Tracing (Agents)

Tracing is the practice of recording the full sequence of steps in a single agent run — each model call, tool invocation, input, and output — as a linked trace for inspection.

A trace makes an agent’s reasoning and actions visible end-to-end, which is what lets engineers find where a multi-step run went wrong.

Tracing is a core part of observability for agents and supplies the raw detail behind monitoring and evaluation.