How to Check Trae Usage and Limits

Trae is ByteDance's AI-native code editor — a free, VS Code-derived environment positioned as an alternative to paid IDEs. Instead of charging per token, Trae paces you with two request tiers. Fast requests route to its quickest premium models and come in a limited allotment; once you exhaust them, you fall back to slow requests, which still work but are deliberately throttled. Because all of this is tracked on ByteDance's side, the only authoritative meter is the panel inside the Trae app itself. There is no local Trae log on your disk for a third-party tool to scan, so npx whoburnedmore does not — and cannot — read your Trae usage. It is built for the command-line agents that write their own usage logs.

How do I check my Trae usage?

Everything you need is built into the editor — you do not visit a separate billing site the way you would for a subscription IDE:

1. 1

   Open the Trae usage panel

   Inside Trae, open the account or usage area (typically reachable from your profile menu or the settings sidebar). The panel surfaces your current fast-request consumption and what remains in the period.
2. 2

   Read the fast-request meter

   This is the number that matters most. It tells you how many premium, low-latency requests you have spent and how many are left before you drop to the slow tier.
3. 3

   Note when slow requests kick in

   Once fast requests run out, Trae keeps responding through the throttled slow path. The panel makes it clear you have crossed that line so the added latency is not a surprise.

What is the difference between Trae fast and slow requests?

The two tiers are not different prices — they are different experiences. A fast request and a slow request can produce the same answer; what changes is latency and how scarce each one is. Think of fast requests as a metered express lane and slow requests as the always-open side road:

The practical upshot: if Trae starts feeling sluggish, you have almost certainly burned through your fast allotment and are now on slow requests. Checking the panel confirms it in a glance, and it resets at the start of your next cycle. To picture it, here is an illustrative split of a heavy session across the two tiers:

How do I track my CLI agents' spend?

Trae lives inside one editor, but most people also drive AI from a terminal — and those agents bill in raw tokens, not request tiers. Unlike Trae, command-line tools write their usage to local log files, which means a scanner can total them honestly. That is the slice npx whoburnedmore owns: Claude Code, Codex, Gemini CLI, opencode, and the newer terminal agents. It reads their logs, converts tokens to an estimated cost, and shows the total — no Trae numbers involved, because Trae has no log to read.

$ npx whoburnedmore↳ reading local agent logs…   AGENT          TOKENS       EST. COST  claude code    11.6M        $38.40  codex           5.2M        $14.10  gemini cli      2.7M         $3.60  trae           (IDE — no local log; check in-app)  ─────────────────────────────────────────  cli total      19.5M        $56.10

The token-to-cost math whoburnedmore applies to each CLI agent is a straight rate calculation per model — the kind of per-token accounting Trae deliberately hides behind its request tiers:

Can I combine my Trae and CLI numbers?

You can build the full picture, but it takes two readings from two honest sources — there is no single tool that pulls both:

1. 1

   Read Trae in-app

   Note your fast-request count and whether you have dropped to slow requests this cycle. That is your Trae usage; only the editor can report it.
2. 2

   Run whoburnedmore for the terminal

   npx whoburnedmore totals your CLI agents from their local logs with estimated costs. None of that touches Trae — the tools simply cover different surfaces.
3. 3

   Compare against other IDEs if you switch

   Evaluating a paid editor alongside Trae? The Cursor usage guidewalks through that IDE's request-based metering so you can weigh it against Trae's free fast/slow model.

Does whoburnedmore read my Trae data?

No — and it is worth being precise about this. whoburnedmore never connects to ByteDance, your Trae account, or any Trae service, and it has no way to see your fast or slow request counts. It only reads the local log files that command-line agents write to your filesystem, in directories like ~/.claude and ~/.codex. Your prompts, your source code, and your file names are never read or transmitted. If you want to see precisely what would be submitted before anything leaves your machine, run npx whoburnedmore --dry-run. For Trae, the in-app panel remains your single source of truth. 🛡️

Curious how a metered free editor stacks up against a paid one on cost and flexibility? The Cursor usage guidecovers the request-versus-token billing distinction in detail with that IDE's specific numbers.