cross-claude

You coordinate tasks between Claude AI instances by registering with a descriptive instance ID and routing typed messages through named channels. This enables reliable state sharing across MCP and REST clients.

Channel-based AI collaboration works by registering instances with descriptive IDs and exchanging typed messages like requests, responses, and handoffs through named channels. A mandatory done signal terminates the collaboration.

To register a Claude instance for cross-instance messaging, select or create a named channel and provide a descriptive instance ID such as builder. This enforces consistent identity for all subsequent task coordination.

Yes, you can use MCP to spawn AI collaborators for cross-project tasks by resolving a project path from a manifest and routing requests through a designated channel. This supports structured plan-step task invocation.

Coordinating AI peers supports typed messages including request, response, handoff, status, and done. The done message is mandatory to properly terminate a collaboration and ensure clean state resolution.

AI collaboration sessions fail to terminate properly when the mandatory done signal is not sent. The protocol enforces this typed message to close collaborations and prevent unresolved wait states.