Key findings
- Build and Business include US$20 of agent usage per seat a month. Max includes US$200 on the plan card and US$240 in the pricing table. Included usage resets monthly and does not roll over.
- Usage has three parts: inference at the model's rates, Warp-hosted compute for cloud sandboxes, and platform charges per agent hour. Platform charges apply to cloud runs and to local runs on Business and Enterprise that use your own inference.
- Purchased usage rolls over for 12 months and is shared across the team. Free-plan purchases cost 20% more. The team-wide monthly spend limit defaults to US$200 when auto-reload is on.
- Running claude or codex yourself in Warp's terminal adds no Warp charge. Run as Warp cloud agents, the same tools incur Warp compute and platform charges, and your provider bills inference.
Budget a Warp seat as a monthly dollar allowance at model API rates, plus purchased usage past it, plus compute and agent-hour charges for cloud runs. Warp's pages still disagree on Max's allowance, which plan gets team-managed keys, and which plans get zero data retention, so write those terms into the order form.
What the plans include
Warp's self-serve plans differ mainly in included usage, seat limits, and controls. Build and Max allow up to 10 team members, Business up to 25, and Enterprise is unlimited on a custom contract. Business adds SAML SSO, per-user usage reporting, admin-configurable data controls, and bring-your-own keys for organizations of any size.
Usage displays in dollars on self-serve plans. Enterprise contracts that bill in credits keep their credit pools and rates, and Warp's pricing table still meters the Build allowance as 1,500 credits.
| Plan | Included usage per month | Monthly price |
|---|---|---|
| Free | None; pay as you go or bring your own inference | |
| Build | US$20 | |
| Business (per user) | US$20 per seat | |
| Max | US$200 (plan card) or US$240 (pricing table) |
Warp pricing page, monthly billing, checked Oct 12, 2026
Where the bill comes from
Inference is charged at each model's input, output, and cache rates, so model choice and context size drive most of the variation. Warp's documentation says two similar prompts can cost different amounts and a credit is not a fixed number of prompts. Each agent response shows its token breakdown.
Cloud runs add two charges. Compute covers the Warp-hosted sandbox. Platform usage bills agent hours. For the Warp Agent, an agent hour counts the time spent generating responses and leaves out command execution. For Claude Code and Codex, it counts active runtime, including commands and time waiting on the provider. Local runs on Free, Build, and Max carry no platform charge.
- Build and Max allow 20 concurrent cloud agents, Business 80.
- Cloud sandboxes get 2 vCPU and 4 GiB on Build, 4 vCPU and 8 GiB on Max, and 8 vCPU and 16 GiB on Business.
- Self-hosted workers avoid Warp-hosted compute charges and still incur platform charges.
Bringing your own inference
Personal API keys for OpenAI, Anthropic, or Google, a SuperGrok subscription, a ChatGPT subscription, or an OpenAI-compatible endpoint move inference billing to your provider. Warp's terms limit these self-serve options to individuals and organizations with 10 or fewer employees. Larger organizations need Business or Enterprise.
On Business and Enterprise, local runs that use your own inference incur Warp platform charges, so Warp still bills platform time after inference moves to your provider. Enterprise BYOLLM bills inference to your AWS or Google Cloud account.
Where Warp's pages disagree
Three buying terms still read differently by page when GILD rechecked them on October 25, 2026. Max includes US$200 a month on the plan card and in the FAQ, and US$240 (18,000 credits) in the comparison table. The enterprise documentation offers team-managed API keys on Business and Enterprise, while the pricing table and FAQ list them for Enterprise only. The pricing page and security documentation apply zero data retention to Warp-provided inference on every plan. But the legal security page says a zero data retention policy is available on the Enterprise plan.
The Free plan's indexing limit also differs, with 3,000 files per codebase in the pricing table and 5,000 in the Codebase Context documentation. Warp's billing docs tell existing customers to read their actual allowance in Settings, Billing and usage, because existing seats can carry a different allowance from new subscriptions.
How to budget a pilot
Run the pilot on the plan you would buy, with the team spend limit set and auto-reload either off or capped. After two weeks, read each engineer's usage from Billing and usage, or from the Analytics API on Enterprise, and price the month from the subscription plus the purchased usage engineers actually drew.
- Record spend per merged change, split by model.
- Separate local agent spend from cloud runs, which add compute and agent hours.
- Check how many engineers exceed the US$20 allowance, and by how much.
Methodology
GILD read Warp's pricing page, including its plan comparison table and FAQ, and the plans and billing, platform usage, enterprise billing, and security documentation on October 12, 2026. GILD rechecked the pricing table, the pricing FAQ, the team-managed keys documentation, and the plans and refunds page on October 25, 2026. Every figure is Warp's published list price or limit, linked in Sources.
Products in this piece
Sources
- Official documentation
- 14
- Warp: Pricing
- Warp docs: Plans and billing
- Warp docs: Usage and billing
- Warp docs: Plans, pricing, and refunds
- Warp docs: Purchasing additional usage
- Warp docs: Platform usage
- Warp docs: Pricing and billing FAQs
- Warp docs: Enterprise billing
- Warp docs: Access, billing, and identity permissions
- Warp docs: Bring Your Own API Key
- Warp docs: Team-managed LLM API keys and endpoints
- Warp docs: Codebase Context
- Warp docs: Security overview
- Warp: Security