Agent Teams

Orchestrate collaborative agent teams using Claude Code's experimental agent teams feature.

When to Use

Is the work parallelizable with inter-agent communication needs?
├─ Single focused output? → Use a subagent (Task tool), not a team
├─ 2-3 independent outputs? → Use parallel subagents, not a team
├─ Parallel work + agents need to share findings? → USE A TEAM
├─ Competing hypotheses to test? → USE A TEAM
├─ Multi-file implementation with integration? → USE A TEAM
└─ Multi-lens review (security + perf + correctness)? → USE A TEAM

Don't use teams for: simple delegation, single-file work, sequential tasks, or anything a subagent handles fine.

The Process