Operating at scale

Watch usage, work, cost, and team activity.

As more agents run, the API gives you a few control surfaces for operations.

What to monitor

SurfaceUse it for
MetricsSessions, users, pull requests, merge requests, searches, and usage trends.
ConsumptionDaily cost and usage by org, user, service user, or session.
Queue statusWhether work is waiting on capacity.
Compute statusWhether the runtime is healthy.
Audit logsWho changed settings, access, and resources.
TagsTeam reporting, filtering, and policy.

Session fields to watch

  • status for completion or failure.
  • cost_usd for model spend.
  • usage for token totals.
  • context for live context occupancy.
  • tags for reporting and workflow grouping.

Team hygiene

  • Use service users for machines.
  • Use narrow scopes.
  • Rotate keys that leave your control.
  • Apply tag policy when reports depend on tags.
  • Review audit logs for sensitive changes.