Cercano
Token metrics
The Token Metrics settings page shows reported token consumption over time, with breakdowns by provider, model, and source. It distinguishes known counters from missing usage and surfaces accounting-health warnings so incomplete data is not silently presented as zero consumption.
Why it matters
Cost control starts with visibility. Compare which models, providers, and work sources consume tokens, then use that evidence to refine routing and delegation. An efficient or local route can be useful, but its name alone does not prove that it saved money or produced an equivalent result.
Inspect usage
Open /config, select Token Metrics, choose a date range, and inspect the
provider, model, and source breakdowns. Read coverage warnings before comparing
workflows.
Use /c, or Context in settings, for the different question of how much of
the current conversation occupies the active context window. Current context
size is not cumulative billed usage.
Accounting boundaries
- Metrics are reported usage figures, not billing-grade accounting.
- Unknown or unreported counters are not zero.
- Cache and reasoning categories can be subsets of other totals; do not add them indiscriminately to input and output tokens.
- Token counts, provider prices, cost estimates, and estimated savings are different quantities. Savings require an explicit comparison baseline.
- External usage reports can overlap internal attempts and should not simply be added together.
See models and routing and delegation.