agent-teams

# Agent Teams Coordinate multiple Claude Code instances working together. One session acts as team lead, spawning and coordinating teammates that work independently in their own context windows and communicate directly with each other. ## Quick Setup ### 1. Enable the feature (experimental, disabled by default) Add to `settings.json`: ```json { "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" } } ``` Or set the environment variable directly: ```bash export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 ``` ### 2. Start a team Tell Claude to create an agent team with a natural language prompt describing the task and team structure: ``` Create an agent team to review PR #142. Spawn three reviewers: - One focused on security implications - One checking performance impact - One validating test coverage ``` Claude creates the team, spawns teammates, coordinates work, and synthesizes results. ## When to Use Agent Teams vs Subagents | Criteria | Use Subagents | Use Agent Teams | |----------|--------------|-----------------| | Workers need to talk to each other | No | Yes | | Task is focused, result-only matters | Yes | No | | Complex work requiring discussion | No | Yes | | Token budget is tight | Yes (lower cost) | No (higher cost) | | Workers need shared task list | No | Yes | Best use cases for agent teams: - **Research and review**: parallel investigation of different aspects - **New modules/features**: each teammate owns a separate piece - **Debugging competin...