Cursor Pricing in 2026: Every Plan, Real Costs and Hidden Fees

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 evaluated 153 tools, updated September 2026. The product writeup on this site is the Cursor review, and the directory record is Cursor on Toolradar.

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, 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 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.

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 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.

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.

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. 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, the models and pricing doc, the Teams pricing doc, the June 2026 Teams note, and the Bugbot doc on September 24, 2026.

Copilot figures come from GitHub's individual plans and individual usage billing. Devin figures come from devin.ai/pricing and the self-serve billing doc. Claude figures come from claude.com/pricing, the Max article linked above, and the Team plan article. 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.