Documentation Grafana Cloud Observe and act Agent Observability Observe and troubleshoot

Observe and troubleshoot

Use analytics to review activity across your agents, then explore agents and individual conversations to investigate what happened.

1. Monitor agents with analytics to monitor activity, performance, quality, and cost.
2. Use the agent catalog to review agents, versions, and prompt changes.
3. Browse and debug conversations to inspect responses, tool calls, failures, and linked traces.
4. Optimize cost and performance to investigate usage and efficiency.
5. Configure your model prices to investigate costs with the prices you pay.

When you find examples you want to test again, refer to Evaluate and improve.

• Monitor agents with analytics
  Monitor agent activity, performance, cost, and quality using the Agent Observability analytics dashboards.
• Use the agent catalog
  Monitor agent versions, track tool and prompt changes, and compare performance across agents in the Agent Observability agent catalog.
• Browse and debug conversations
  Search, filter, and drill into conversations to understand what your agents did, where they failed, and how they performed.
• Optimize cost and performance
  Use Agent Observability data to reduce LLM costs, improve cache efficiency, and tune agent performance.
• Configure your model prices
  Set the prices you pay for LLM models so Agent Observability cost figures match your contract.

Next steps

• Evaluate and improve agent quality
• Add deeper instrumentation