repeatedly, and each one has a reason.

One: system design and tradeoffs across a codebase it cannot hold. Choosing between a queue and a cron, deciding where a service boundary goes, recognizing that this "simple" feature will fork your data model in six months. These depend on constraints that are not in any file: team size, on-call tolerance, migration history, what the roadmap wants next quarter, which part of the system nobody understands anymore. Even where a large context window can technically ingest the repository, the relevant information is mostly not in the repository. A model will happily produce an architecture. It has no stake in whether that architecture survives eighteen months of contact with reality, and no memory of the last three times a similar decision went badly here.

Two: correctness under real load and at the edges. This is where models are simultaneously weakest and most confident. A flaky failure that appears only above a certain concurrency, a race between two services, a memory leak that surfaces after nine hours, a query plan that changes when the table crosses a size threshold. Diagnosing these means forming a hypothesis, instrumenting, reading live telemetry, and reasoning about a system state that exists nowhere in any training set. The [DORA 2024 report](https://dora.dev/research/2024/dora-report/) found the tension directly: AI adoption raised individual productivity and job satisfaction, yet it "negatively impacts software delivery stability and throughput," and the fundamentals of small batches and thorough testing still drove outcomes. More code, faster, is not the same thing as a system that stays up.

Three: accountability when it breaks in production. When a deploy takes down checkout, a model does not get paged, does not talk to the customer whose order vanished, does not write the incident review, and does not carry the fix through review at midnight. Someone whose name is on the commit does. This is not a sentimental point about human value, it is a structural one: organizations pay for software that can fail expensively, so they pay for someone who can be held to it. As long as that is true, there is a person at the end of the chain, and that person needs to actually understand the code they are answering for.

Notice that all three failure modes get worse, not better, as more code is generated. A larger volume of code written faster, by something with no memory of why the previous decision was made, increases the design surface, the number of edge cases, and the amount of production behavior somebody has to be accountable for.

Devshot unpacks which coding tasks AI is actually taking over, every weekday morning in five minutes.

## Which parts of the work are most exposed

Development is not one job, and the exposure is uneven enough that a blanket answer is useless. Here is the split.

Dev work or role | AI exposure | Where it heads |

CRUD, boilerplate, glue code | High | Absorbed into autocomplete and agents |

Junior "ticket to code" grind | High | Shifts toward reviewing and specifying |

Mid-level feature work | Medium | Faster output, more review and integration |

Senior and staff engineering | Low | Architecture, tradeoffs, direction, mentoring |

Debugging novel prod issues | Low | Stays human, AI assists on the edges |

Security and infra decisions | Low | Human accountable, AI drafts under scrutiny |

Most exposed: repetitive, self-contained, low-context code. Standard CRUD endpoints, glue between two well-documented APIs, the fifth variation on a form, and the "translate this ticket directly into a diff" work. These are pattern-heavy and require almost nothing that is not visible in the immediate file, which is precisely the shape models handle well.

Least exposed: anything that requires holding the system in your head. Architecture, cross-cutting changes, migrations with data at stake, security and infrastructure decisions, and the unglamorous work of turning "make onboarding better" into a correct, shipped change. Requirements arrive vague, contradictory, and half-wrong, and converting them into something buildable is a conversation with stakeholders rather than a prompt. Demand for this is rising, partly because a codebase carrying a lot of generated code needs more architectural discipline and more review, not less.

Between those two poles, the mid-level day changes character rather than volume. The output goes up. So does the share of the day spent reading, verifying, and integrating rather than producing. In practice a strong engineer now spends less time writing a function and more time specifying it precisely, then reading what came back with the skepticism of a reviewer who assumes a subtle bug is present until proven otherwise. Directing an agent well looks a lot like writing an excellent ticket for a fast, literal, tireless colleague who never learns from yesterday and never asks a clarifying question.

The typing shrinks. The judgment on both ends grows, and it is where the accountability sits.

## The junior hiring question

This is the real anxiety behind the search, and it deserves a straight answer rather than a reassuring one.

The structural problem is genuine and it is not about whether juniors can code. For decades the entry-level path ran through exactly the work that is now most automatable: small, self-contained tickets, well-specified bugs, boilerplate with a senior reviewing the diff. That work was never valuable in itself. It was the training apparatus. It was how people accumulated the pattern library that later lets them look at a diff and feel that something is wrong before they can articulate why. If a large share of that work is absorbed by tooling, the apparatus goes with it, and the shortage arrives years later at the senior end, where it is much more expensive to fix.

What is not established is the claim you see everywhere that AI has already collapsed junior hiring. Entry-level technical hiring moves with interest rates, post-2021 overhiring corrections, offshoring, and the general state of the industry, all of which have been in play at once. Anyone attributing the whole movement to one cause is guessing. The defensible statement is narrower: the specific tasks juniors were historically hired to do are cheaper than they were, so the case for a junior now has to rest on something other than throughput on small tickets.

Which is workable, because that case exists. What follows is what actually makes it, and it is the same list that keeps any engineer valuable, just compressed into a shorter runway.

Get sharp at reviewing code you did not write. The bottleneck moved from producing a first draft to verifying one quickly and catching the "almost right, but not quite" bug before it merges. Treat every generated diff like a pull request from a competent stranger with no context on your system: read the error paths, check the assumptions it made silently, and confirm it did what you asked rather than something adjacent. This used to be a skill people built over years of reviewing colleagues. It is now the daily core of the work, which means it has to be built deliberately and early. The best AI for code review guide covers the tooling, but the judgment is yours to develop.

Learn to debug systems, not snippets. The durable form of debugging is reasoning about live, novel failures with real telemetry, which is precisely what models are worst at. Invest in observability, profiling, and hypothesis-driven investigation. Learn to read a flame graph, a slow query log, and a trace across services. The best AI for debugging tools assist at the edges, but the mental model that finds a race condition is the part that compounds.

Move up a layer into design and ownership, early. Volunteer for the ambiguous problems: the schema decision, the migration plan, the "should we even build this" conversation. If you wait to be handed architectural work when you are senior, the path that used to deliver you there may no longer exist. Ask to sit in the design discussion long before you are expected to lead one.

Direct agents deliberately. Being effective with these tools is a real engineering skill with real technique: scope the task tightly, supply the right context, state the constraints that are not visible in the code, and verify the output before it accumulates. Done well you get the speed without inheriting the bugs. Done badly you get a large volume of plausible code that nobody understands, which is the worst possible position to be in when it breaks. The ChatGPT for coding guide covers the prompting habits that make that loop productive, the best AI coding assistants roundup compares the agents worth directing, and the AI for developers hub maps the tooling across the lifecycle.

For teams, the implication is uncomfortable but simple: if you stop hiring and training juniors because the tooling covers their old output, you are borrowing from your own senior pipeline at a rate you will not enjoy paying back. The teams that handle this well are the ones deliberately putting juniors on review, on incidents, and on design conversations far earlier than tradition suggested, because the slow apprenticeship path is the thing that broke, not the people on it.

## FAQ

### Will coding be replaced by AI?

Not replaced, but restructured. AI is automating the production of code, not the engineering around it. Deciding what to build, designing systems, weighing tradeoffs, diagnosing novel failures, and owning what ships still require developers, and all five of those get harder as the volume of generated code grows. The job is shifting from typing code to specifying, reviewing, and directing it. The BLS still projects 15 percent growth in software developer employment from 2024 to 2034, a projection made with these tools already in wide use.

### Will junior developers still get hired?

Yes, but for different reasons than before. The classic entry path ran through small, self-contained tickets, which is exactly the work that is now cheapest to automate, so throughput on small tasks is no longer a compelling case for a junior hire. What is compelling is someone who reviews generated code critically, learns the system quickly, and grows into design thinking faster than the previous generation had to. Be skeptical of anyone attributing the whole state of entry-level technical hiring to AI, since rates, post-boom corrections, and general industry conditions all moved at the same time.

### Is learning to code still worth it in 2026?

Yes, and the reason has changed. You cannot direct, review, or debug generated code without understanding code, and that understanding is now the core skill rather than the raw typing. What is less worth your time is memorizing syntax you can generate on demand. Spend the effort on reasoning about programs, reading unfamiliar code fast, understanding how systems fail, and designing them, because those are the parts that determine whether you can catch what a model got subtly wrong.

### Which coding tasks should I hand to AI first?

Start where the intended behavior change is small and verification is cheap: boilerplate, unit test drafts, mechanical refactors and renames, and explaining unfamiliar code to yourself. Keep a much tighter hand on anything security-adjacent, infrastructure changes, database migrations, and novel production bugs, because that is where a confident wrong answer is expensive and hard to detect. The rule of thumb is to let the model draft and to make yourself the reviewer who finds what it got almost right.

### Does using AI actually make developers faster?

It depends on the work and on the discipline around it. AI clearly speeds up boilerplate, tests, and first drafts. But the 2024 DORA report found that AI adoption improved individual productivity while it worsened overall software delivery stability and throughput, and 45 percent of Stack Overflow respondents say debugging AI-generated code takes longer, not less. Individual speed and system-level delivery are different measurements, and the tool can improve one while degrading the other. The net is positive when you review rigorously and negative when you trust output blindly.

### What is the one skill that compounds from here?

Critical reading of code you did not write. Every other durable skill in this article routes through it: you cannot verify a generated diff, review a junior's pull request, debug an unfamiliar service, or evaluate an architecture proposal without it. It is also the skill developers historically built slowest, through years of reviewing colleagues, and the one the current tooling demands from day one. If you invest deliberately in a single thing, invest in becoming the person who can look at unfamiliar code and say precisely what is wrong with it.

Some offers on this page may be paid placements or contain affiliate links.

---

# AI for Developers in 2026: The Tools, and the Review Problem They Create

URL: https://devshot.co/blog/ai-for-developers
Type: blog
Published: 2026-07-17
Updated: 2026-07-17
Summary: Coding assistants, PR review bots, debugging and DevOps AI in 2026, and how to keep review from becoming the bottleneck that agent-written code creates.

Guide

## AI for Developers in 2026: The Tools, and the Review Problem They Create

Coding assistants, PR review bots, debugging and DevOps AI in 2026, and how to keep review from becoming the bottleneck that agent-written code creates.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated July 17, 2026 · 7 min read

Writing code stopped being the constraint on most teams, and reviewing it became one. That single shift explains nearly everything about how AI tooling has developed since agents started producing multi-file changes: a diff an agent produced in seconds takes a human exactly as long to read as it always did, and now there are far more of them arriving. Throughput moved downstream. If your team has adopted an agent and things feel slower rather than faster, the bottleneck did not disappear, it relocated to your pull request queue.

That is worth holding onto while you evaluate tools, because most of them are sold on generation speed, which is no longer the scarce resource. The useful questions are narrower. How much of this output can I verify cheaply? Where does the tool fit into review and CI rather than around it? And which parts of my system am I willing to let something confidently wrong touch?

## Where the tooling stands

Pick the part of the stack you care about. Each guide compares tools on the vendor's published pricing and documented behavior, and is specific about the failure modes each one has.

- Best AI Coding Assistants: the agentic and autocomplete tools (Claude Code, Cursor, Copilot, Codex and more), sorted by whether you want terminal-native agents, IDE integration, or fast completion.
- Best AI for Code Review: the PR-review bots (CodeRabbit, Greptile, Graphite, Qodo), and the honest reality of their false-positive noise.
- Best AI for Debugging: local debug assistants versus production observability AI ([Sentry Seer](https://docs.sentry.io/product/ai-in-sentry/seer/), Datadog Bits), and which fits a stack trace versus a live incident.
- Best AI for DevOps: IaC generation, AIOps and CI/CD copilots (Pulumi AI, PagerDuty, k8sGPT), and why AI-generated infra needs the most scrutiny.
- ChatGPT for Coding: ten real use cases with prompts, when a coding agent beats raw ChatGPT, and where it hallucinates.

One warning about all five: this space renames and repackages faster than any other in the guides. [Windsurf became Devin Desktop](https://devin.ai/blog/windsurf-is-now-devin-desktop), several tools restructure pricing quarterly, and free tiers get quietly capped. Treat any figure you read, here or anywhere, as something to confirm on the vendor's live page before you commit a team to it. Free tiers across LLM APIs, with their caps and training terms, are compared in Free LLM APIs in 2026.

## The review bottleneck

Start with one assistant that matches how you work. If you live in a terminal and want something that reads the whole repo and runs your test suite, that points one way. If you want fast in-editor completion with your hands on the wheel, another. Most engineers get the bulk of the value from one strong assistant, then add a review bot, then an observability or DevOps AI if they own infrastructure. Running three assistants in parallel mostly produces three opinions and one confused codebase.

The harder problem is what to do with the volume. A few things that help. Keep AI-generated changes small enough to actually read; an agent will happily produce a diff of several hundred lines across a dozen files, and a diff that size gets approved rather than reviewed. Make the agent explain its reasoning in the PR description, since a wrong explanation is much easier to spot than a wrong line of code buried in a refactor. Insist on tests you understand rather than tests the same model wrote to satisfy itself. And route review effort by blast radius instead of by line count: auth, migrations, permissions, billing and infrastructure get read line by line, while a generated test fixture does not.

Review bots help with the mechanical layer, and they come with a cost that guide covers in detail. They catch the obvious things reliably and generate enough noise that teams start scrolling past them, which is worse than not having one. Tune what they comment on early, before your engineers learn to ignore the bot. (Devshot covers AI and dev tooling daily, in five minutes, if you want the changes without the release-note archaeology.)

The recurring failure mode underneath all of this is unchanged since the autocomplete era: models invent APIs that do not exist, cite library versions that shipped two years ago, and produce plausible code with subtle bugs in the exact places where plausible is not good enough. The tools have improved. The requirement that someone understands the change before it merges has not moved at all.

## What not to hand a model

Three categories, in increasing order of how badly it goes.

Secrets, first, and they leak by accident more often than by decision. A .env file open in the editor, a config with a live key, a stack trace with a token in a header: agents pull context automatically, so this is a tooling problem, not a discipline problem. Configure ignore rules for secret paths, use a scanner in pre-commit, and rotate anything that has been in a prompt rather than hoping.

Proprietary code, second. Consumer accounts can retain what you send and use it to improve models unless that has been switched off in the data controls, which is a per-account setting somebody chose in a hurry. Before your source goes anywhere, get the plan and the data policy in writing: business, team and enterprise tiers are contractually different from personal ones, and several coding tools offer zero-retention or self-hosted deployment specifically for this. If you are under a customer contract or an NDA that constrains where code can be processed, that constraint applies to your assistant too, and "the developer pasted it" is not a defense anyone has enjoyed making.

Third, and least discussed: anything you cannot review. If a change lands in a subsystem nobody on the team currently understands, an agent has not saved you work, it has converted an afternoon of reading into an unbounded liability. Generate into code you can evaluate. Where you cannot evaluate it, read the code first and generate second.

## What these tools actually cost

We price every tool we review, so this is measured rather than estimated. Across 429 tools, 293 publish a price and 33% offer a free tier. Among developer tools, the median entry plan is $24.50 a month, which runs above the $24 median across every category we price.

The spread matters more than the median. Half of the developer tools sit between $10 and $49, and the range runs from $2.49 to $299. A quoted "starting at" price near the bottom of that range usually means per-seat add-ons land on top of it.

Price point | Developer tools | All tools |

Cheapest paid plan | $2.49 | $1 |

Lower quartile | $10 | $10 |

Median | $24.50 | $24 |

Upper quartile | $49 | $49 |

Most expensive | $299 | $990 |

Tools measured | 18 | 293 |

Median advertised entry price/mo. Source: Dupple pricing index, 293 tools with public pricing out of 429 reviewed, 2026-08-19.

## FAQ

### What is the best AI coding setup in 2026?

It depends on the job rather than on a ranking. For agentic work inside a repo, Claude Code and Cursor lead. For pull request review, CodeRabbit and Greptile. For production debugging, Sentry Seer and Datadog Bits AI. For quick explanations and one-off scripts, a general assistant in a browser is fine. Most teams need one assistant and one review bot, not five subscriptions.

### Do I need a coding agent, or is a chat window enough?

For explanations, snippets and throwaway scripts, a chat window is genuinely enough. For work inside a real repository, dedicated agents win clearly, because they see your files, run your tools and make coordinated changes across several of them. The difference is not the model, it is the context and the ability to execute.

### Why did my team not get faster after adopting an agent?

Usually because generation was never the constraint. If pull requests were already queuing, adding output makes the queue longer. Look at where changes wait: review capacity, CI time, deploy windows. Shrinking diffs and routing review by risk buys more than a better model does.

### Is it safe to use AI on proprietary code?

It depends on the account and the tool, and it is worth ten minutes to find out rather than assuming. Consumer plans may retain and train on what you send unless that is disabled; business and enterprise plans, and coding tools with zero-retention or self-hosted options, exist precisely for this. Check the data policy, configure your ignore rules so secrets never enter the context in the first place, and treat anything covered by a customer contract as off limits until legal says otherwise.

Some offers on this page may be paid placements or contain affiliate links.

---

# CNCF announces Kubeflow graduation

URL: https://devshot.co/blog/cncf-kubeflow-graduation
Type: news
Published: 2026-08-19
Updated: 2026-08-19
Summary: The Cloud Native Computing Foundation has graduated Kubeflow to its highest maturity tier, cementing the toolkit as the standard for cloud-native AI…

News

## CNCF announces Kubeflow graduation

The Cloud Native Computing Foundation has graduated Kubeflow to its highest maturity tier, cementing the toolkit as the standard for cloud-native AI and machine-learning operations on Kubernetes.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated August 19, 2026 · 3 min read

The Cloud Native Computing Foundation (CNCF) has announced that Kubeflow has graduated, reaching its highest tier of project maturity. In its [official announcement](https://www.cncf.io/announcements/2026/08/17/cncf-announces-kubeflows-graduation-solidifying-the-standard-for-cloud-native-ai-operations/), the CNCF frames the move as a signal that Kubeflow is now the standard for running cloud-native AI and machine-learning operations on Kubernetes. As [DevOpsDigest reports](https://devopsdigest.com/cncf-announces-kubeflow-graduation), graduation places Kubeflow in the same company as foundational projects like Kubernetes and Prometheus.

## What CNCF graduation means

CNCF projects move through three stages: sandbox, incubating, and graduated. Graduation is the top rung, and it is not handed out lightly. To reach it, a project has to prove production adoption at scale, a healthy multi-vendor community, documented and mature governance, a committed security posture (including a third-party audit), and adherence to the CNCF code of conduct. In plain terms, a graduated project is judged stable, well-governed, and safe to build a business on. It is the same bar cleared by the projects most enterprises already run in production, so the tier carries real weight with the teams who sign off on infrastructure.

## Why it matters for MLOps teams

For teams doing MLOps, the label is a de-risking signal. Kubeflow bundles the pieces of a machine-learning platform, notebooks, pipelines, training operators, hyperparameter tuning, and model serving, into a toolkit that runs natively on Kubernetes. Graduation tells a platform engineer that this stack has staying power, a broad contributor base that no single vendor can yank out from under them, and the governance maturity that enterprise procurement and security reviews demand. That makes it far easier to standardize on Kubeflow instead of stitching together bespoke tooling for every model. For organizations already invested in Kubernetes, it also means their AI platform can share the same operational patterns, monitoring, and access controls as the rest of their infrastructure, rather than living as a separate island that a small team has to babysit.

## The bigger picture

The timing fits a wider trend: AI workloads are moving from experiments into production, and the infrastructure beneath them is consolidating around Kubernetes. A graduated Kubeflow gives the industry a common, vendor-neutral backbone for shipping models, much as Kubernetes did for containers. It also lands in a year of fast movement across developer tooling, from Cursor's push into code hosting to the steady mainstreaming of AI-assisted coding.

## The takeaway

Kubeflow's graduation will not change how the software works overnight, but it changes how teams can justify choosing it. A CNCF graduated stamp is the closest thing the cloud-native world has to a maturity guarantee, and it puts Kubeflow firmly in the default column for anyone building an ML platform on Kubernetes. If you are mapping out where the field is headed for engineers, our overview of AI for developers is a good next read.

Some offers on this page may be paid placements or contain affiliate links.

---

# Cursor launches Origin, a code hosting platform built for AI coding agents

URL: https://devshot.co/blog/cursor-launches-origin-code-hosting
Type: news
Published: 2026-08-18
Updated: 2026-08-18
Summary: Cursor has launched Origin, a GitHub-style code hosting platform designed for AI coding agents and switched on by default for paid users.

News

## Cursor launches Origin, a code hosting platform built for AI coding agents

Cursor has launched Origin, a GitHub-style code hosting platform designed for AI coding agents and switched on by default for paid users. It arrived the same day GitHub suffered a major outage, and its data terms are drawing scrutiny.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated August 18, 2026 · 3 min read

Cursor has launched Origin, a code hosting platform built for AI coding agents rather than human contributors alone. As [VentureBeat reports](https://venturebeat.com/infrastructure/cursor-launches-origin-code-hosting-platform-as-github-outage-exposes-opening-in-ai-coding-race), it is a direct challenge to GitHub, and the timing was pointed: Origin arrived the same day GitHub went through a major outage. [TechStartups](https://techstartups.com/2026/08/17/cursor-launches-origin-a-github-rival-built-for-ai-coding-agents/) calls it a GitHub rival designed from the ground up for agents.

## A forge built for agents

The pitch is that traditional Git hosting was built for people opening pull requests, while Origin is built for AI agents that read, write, and review code at machine speed. Cursor has turned it on by default for paid users, which means a large base of developers is being nudged onto its forge without opting in, an aggressive distribution move that instantly gives Origin scale a new platform would normally spend years building.

## The data terms to read closely

That default-on posture is exactly why the fine print matters. [The Next Web reports](https://thenextweb.com/news/cursor-origin-opt-out-data-terms-spacex-github-outage) that Origin's data terms are opt-out and not fully published, so code and activity may feed Cursor's systems unless a developer actively turns it off. For teams with proprietary code, that is the detail to settle before Origin becomes the default place their repositories live.

## Timing that stung GitHub

Launching a GitHub alternative during a GitHub outage is not subtle. [Neowin argues](https://www.neowin.net/news/elon-musks-github-alternative-couldnt-have-arrived-at-a-worse-time-for-github/) the outage could not have come at a worse moment for Microsoft's platform, handing Origin a live demonstration of why a second option matters. Cursor is now inside Elon Musk's orbit after its acquisition by SpaceX closed, which gives this forge serious compute and capital behind it.

## The takeaway

Origin turns Cursor from a coding assistant into a place your code actually lives, which is a much stickier position and a direct shot at GitHub. The default-on rollout gets it users fast, but the opt-out data terms deserve a careful read before you migrate anything sensitive. If you are weighing your setup, our Cursor versus GitHub Copilot comparison and our guide to AI-assisted coding both help, and the broader shift toward tools that act on their own is covered in our cross-network explainer on [agentic AI](https://aipresso.com/blog/what-is-agentic-ai).

Some offers on this page may be paid placements or contain affiliate links.

---

# SpaceX has officially closed its $60 billion Cursor acquisition

URL: https://devshot.co/blog/spacex-closes-cursor-acquisition
Type: news
Published: 2026-08-18
Updated: 2026-08-18
Summary: Cursor, the AI coding tool, is now officially part of SpaceX. The $60 billion deal first structured in April has closed…

News

## SpaceX has officially closed its $60 billion Cursor acquisition

Cursor, the AI coding tool, is now officially part of SpaceX. The $60 billion deal first structured in April has closed, folding one of the most-used coding assistants into Elon Musk's empire alongside xAI, with compute as the clear prize.

L

[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)Founder, Dupple · Updated August 18, 2026 · 3 min read

Cursor, the AI coding tool that became a fixture of the "vibe coding" era, is now officially part of SpaceX. As [TechCrunch reported](https://techcrunch.com/2026/08/15/spacex-officially-closes-its-cursor-acquisition/), the two companies confirmed the deal has closed in an announcement on the Cursor blog, completing a $60 billion acquisition that Elon Musk's rocket company first structured earlier this year.

## How the deal got here

The path was unusual. In April, SpaceX and Cursor announced a partnership to build technology together, a deal that also handed SpaceX an option to acquire Cursor for $60 billion. SpaceX, which had already absorbed Musk's AI company xAI earlier in the year and had since gone public, moved forward on that option. The closing this month simply finishes what April set up. For Cursor's co-founders it is a life-changing outcome, and the team lands inside one of the most compute-rich organizations on the planet.

## Compute is the point

Read the Cursor announcement and one theme repeats: SpaceX's computing infrastructure. That is the real logic of the deal. Cursor's team gains access to SpaceX's AI chips, the same hardware backbone that came with xAI. For a company building coding agents, the binding constraint is the cost of training and serving models at scale, so cheap, abundant compute is worth more than almost anything else. Owning the stack that runs your models, rather than renting it, changes the unit economics of every request.

## What it means for developers

Cursor sits at the center of how a lot of engineers now write code, so its ownership matters to the people who depend on it. The open questions are the familiar ones after any acquisition: how independent the roadmap stays, which underlying models it leans on, and whether pricing shifts once the compute bill is internal. If you are weighing your editor, our head-to-head on Cursor versus GitHub Copilot still holds up, and our guide to AI-assisted coding covers the workflow either way. The broader shift toward tools that act on their own is worth understanding too, which our cross-network explainer on [agentic AI](https://aipresso.com/blog/what-is-agentic-ai) unpacks.

## The takeaway

A $60 billion price tag on a coding-tool startup would have sounded absurd two years ago. What it buys Musk is distribution into millions of developer workflows plus a reason to keep his own compute humming. Watch what happens to Cursor's model choices and pricing next, because that is where a compute-first owner tends to leave fingerprints.

Some offers on this page may be paid placements or contain affiliate links.