How to Check Warp AI Usage and Cost

Warp is a modern, Rust-based terminal whose headline feature is Agent Mode — an AI assistant that lives in your command line and can plan, run, and fix shell work for you. Unlike agents that expose raw token logs, Warp bills its AI as requestson a tiered allowance: the Free plan ships with a small monthly bucket, and paid tiers raise the ceiling. That design is why so many users went looking for “where is my Warp usage” in the first place — and why the answer is an in-app meter rather than a file on disk.

How do I check my Warp AI usage?

Warp keeps its AI allowance server-side and surfaces it inside the app rather than in a log file you can grep. Here is the path to the live number:

1. 1

   Open the Settings panel

   Launch Warp, then open Settings(the gear icon, or the command palette → “Open Settings”). Warp groups its panels down the left rail.
2. 2

   Select the AI section

   Choose the AI (sometimes labelled Agent or Warp AI) section. This is where Agent Mode behaviour and your allowance both live.
3. 3

   Read requests used vs your plan ceiling

   Warp displays how many AI requests you have spent in the current billing period against the cap that comes with your tier. When that bar fills, Agent Mode prompts pause until the period resets or you upgrade.

What are Warp's AI request limits?

Warp's AI allowance scales with your plan. The exact ceilings shift as Warp updates its pricing, so always confirm the current numbers on Warp's own pricing page — but the shape is consistent: a modest free bucket, then progressively larger monthly request pools on paid tiers.

This tiered-request model is also what sparked very public pushback: when Warp tightened or re-shaped its caps, vocal users protested on social channels that the allowance ran out faster than they expected for real agent work. The takeaway for you is practical — the only authority on “how much is left” is Warp's in-app meter, and a request can represent far more underlying model work than a single chat reply.

Why a request can cost more than it looks

Because Agent Mode autonomously chains steps — read a file, run a command, inspect the error, retry — one of your metered requests may drive several model calls behind the scenes. That is great for getting work done hands-free, but it means the request bar can move faster during an agentic loop than during plain question-and-answer use.

How do I estimate what my Warp usage is worth?

Since Warp sells requests rather than tokens, the effective cost of your usage is simply your subscription divided across the requests you actually spend:

If a paid tier costs a flat monthly fee and you spend, say, half of your allotted requests, your effective price per request is double the rate you would pay if you used the whole bucket. That is the genuinely useful number Warp's meter helps you compute: am I leaving allowance on the table, or am I hitting the wall and paying upgrade prices for headroom I rarely need?

How do I track my terminal agents' token spend?

Warp's in-app meter answers “how many requests have I burned?” but it cannot tell you what you spent in the other AI tools running in that same terminal. That is the gap whoburnedmore fills — and to be clear, it covers those other agents, not Warp itself:

$ npx whoburnedmore↳ scanning local agent logs (~/.claude, ~/.codex, …)   TOOL          TOKENS       EST. COST  claude code   11.6M        $38.90  codex          5.2M        $13.70  gemini cli     2.4M         $3.20  warp          (request-metered — check in-app)  ─────────────────────────────────────────  cli total     19.2M        $55.80

Notice the warprow reads “check in-app” — whoburnedmore deliberately does not invent a Warp number, because Warp does not expose a local token log to read. For the agents it does parse, you get a real daily and monthly token breakdown with an estimated API-equivalent cost.

1. 1

   Read your Warp requests in-app

   Settings → AI shows requests used against your tier ceiling. Note the figure for the period.
2. 2

   Run whoburnedmore for your logged agents

   npx whoburnedmore reads Claude Code, Codex, Gemini CLI, opencode and other local-logging agents — see the cross-tool token usage guide for the full list.
3. 3

   Keep the two views side by side

   Your Warp request meter plus your whoburnedmore token totals together describe your real AI shell footprint, even though they come from two different sources.

Does whoburnedmore read my Warp account?

No. whoburnedmore never touches your Warp account, your Warp AI requests, or anything stored on Warp's servers. It only reads the local log files that certain CLI agents write to your filesystem — directories like ~/.claude and ~/.codex. Warp keeps its AI usage server-side and in-app, so it is simply outside what whoburnedmore can see. Your prompts, code, and file names are never read or transmitted; run npx whoburnedmore --dry-run to preview exactly what would be submitted before anything leaves your machine. 🛡️

For a side-by-side look at how request-metered tools compare with raw-token terminal agents on cost and control, the Cursor usage guidewalks through the same requests-versus-tokens distinction with a different tool's numbers.