Skip to content

P01: Agent Trace ​

What You Do ​

Run one narrow task through Agent Canvas, then read the agent trace. Count tool calls, inspect observations, and save the trace as your baseline.

Harness Mechanism ​

Agent Server typed events, persisted conversation events, and the Canvas trace viewer.

Open First ​

Keep ​

A baseline trace and a trace-reading checklist. This becomes the comparison point for every later project.

Built as a friendly front door for the runnable OpenHands harness lab.