Claude Code has an experimental feature called Agent Teams that lets you spawn multiple AI agents that work together on your project. Each agent takes on a different role, and they coordinate with each other automatically.
This guide walks you through enabling the feature and spawning your first team.
Official Documentation
For the full reference, see the Claude Code Agent Teams docs.
Agent Teams DocsEnable Agent Teams
Step 1: Open Claude Code
Open Terminal and start Claude:
claudeStep 2: Ask Claude to Turn On Teams
Paste this into Claude Code:
Turn on the Agent Teams feature for me. Add the experimental agent teams setting to my Claude Code settings file.Claude will update the settings file for you. You don't need to touch it yourself.
Step 3: Restart Claude Code
Close your current Claude Code session and start a new one for the setting to take effect.
Spawn a Team
Step 4: Switch to a Lower Model (Recommended)
Teams use a lot of credits. To save costs while testing, switch to a lighter model:
/model sonnetYou can always switch back to a more powerful model later.
Step 5: Tell Claude to Spawn a Team
Describe the roles you want each agent to take on. For example:
spawn a team of 3 agents named UI designer, UX engineer, and a front end developer to work on my siteClaude will create the team members and you'll see them appear with an @ symbol in front of their names, like @ui-designer, @ux-engineer, and @front-end-developer.
Step 6: Collaborate
From here, you can give instructions and the team members will coordinate with each other to complete the work. Each agent focuses on their role while staying aware of what the others are doing.
Tips
- Start small. Try a team of 2 or 3 agents before scaling up.
- Be specific about roles. The clearer you define each agent's responsibility, the better they collaborate.
- Watch your usage. Teams use credits quickly, especially on larger models.
