Orchestration Trace
A trace that records coordination events in a multi-agent system, such as spawning sub-agents, delegating work, communicating, aggregating results, and stopping.
Plain English
It is the map of how the agent team actually worked, not only a list of model calls.
Example
A research supervisor's trace shows which specialist agents were spawned, which sources each used, and why the supervisor stopped the run.
Why it matters
Multi-agent systems fail in routing and coordination, so teams need traces that expose those decisions for debugging, evals, and optimization.