Observe agents
Find out what your agents are doing in Orca Agent Engine - session status, token usage, transcripts, and platform-wide health.
Observability in Agent Engine works at two levels. Per agent, a session carries its own status, statistics, token usage, and full event history, so you can answer "what did this run actually do" without external tooling. Across the platform, health and throughput signals span agents, functions, and connectors together.
Monitor sessions and agents
Session status, statistics, token usage, and event history for a single agent.
Platform monitoring
Where to find health, throughput, usage, and logs across the whole deployment.
For traffic-level observability - per-call latency, cost, OpenTelemetry spans, and replayable recordings of model and tool calls - see AI Gateway observability.