Skip to main content

Monitoring and analysis

Review the inference that flows through the LLM gateway. Every gateway call is logged to the Agent Monitor as part of a unified lifecycle with your other agent events, so inference sits alongside tool calls rather than in a separate silo.

Individual request logs​

LLM gateway calls appear in the Agent Monitor's Live activity view. For each call you can see:

  • Latency, including the gateway's own overhead
  • Usage and cost for the call
  • Provider and model selected
  • Prompt and tool-result data, when logging is enabled
Request details panel for an LLM gateway inference call, showing provider, model, latency, and token usage

Usage dashboard​

The same Agent activity page has a Usage tab that gives high-level data on usage across your policies, for a quick read on volume and spend without opening individual calls. See Agent Monitor usage for the broader usage reference.

LLM gateway Usage dashboard showing spend and request volume charts across policies

Next steps​