Review activity

Every run — your personal assistant's or a managed agent's — leaves a full trail: who or what triggered it, exactly what it was allowed to use, every action it took along the way, and how it ended. You can always answer "what did this agent do, and who let it" without guessing.

What to inspect

• agent service and immutable revision;
• organisation and delegated subject;
• run state, attempt and terminal reason;
• frozen input and capability digests;
• approval requests and action receipts;
• token use, cost and cancellation evidence; and
• ordered conversation events where the caller is authorised to replay them.

Use the authenticated /api/v1/audit surface. Retrieve current filters and pagination from the API reference; the current UI does not expose an audit view.

TIP

Search by run id first. Pod names are replaceable execution details and do not identify the durable product record.