Warp Pricing 2026: Free Terminal, Build $20, Max $200, Business $50

Warp's terminal is still free. The Free plan costs $0, Build is $20 a month or $18 billed annually, Max is $200 a month or $180 annually, and Business is $50 per user a month or $45 annually with a 25-seat cap, per Warp's pricing page read on October 26, 2026. Enterprise is a custom contract. Paid plans no longer sell credits: they include a dollar amount of agent usage billed at public API rates.

Our view: Build is the plan for almost every developer who uses the agent, because its included usage covers a normal month on a mid-priced model and anything beyond it costs the same per token as the model vendor charges. Bringing your own API key saves money only on the Free plan, and only for individuals and companies of 10 people or fewer, which is the cap Warp puts on BYOK.

Toolradar data: on Toolradar, the software directory we run, Warp is listed as an alternative to 13 other tools, 18th of the 161 products in its terminal tools category, as of October 2026. Most of those 161 are free, so the paid question for Warp is about the agent, not the terminal.

What "usage at API rates" means

Warp moved its self-serve plans from credits to dollars during 2026. The pricing cards now describe Build's allowance as agent usage included each month "at API rates", and Warp's usage and billing doc spells out the rule: on paid self-serve plans billed in dollars, inference that Warp provides is priced at the public API rates of the model you pick. A token on Claude Sonnet costs the same inside Warp as it does on Anthropic's own invoice.

The old unit has not disappeared. The comparison table lower on the same pricing page still prints the included usage as "$20 (1,500 credits)" for Build and Business, so 75 credits equal one dollar by our arithmetic, and the docs say a converted credit pack keeps its face value or its inference value, whichever is higher. One line on that table is out of step with the card above it: it lists Max at "$240 (18,000 credits)" while the Max card says $200 of usage. Warp's pages differ there, and this article uses the figure on the card.

Three things draw on the balance, not one. Inference is the model call. Compute is the sandbox when an agent runs on Warp-hosted machines, and it does not apply to local runs. Platform usage is billable agent time, charged by the agent hour for cloud runs on every plan and for local runs on Business and Enterprise that use your own inference. A developer on Free, Build or Max who runs the agent locally pays for inference only.

Free: the terminal, with the agent on pay-as-you-go

Free includes the terminal, Warp Agent CLI access, bring-your-own inference, limited cloud agents, limited Warp Drive sharing and limited cloud conversation storage. The comparison table puts numbers on the limits: 3 indexed codebases of up to 3,000 files, 30 stored cloud conversations, a shared Drive of 10 workflows and 3 notebooks, and up to 5 shared sessions.

What Free does not include is any agent usage. Warp's docs say the plan has no bundled usage for the Warp Agent, so a conversation on a Warp-provided model needs either purchased usage or your own key. Free users can buy usage without subscribing, and the pricing FAQ states that Free-plan purchases cost 20% more than the same purchase on a paid plan. That premium is the whole reason the Build plan exists for a light user: the subscription is cheaper than buying the same usage at a markup.

The other route is a key. Free users can plug in their own OpenAI, Anthropic or Google key, connect a ChatGPT or SuperGrok subscription, or point Warp at an OpenAI-compatible endpoint such as OpenRouter. Warp's FAQ limits all of that to individuals and organizations of 10 or fewer employees; a larger company must be on Business or Enterprise to use it.

Build: the plan to default to

Build adds the included usage, full Warp Agent access to frontier and open-source models from OpenAI, Anthropic, z.ai and others, add-on usage with auto-reload and a team-wide spend cap, extended cloud agents, the highest codebase indexing limits, unlimited Warp Drive objects and conversation storage, and private email support. The table lists 40 indexed codebases of up to 100,000 files, 20 concurrent cloud agents on 2 vCPU and 4 GiB of RAM, and a 10-member team limit.

The included usage resets every 30 days from the subscription date, even on an annual plan, and unused included usage does not roll over. Purchased usage is different: it rolls over, stays valid for 12 months, and is pooled across the team. A developer who buys a top-up in a heavy month keeps whatever is left for the quiet months that follow.

Max: ten times the usage, same features

Max is Build with ten times the included usage and the same add-on controls. Its cloud agents get 4 vCPU and 8 GiB of RAM, but the seat limit, codebase limits and collaboration features match Build. Warp's own FAQ describes it as the plan for heavy AI users who consistently need more capacity, and that is the honest test: the plan pays for itself only when you would otherwise buy close to the same amount in add-ons every month. Our Warp review works through the break-even in detail.

Business and Enterprise: seats, SSO and the BYOK carve-out

Business costs $45 per user a month on annual billing, or the monthly rate in the table above, for up to 25 seats, and each seat carries the same monthly usage allowance as Build. The extra money buys team usage metrics, admin-configurable data controls, SAML-based SSO, 80 concurrent cloud agents on 8 vCPU and 16 GiB machines, and the right for a company of any size to bring its own API keys and custom inference endpoints. Our arithmetic for a five-seat team: $250 a month on monthly billing, or $225 a month, $2,700 a year, on annual billing, before any add-ons.

Business is also where bringing your own inference starts to cost something. Local agent runs on Business and Enterprise that use a customer-supplied key, subscription or endpoint incur platform charges for billable agent time, measured in agent hours with partial hours billed proportionally. Warp does not print the agent-hour rate on its pricing page, so a team planning to run Business on its own OpenAI contract should check the rate in Settings before assuming the key makes Warp usage free.

Enterprise adds unlimited seats, custom shared usage pools, per-user spend limits, an analytics API, domain capture, routing inference through your own AWS Bedrock or Vertex AI account (BYOLLM), self-hosted cloud agents and a dedicated account manager, at a negotiated price.

Included usage or your own key: a computed month

Because Warp bills inference at API rates on paid plans, bringing your own key does not lower the per-token cost. It only changes who sends the invoice. The decision therefore turns on the subscription fee and the premium, which we can compute from Anthropic's pricing page, read October 26, 2026: Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, and Claude Opus 5.5 costs $4 and $20.

Take a working month of 5 million uncached input tokens and 500,000 output tokens, a plausible load for a developer who runs the agent on a few tasks a day. Our arithmetic:

Two lessons fall out. On the Sonnet month, Build and BYOK on Free land within $5 of each other, and Build brings 40 indexed codebases, unlimited Drive and the Auto model router that BYOK cannot use, since Auto always draws on Warp-provided usage. On the Opus month, Build with an add-on and BYOK cost the same $30, while buying that usage on Free costs $6 more. The premium makes Free the most expensive way to pay Warp for inference; the key makes it the cheapest, but only for the tokens you route through it.

Cached tokens shift the numbers in Warp's favor on long sessions: Anthropic prices a cache read on Sonnet 5.5 at $0.10 per million tokens, and Warp reports cache reads separately in each turn's usage breakdown. If most of your context is reused, the same month costs far less than the table on either route. Rates for the other providers are in our Claude API pricing and OpenAI API pricing explainers.

Add-ons, auto-reload and the spend cap

Free, Build, Max and Business users can buy additional usage; the amounts are chosen on the upgrade page or in Settings, and the purchase charges the team's saved card, not the buyer's own. Auto-reload is off by default. When an admin switches it on, it rebuys the chosen amount whenever included usage is spent and the purchased balance runs low, inside a team-wide monthly spend limit that defaults to $200. Warp's docs warn that the limit is not a guaranteed ceiling, because Free-plan premiums and taxes sit on top, and that heavy use by any one member can trigger repeated reloads.

Refunds are narrow. A monthly plan bought by mistake is refundable in full within 24 hours if no usage was consumed; an annual plan within 15 days on the same condition, and after that Warp refunds the remaining full months minus the first. Purchased usage is refundable only while none of it has been used. Cancelling keeps the paid features until the end of the billing cycle, and a downgrade takes effect at once with a prorated balance for the unused time.

Dupple data: 31,000+ developers read Devshot, and the audience file dated November 8, 2026 puts its open rate at 32%. Terminal and agent pricing changes are a regular item in the Devshot brief.

Which Warp plan to pick

Pick Free if you want the terminal and will run the agent through your own key, which Warp allows for individuals and teams of 10 or fewer. Pick Build as soon as you want Warp's own models, the Auto router or the codebase index, because the subscription costs less than buying the same usage at the Free premium. Pick Max only after a few months of add-on receipts show you spending near its price anyway. Pick Business when SSO, admin data controls or a company larger than 10 people that wants its own keys is the requirement.

For the product itself, read the Warp review; for the editors and agents it competes with, see Cursor pricing, Claude Code pricing and GitHub Copilot pricing, the Zed review for the other Rust-based alternative, and our best AI coding agents ranking. The whole category's price moves sit in AI coding assistant pricing 2026 and on the AI coding statistics page.

How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this page from Warp's pricing page and its plan comparison table, and from the Warp docs on plans, pricing and refunds, usage and billing, purchasing additional usage, platform usage, bring your own API key and the pricing FAQs, all read on October 26, 2026. Model rates come from Anthropic's pricing page on the same day. Team totals and the computed month are our arithmetic on those printed figures; we did not measure real token use.

FAQ

How much does Warp cost in 2026?

Warp is $0 on Free, $20 a month on Build or $18 billed annually, $200 a month on Max or $180 annually, and $50 per user a month on Business or $45 annually, per Warp's pricing page on October 26, 2026. Enterprise is custom. Build and Business seats each include a month of agent usage at API rates equal to the Build fee, and Max includes ten times as much.

Is the Warp terminal still free?

Yes. The Free plan includes the terminal, Warp Agent CLI access and the option to bring your own AI inference, with no time limit. What it lacks is included agent usage: Warp-provided models need purchased usage, bought at a 20% premium over paid-plan prices, or your own API key.

What happened to Warp credits?

Self-serve plans now bill in dollars, and inference on paid plans is priced at public API rates. The pricing page's comparison table still shows the conversion, the Build allowance as 1,500 credits, and Warp's docs say an existing credit pack converted to dollars keeps its face value or its inference value, whichever is higher. Enterprise contracts that bill in credits keep their terms.

Can I use my own API key with Warp?

Yes, on every plan, for individuals and organizations with 10 or fewer employees; larger companies need Business or Enterprise. Your provider bills the inference and it does not consume Warp usage. Warp's Auto models still use Warp-provided inference, and on Business and Enterprise local runs on your own key incur platform charges for agent time.

Does Warp usage roll over?

Included monthly usage does not: it resets every 30 days, even on an annual subscription, and unused amounts expire. Purchased add-on usage rolls over, stays valid for 12 months, is shared across the team and stays with the team that bought it.

Is Warp Max worth it?

Only for someone who already spends close to its price on add-ons. Max costs ten times Build and includes ten times the usage, with the same seat limit, codebase limits and collaboration features, so it buys capacity rather than features. Check two or three months of your billing page before upgrading.

Cite this: Devshot, "Warp Pricing 2026: Free Terminal, Build $20, Max $200, Business $50", November 2026. Prices verified on vendor pages, October 26, 2026.

Devshot tracks what developer tools charge and what changes on the invoice. Subscribe at devshot.co.

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