SplitMind MCP Agent Communication Server (Redis Edition)

What is SplitMind MCP Agent Communication Server (Redis Edition)?

A persistent, multi-project MCP server that enables real-time communication between AI agents working on parallel tasks in SplitMind. Built with Docker and Redis for reliability and scalability.

How to use SplitMind MCP Agent Communication Server (Redis Edition)?

Clone the repository, make entrypoint.sh executable, and start services with docker-compose up -d. Each Claude Code agent must be configured with the JSON MCP server configuration pointing to the Docker container. Agents then use provided tools like register_agent, heartbeat, add_todo, and query_agent to coordinate.

Key features of SplitMind MCP Agent Communication Server (Redis Edition)

• Multi-project isolated namespaces for different projects
• Persistent state via Redis that survives restarts and crashes
• Per-agent todo list management with priority and status
• Automatic dead agent detection and cleanup via heartbeats
• Docker deployment for easy setup and scalability
• Optional Redis Commander web UI for real-time monitoring

Use cases of SplitMind MCP Agent Communication Server (Redis Edition)

• Parallel AI agents working on separate tasks in the same project
• File coordination with locks to prevent conflicts
• Inter-agent queries and broadcasts for status and help
• Sharing TypeScript interfaces and type definitions across agents
• Tracking progress and task breakdown across all agents

FAQ from SplitMind MCP Agent Communication Server (Redis Edition)

What are the runtime dependencies?

Docker and Docker Compose are required. The server uses a Redis container; optional Redis Commander container for monitoring.

How does dead agent cleanup work?

Agents must call heartbeat() every 30–60 seconds. If a heartbeat is missed, the agent is automatically cleaned up.

How can I monitor the system?

Use Redis Commander at http://localhost:8081 (with debug profile), Docker logs, or direct Redis CLI via docker exec.

How can I secure the deployment in production?

Remove Redis port exposure, add a Redis password via REDIS_URL, and set resource limits on containers.

What data is stored per project?

Each project has a Redis namespace containing: agents hash, heartbeat timestamps, file locks, shared interfaces, agent todo lists, message queues, and a list of recent file changes.