Fastest accurate answer
Send /status in OpenClaw for the current session, enable /usage tokens or /usage full for a footer on each reply, and use /usage cost for a local cost summary. Run openclaw status --usage in your terminal when the question is provider quota rather than tokens.
Usage map
Follow the number to its source of truth
Session tokens
Input, output, and cache accounting from model turns in the session transcript.
/usage tokens · /usage cost
Context occupancy
The latest prompt snapshot compared with the active model's runtime context budget.
/status
Provider quota
A provider-reported usage window, normalized by OpenClaw when a supported credential and endpoint are available.
openclaw status --usage
The arrows show a diagnostic sequence, not a mathematical conversion: you cannot derive quota remaining from local token totals.
OpenClaw usage commands and what they answer
Run slash commands as standalone messages in an OpenClaw chat. Run the CLI command in a terminal connected to the same OpenClaw setup. This table describes the real surfaces without inventing example totals that could be mistaken for yours.
A repeatable OpenClaw token audit
- 1
Capture the current session state
Send
/status. Record the runtime model, last-response token fields, and context usage together; a token number without its model and session is hard to compare. - 2
Turn on per-reply accounting before the next task
Use
/usage tokensfor the lean footer or/usage fullwhen model and context detail matter. The preference persists for that session until you change or reset it. - 3
Check local estimated cost separately
Send
/usage cost. Treat the result as a local, API-equivalent estimate based on available transcript usage and configured model pricing—not as an invoice. - 4
Ask the provider when the question is quota
Run
openclaw status --usage. A missing provider window can mean the provider, auth profile, or credential path does not expose a supported quota endpoint; it does not prove unlimited usage. - 5
Build a cross-agent local history
Run
npx whoburnedmore --localto read directly supported OpenClaw records alongside other supported coding-agent logs and view daily/model aggregates without an automatic aggregate submission. In an interactive terminal, decline the optional publish offer to keep the run local.
Keep OpenClaw history local or add it to your aggregate profile
Offline aggregate view
npx whoburnedmore --local
Reads supported local records and displays cross-agent daily and model history without initial sign-in or automatic submission. Decline the optional interactive publish offer to keep it local.
Aggregate profile workflow
npx whoburnedmore
The standard command signs you in and submits daily usage aggregates for your profile. Use the local flag when no aggregate submission is wanted.
Do not collapse these OpenClaw metrics into one number
- Session token usage
- Provider usage attached to model turns in the stored session. It can include input, output, cache activity, and several calls made during a tool loop.
- Context used
- The latest prompt-oriented snapshot occupying the active model's runtime context budget. It is not a cumulative lifetime token total.
- Provider quota
- A provider-reported allowance window such as a percentage remaining. Availability and reset rules depend on the provider and credentials.
- Estimated cost
- A calculation from usage metadata and configured per-million-token rates. For subscriptions, credits, negotiated rates, or missing pricing, the provider's billing page remains authoritative.
Troubleshooting missing or surprising usage
Limits and interpretation
OpenClaw's native session summaries are the first stop for a live run. Transcript-backed fields can fill gaps when a current session snapshot is sparse, but they still describe the local records available to OpenClaw—not every request ever billed to an account.
whoburnedmore's direct OpenClaw support is useful for durable daily/model comparisons across supported agents. It does not turn local token counts into a subscription entitlement, and its cost output remains API-equivalent unless it is backed by a provider bill.
Command behavior and provider support were reviewed against the official OpenClaw documentation on August 28, 2026.
Sources
- OpenClaw usage tracking documentation
Official explanation of OpenClaw token usage, context, and cost visibility.
- whoburnedmore CLI source
Canonical CLI commands and export behavior; use --local when no aggregate upload is wanted.