Cost tracking

Anvil tells you what it costs as you go. Spend is measured from real token counts and published provider rates, rolled up per session and across your whole fleet — so there are no surprises.

How the estimate is built

Cost isn't a guess. For each turn, Anvil multiplies the measured token counts by the provider's published rate — the rate in force when the turn ran — and reports it. The pricing page shows those rates, how they've moved, and exactly where the estimate stops being certain.

• Tokens are measured, not estimated.
• Rate keyed on the provider that actually served the turn.
• Priced at the rate in force when the turn ran, so historical turns stay accurate.
• Unknown is reported as unknown — never passed off as a certain number.

In the session

See the running cost of the current session from the workspace:

/cost        # token costs accumulated this session
/status      # model, token usage, and provider at a glance

Across your fleet

Your command centre rolls cost up across every deployment and every session, so you can see where spend actually goes — not just one terminal, but the whole operation.

[!NOTE] Fleet-wide cost lives in your command centre — spend rolled up across every deployment and session, not just one terminal.

Related

• Pricing — the provider rates the estimate is built from.
• The workspace — the per-session cost you see while working.