Warp Pricing (2026): Free Terminal, Metered AI

Warp's terminal is free, but Warp-managed AI starts at $20 per month and heavy cloud work can draw inference, compute, and platform credits from one balance. Build is the right default until usage crosses roughly 12,000 credits in a month; above that line, Max becomes the cleaner budget choice.

Warp pricing at a glance (August 2026)

Warp has five current tiers: Free, Build, Max, Business, and Enterprise. Prices and limits below were verified against Warp's live pricing page and billing documentation on 20 August 2026.

The sticker price hides the central decision. Warp is not selling a fixed number of agent tasks. It is selling a credit balance whose consumption changes with the model, context, tool calls, and whether work runs locally or in Warp's cloud.

Which Warp plan should you choose?

Build is the sensible paid default for most individual developers. The other tiers solve narrower problems: Free removes the subscription, Max lowers the unit cost at sustained volume, Business buys administration, and Enterprise buys custom infrastructure.

Free: enough when Warp is your terminal, not your AI provider

Warp Free costs $0 and keeps the core terminal. It includes limited cloud-agent access, 3 indexed codebases with up to 3,000 files each, 30 stored cloud conversations, 10 Warp Drive workflows, 3 notebooks, and up to 5 shared sessions.

The limit that matters is simpler: Free includes no bundled Warp-managed AI usage. You can still buy add-on credits or connect your own inference. That makes Free a legitimate working plan, not just a trial, when your model bill already lives elsewhere.

Build: the $20 default

Warp Build costs $20 monthly or $18 per month equivalent on annual billing. Its 1,500 monthly credits represent $20 of agent usage at API rates. It raises indexing to 40 codebases with up to 100,000 files each, allows 20 concurrent cloud agents with 2 vCPU and 4 GiB RAM per run, and removes the limits on Warp Drive objects, shared sessions, and cloud conversation storage.

Build fits a solo developer or small team that wants Warp to provide the models and agent runtime but has not proven heavy, steady credit demand. It also gives you the lowest-risk way to learn your own usage curve before committing $2,160 to annual Max.

Max: capacity, not extra governance

Warp Max costs $200 monthly or $180 per month equivalent on annual billing. It includes 18,000 credits, 12 times Build's allowance, and Warp values that pool at $240 of agent usage. Cloud concurrency stays at 20, but each cloud run gets 4 vCPU and 8 GiB RAM instead of Build's 2 vCPU and 4 GiB.

Max does not expand the codebase limits beyond Build's 40 codebases and 100,000 files each. Its job is capacity. If your monthly work remains well below the break-even line, Max is prepaid inventory you may lose at reset.

Business: pay for control, not more AI per person

Warp Business costs $50 per user monthly or $45 per user monthly equivalent on annual billing. Each seat still receives the same 1,500 credits as Build. The premium buys up to 25 paid seats, 80 concurrent cloud agents with 8 vCPU and 16 GiB RAM per run, 80 indexed codebases, SAML-based SSO, team usage metrics, admin-configurable data controls, spend caps, and team-configured model controls.

A five-seat Business team therefore starts at $250 per month while receiving 7,500 plan credits across individually allocated seats. Do not buy it because one developer is heavy. Buy it when security and administration are requirements.

Enterprise: custom pools and infrastructure

Warp Enterprise has no public list price. It adds unlimited seats, custom shared credit pools and usage terms, custom indexing, advanced spend controls, an analytics API, multi-admin governance, self-hosted cloud agents, managed bring-your-own-LLM routing, and dedicated implementation and account support.

Enterprise is the only reliable quote when the requirement is a pooled allowance, self-hosted workers, or inference routed through your own cloud. Those are contract questions, not self-serve pricing questions.

Is Warp free?

Warp is free as a terminal, but not as a source of managed AI. The Free plan never needs to become a paid subscription if you use shell features, run a third-party agent CLI directly, or connect inference you already pay for.

Free users can add an OpenAI, Anthropic, or Google API key, connect a SuperGrok subscription, or point Warp at a custom OpenAI-compatible endpoint. Requests then create a second bill with that provider. This guide to the cheapest AI APIs helps size that separate model cost.

There is an organization-size condition. Bring-your-own-key and custom inference are available on every plan for individuals and organizations with 10 or fewer employees. Larger organizations need Business or Enterprise to use customer-supplied inference inside Warp.

Who never needs to pay Warp? A developer who wants the terminal, runs agents locally, and is comfortable paying the model provider directly can remain on Free indefinitely. A developer who wants Warp-managed models can also stay Free and buy add-on packs without subscribing, but Warp says Free-plan pack rates are 20% higher than paid-plan rates. Repeated add-on buying is the signal that Build may be cheaper.

How Warp credits translate into work

A Warp credit is a unit of processing, not one prompt and not one completed ticket. Warp says every Agent interaction consumes at least 1 credit, while larger models, more context, more tool calls, multi-step work, and cache misses can make a similar-looking task consume several.

Warp now meters three buckets:

• AI credits pay for the model call when Warp supplies inference.
• Compute credits pay for Warp-hosted sandboxes used by cloud agents.
• Platform credits pay for the run lifecycle, integrations, dashboards, APIs, and observability around cloud work.

All three draw from the same plan balance and then the same add-on balance. That means Build's 1,500 credits are not automatically 1,500 requests, and a cloud-heavy workflow can buy fewer coding outcomes than a local workflow using the same model.

The subscription still supports useful cost-per-outcome scenarios. If one representative outcome averages 10 credits, Build's allowance covers 150 outcomes and the $20 subscription contributes about $0.13 per outcome. At 25 credits, it covers 60 outcomes at about $0.33 each. At 50 credits, it covers 30 outcomes at about $0.67 each. These are planning scenarios, not promised task counts.

At the plan level, monthly Build costs about 1.33 cents per included credit. Monthly Max costs about 1.11 cents, while the largest published add-on pack costs about 1.54 cents per credit. Annual billing takes those included-credit rates to 1.20 cents for Build and 1.00 cent for Max.

Build versus Max: the break-even

Max becomes the better monthly commitment once normal usage exceeds about 12,000 credits. That threshold comes from Warp's discrete add-on packs, not from dividing one headline price by another.

Paid plans can buy 400 credits for $10, 1,000 for $20, 3,000 for $50, or 6,500 for $100. Larger packs lower the unit cost, but you cannot buy an arbitrary number of credits.

At 12,000 monthly credits, Build plus the cheapest sufficient mix of packs costs $190. At 12,001 credits, the next sufficient pack combination takes the bill to $200, tying Max. Max still includes 18,000 credits, so it gives 5,999 more credits at the same price. At the full 18,000-credit level, Build plus packs costs $290, making Max $90 cheaper for that month.

Annual-equivalent pricing moves the line slightly lower. Build remains cheaper through 11,400 credits in a typical month. At 11,401 credits, its $18 base plus sufficient packs reaches $188, while annual Max is $180 per month equivalent.

Irregular use changes the judgment. Add-on credits roll over for 12 months, so a developer with one 15,000-credit release month and several light months can stay on Build and carry the unused pack balance. Max wins when high consumption is persistent enough that included credits will be used before each reset.

Hidden costs in Warp pricing

The surprise bill comes from credit behavior, not the base subscription. Four rules decide whether the headline price survives contact with your workflow.

Included credits disappear; add-on credits do not last forever

Plan-included credits reset every 30 days, do not roll over, and cannot be transferred. Build is therefore a use-it-or-lose-it allowance.

Add-on credits behave differently. They roll over between cycles, remain valid for 12 months from purchase, and stay available if you move back to Free. On self-serve plans they are scoped to the individual user, not pooled across the team.

That user scope creates a team-budget wrinkle. Any active member can purchase their own add-on credits, the team's saved payment method is charged, and each purchase counts against the team-wide cap. A manager who assumes only admins can spend may discover a legitimate charge from a non-admin user.

Auto-reload can turn a cap into the working budget

Auto-reload is disabled for new subscribers. When enabled, it buys the selected pack whenever the balance reaches 100 credits. Warp starts the setting with a $200 monthly spend cap, which an admin can change.

The cap blocks purchases above it, but it is not a forecast. A $200 cap plus a $20 Build subscription authorizes a $220 month. Leave auto-reload off during the first measurement cycle, then set a cap from observed work rather than from the default.

Cloud work can consume three kinds of credits

Every cloud-agent run on every plan consumes platform credits. A Warp-hosted cloud run can also consume compute credits, and a Warp-managed model call consumes AI credits. Those charges share one balance.

Local runs on Free, Build, or Max do not consume platform credits, even with your own key. Local Business and Enterprise runs using customer-supplied inference do consume platform credits, while local runs using Warp-managed inference do not. That is the hidden split a team can miss when it assumes BYOK removes Warp usage charges.

Warp does not promise one fixed credit cost per task. The duration, model, context, and tools decide consumption. Cloud automation should therefore be budgeted from measured credits per completed job, not from concurrent-agent capacity alone.

BYOK replaces one meter with two bills

The Warp Agent app and CLI share the same plan, credits, models, and conversations. When Warp supplies inference, usage comes from Warp credits. When you bring a key or endpoint, the provider bills the model call directly.

BYOK can make Free the cheapest Warp plan for local work, but it does not make inference free. It also moves data-retention terms to the provider behind your key. The right comparison is the Warp subscription plus provider usage, not the Warp subscription alone.

Warp Factories shares the same balance

Warp Factories is in Early Access and does not require a separate subscription. Its runs draw from the same plan credits. Free pays a 20% markup over API rates; Build, Max, and Business include factory work in plan credits and charge additional use at API rates. Selected early-access teams can receive up to $10,000 in factory usage, but that offer is not a permanent unit-cost assumption.

Annual billing is discounted, but cancellation timing matters

Annual pricing reduces Build from $20 to $18 per month equivalent, Max from $200 to $180, and Business from $50 to $45 per user. Cancelling leaves the subscription active through the end of the yearly billing cycle.

Warp's refund policy is more flexible than a strict no-refund annual lock. Monthly subscribers can receive a full refund only when they cancel or downgrade within 24 hours of the charge and have used no credits. Annual subscribers can receive a full refund within 15 days when no credits were used; after that, Warp refunds remaining full months, excluding the first month. Add-on credits are refundable only when none from the purchase were used.

Warp versus Cursor and Claude Code on price

Warp Build, Cursor Pro, and Claude Pro all start at $20 per month, but they sell different centers of gravity. The best choice follows where you want the agent to live and which second bill you can tolerate.

Cursor Pro is the editor-first choice at $20 monthly. Its live page includes extended Agent limits, frontier models, MCPs, skills, hooks, and cloud agents. When included usage is exhausted, on-demand use is billed in arrears. Pick Cursor when daily work is organized around an AI editor and you want the agent inside that interface.

Claude Pro is the Claude-first terminal choice at $20 monthly or $17 per month on a $200 annual subscription. Claude Code is included, but Claude usage across web, desktop, mobile, and Claude Code shares one pool. Pick it when one Claude subscription should cover both coding and broader knowledge work, and accept that unrelated Claude activity can reduce coding capacity.

Warp Build is the terminal-first multi-model choice at $20 monthly. It combines the terminal, Warp Agent, cloud agents, and model choice under a credit meter. Pick it when terminal context and cloud execution matter more than an editor-native experience.

At five team seats, the governance premium becomes visible: Warp Business starts at $250 per month while Cursor Teams starts at $200. Warp earns the extra $50 only when its SSO, admin-configurable data controls, larger cloud-agent resources, or terminal-centered workflow are the better fit.

If you are also considering an agent-first development environment at the same price point, the Antigravity pricing breakdown shows how a quota model changes the decision.

The rule is direct. Choose Cursor for the editor, Claude Code for a shared Claude subscription, and Warp for a terminal plus cloud-agent platform. A matching $20 price does not make them interchangeable.

What changed in Warp pricing

Warp changed from fixed legacy tiers to a usage-metered budget, then widened who could bring their own inference. The result is more control for light users and more responsibility for anyone automating work in the cloud.

On 30 October 2025, Warp introduced Build at $20 per month with 1,500 credits and deprecated Pro, Turbo, and Lightspeed. Existing users moved at their first renewal after 1 December 2025. Add-on credits replaced legacy overages and gained 12-month rollover.

On 21 May 2026, Warp opened BYOK to every plan, made new self-serve add-on purchases user-scoped, and added explicit seat limits. It also announced platform-credit metering for self-serve cloud runs beginning 1 July 2026.

That July change is the business consequence. A cloud agent can now consume platform and compute capacity even when your own key pays for the model. Teams that priced cloud automation as model tokens alone need to add Warp's shared credit pool to the budget line.

Frequently asked questions

How much does Warp cost?

Warp Free costs $0. Build costs $20 monthly or $18 per month equivalent annually. Max costs $200 monthly or $180 equivalent annually. Business costs $50 per user monthly or $45 equivalent annually. Enterprise pricing is custom.

Is Warp totally free?

The core terminal is free. The Free plan includes no bundled Warp-managed AI usage, so AI requires purchased add-on credits or inference you pay for through your own key, subscription, or endpoint.

Who never needs to pay for Warp?

A developer using Warp as a terminal, running third-party agents directly, or using local Warp Agent with their own inference can stay on Free. Organizations with more than 10 employees need Business or Enterprise to use customer-supplied inference inside Warp.

Does Warp offer a student discount?

No. Warp currently lists no discount for students or nonprofits. The standard Free plan is the available no-subscription option.

What is Warp's refund policy?

Monthly subscriptions can receive a full refund within 24 hours only when no credits were used. Annual subscriptions have a 15-day full-refund window with no credit use, followed by refunds for remaining full months excluding the first month. Unused add-on purchases can be refunded only when none of those credits were consumed.

Does Warp offer annual pricing?

Yes. Annual billing is 10% lower than monthly pricing: Build is $216 per year, Max is $2,160, and Business is $540 per user. Measure a monthly cycle first because plan credits reset and do not roll over.

Did Warp pricing change in 2026?

Yes. On 21 May 2026 Warp expanded BYOK to every plan, changed new add-on credits from pooled to user-scoped on self-serve teams, added seat limits, and scheduled platform-credit charging on self-serve cloud runs from 1 July 2026.

Is Warp better than Claude Code?

Warp is the better fit when you want a multi-model terminal, cloud agents, and one Warp credit system. Claude Code is the better fit when you want Claude in the terminal and already value the wider Claude Pro subscription. Both start at $20 monthly, but Claude shares usage across its products while Warp meters AI, compute, and platform work.

The Monday move

Start with one controlled week and make the plan follow the work. The purpose is to establish credits per accepted outcome before auto-reload turns uncertainty into a standing budget.

1. 1

   Choose one repeated outcome

   Pick a unit you can recognize, such as one accepted bug fix, one reviewed pull-request draft, or one completed investigation. Do not measure prompts because one prompt can trigger very different amounts of work.

2. 2

   Run the cheapest viable setup

   Stay on Free if local BYOK covers the job. Otherwise buy one monthly Build seat at $20. Keep cloud and local work separate in the record because cloud runs can consume additional credit buckets.

3. 3

   Leave auto-reload off

   Use the initial balance as a hard observation window. Record the credits consumed by each accepted outcome and note whether a cloud sandbox was involved.

4. 4

   Let the 12K line decide

   Project a normal month from the measured week. Keep Build at 12,000 credits or less. Move to Max above 12,000 when that pattern repeats. Choose Business only when SSO, team controls, or organization-size rules require it.

The decision next Monday is a budget rule, not a product preference: free for terminal and local BYOK, Build for measured moderate use, Max for sustained volume, and Business for governance.

Get the AI Business Workflow Audit Checklist

Turn software subscriptions into a measured workflow budget with a clear owner, outcome, quality bar, and upgrade trigger. Subscribe to get the checklist free.