danieldoderlein/llm-bus

What it does

LLM Bus is a multiagent coordination bus built on the Model Context Protocol (MCP) that enables seamless collaboration between multiple AI coding agents. It prevents agents from duplicating work and colliding by providing a centralized coordination layer with atomic gapfree claims, advisory file leases, a shared event ledger, presence tracking, prose handoffs, and a task graph visualization.

How to set up

LLM Bus operates as an MCP server and can be integrated into your Claude, Cursor, or Codex environment. Access the GitHub repository at danieldoderlein/llmbus to review installation instructions. The server can be deployed remotely to coordinate agents across different machines and environments. Detailed setup steps should be referenced in the project's documentation.

Key features

• Atomic gapfree claims system for task assignment
• Advisory file leases to prevent concurrent modifications
• Shared event ledger for coordination history
• Presence tracking across multiple agents
• Prose handoffs for natural agent-to-agent communication
• Task graph visualization and management
• Remote deployment capability
• Support for Claude Code, Cursor, and Codex agents