The team lead is the coordinator agent in an Agent Teams setup. It decomposes the job, hands out tasks to teammates, and stitches the results into a coherent answer.

What it does

The lead reads the user's goal, plans the decomposition, publishes tasks to the shared task list, and monitors progress. When teammates report back, the lead synthesizes findings, resolves conflicts, and returns the final result. It can also require plan approval before teammates execute.

When to use it

Any multi-agent run - every team has a lead by default.
Complex jobs where a human doesn't want to manage the teammates directly.
Tasks where synthesis is as hard as the underlying work.
Patterns where one "lead + N workers" model is a clean fit.

Gotchas

A bad decomposition wastes everyone's turns. The lead's plan is load-bearing.
Leads can become bottlenecks if they try to micromanage. Trust teammates with scope.
The lead's synthesis is where quality wins or loses. Budget model strength accordingly.

Official docs: https://code.claude.com/docs/en/agent-teams.md#how-claude-starts-agent-teams