Agent Monitoring (Beta)

Agent Monitoring is now available in Beta. Agent Monitoring gives developers and AI engineers full visibility into how their agents behave across every stage, from development to production. Plus, it's configurable per agent, giving you total control to choose which ones you monitor.

Once enabled for an agent, you’ll be able to see:

• Input and output tokens for every agent call, with historical usage trends over time
• Two new quality signals, relevance and safety, scored automatically on your agent traffic alongside your usage data
• The full input and output of every LLM call, plus which tools the agent used to get there

Together, this means you can inspect exactly what happened on any past execution, and track whether your agents are getting slower, more costly, or less accurate before your users notice.

This feature is fully integrated into the Analytics console, where you can now view detailed agent information.