# Monitor & debug

See what agents said, what they cost and who changed what, and find out why a run went wrong.

Three records answer most questions about a running system: **transcripts** (what an agent and a person said), **usage** (what it cost and how long it took) and the **audit log** (who did what, and what the gateway allowed or refused). Each one links to the others through the conversation id and the request's correlation id.