Usage analytics, costs and monitoring

Three separate things, often asked about as one:

Analytics — team usage reporting. Who is using Claude Code, how much, on what. The number an engineering manager wants when justifying seats.

Cost management — controlling spend. Model selection, prompt caching, batching where applicable, and the plan-level spend controls on Enterprise.

Monitoring — operational telemetry. The Agent SDK emits OpenTelemetry, which is the route to putting agent behaviour on the same dashboards as everything else you run.

The audit trail nobody thinks about until they need it

For API-billed usage, the Console’s Usage page exports a CSV broken down by API key and model. That is how you find which service is still calling a model that retires next month, and it is a five-minute task that has saved a lot of production incidents.

What this page does not claim

Which analytics are available on which plan is only partially documented in the sources checked. Pro and Max are recorded as limited rather than yes or no, because team-level reporting is a business-plan concept and individual plans surface something narrower.