Usage views answer two questions: how much capacity are you consuming, and where is model spend going.
Find them under Settings → Billing → Usage and Detailed reports.
Capacity usage
Shows slot utilisation over time — how much of the capacity you are paying for is actually working.
Read it alongside queue depth. High utilisation with a long queue means you are under-provisioned. Low utilisation means you are paying for slots that sit idle. High utilisation with an empty queue is the target: capacity matched to demand.
Agent-hours, on plans that meter them, count only while an agent is actively working. Queue time is free.
Model spend
Broken down so you can attribute it rather than just observe it. The dimensions worth looking at:
- By project — which codebase is consuming the budget
- By agent — almost always the developer agent, given how often it runs
- Over time — the shape matters more than any single day
Reading it usefully
The useful ratio is spend against delivered work, not spend alone. A month with more spend and proportionally more merged pull requests is working as intended.
When spend rises without matching delivery, the causes in rough order of likelihood:
- Rework. Work that runs twice costs twice. Rising rework usually traces to loose specifications — see Specifications.
- Profile drift. Routine work being run under a more capable profile than it needs, which raises the cost of every stage. See AI profiles.
- Task sprawl. Broad tasks explore more to work out their own scope.
Exporting
Detailed reports can be exported for your own analysis or to feed internal chargeback. If you need this on a schedule rather than on demand, see Reports.
Billing history
Invoices and payment history live under Billing history, with payment methods managed through the billing portal. See Billing history.