What is Observability?

Logging and tracing everything an agent does — every step, tool call, token, and cost — so you can see what actually happened, not just the final answer.

How it works

Agents fail silently: a loop that spins, a tool that returns garbage, costs that spike. Without traces you're debugging blind. Observability turns 'it felt slow and expensive' into a specific step you can fix.

Example

In practiceYou trace a run and see the agent called the same search tool nine times with near-identical queries — a loop bug costing tokens.

What's the minimum I should log?

Every iteration's decision, every tool call with inputs/outputs, token counts, and cost. If you can't replay a run, you can't debug it.

Go deeper: Deploy Your First AI Agent

The complete, step-by-step guide — every command, config, and the reasoning behind it.

Get the guide