Read activity and metrics
Understand what happened, when it happened, and which system measured it.
Use Activity to inspect operations and Metrics to inspect measurements for the selected Mind. Use account usage for billing context. Engine activity and billable usage are related, but they are not the same dataset.
Start with the scope and time window
Confirm the organization, instance, and reporting period. Record the units of a value before comparing it across charts. A rate, a total, a capacity limit, and a point-in-time reading answer different questions.
For a failed request, gather the endpoint, method, time, status, and any request or operation ID. Compare these with the instance's health and recent activity. Keep credentials and private payloads out of exported diagnostics.
Confirm that a measurement is live
The audited source includes placeholder MCP statistics and a region-health fallback that previously returned healthy values when upstream checks failed. Treat unavailable telemetry as unknown. Do not turn a default value into evidence of uptime, zero errors, or unused capacity.
Use instance operations for engine health and metrics. Use platform billing and usage for account-side metering and integrations.