Inter-Agent Messaging - Claude Code

Inter-agent messaging lets teammates talk to each other directly. The lead doesn't have to relay every question.

What it does

Teammates can send messages to named peers or broadcast to the team. Messages appear in the recipient's context on the next turn. This is how a researcher can hand a finding to an implementer, or two implementers can sync on an interface, without funneling everything through the lead.

When to use it

• Tasks that need a brief handoff between peers.
• Collaborative patterns where two agents share an interface contract.
• Reducing lead overhead on trivial coordination.
• Debugging - ask one agent to query another's state.

Gotchas

• Messages add context to both sides. Keep them tight.
• Misuse can create tangled threads - prefer lead-mediated coordination for big decisions.
• Messaging is not a replacement for the shared task list. Tasks are the source of truth.

Official docs: https://code.claude.com/docs/en/agent-teams.md#context-and-communication