CoordMCP - Multi-Agent Code Coordination Server

CoordMCP is a coordination server that helps multiple AI coding agents work together on the same project without conflicts.

Why CoordMCP?

When you use AI coding assistants (OpenCode, Cursor, Claude Code, Windsurf) on a project:

• Lost decisions - The AI forgets what was decided in previous sessions
• Inconsistent choices - Different sessions make different architectural decisions
• No coordination - Multiple AI agents don't know what each other is doing
• No history - There's no record of why certain decisions were made

CoordMCP solves this by giving your AI agents a shared brain that persists across sessions.

How It Works

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│     YOU     │────▶│  AI AGENT   │────▶│  CoordMCP   │
│             │     │             │     │   Server    │
└─────────────┘     └─────────────┘     └──────┬──────┘
                                               │
                                               ▼
                                    ┌─────────────────┐
                                    │  Shared Memory  │
                                    │  • Decisions    │
                                    │  • Tech Stack   │
                                    │  • File Locks   │
                                    └─────────────────┘

You just talk to your AI agent normally. CoordMCP works automatically in the background:

• Remembers decisions across sessions
• Prevents file conflicts between agents
• Provides architecture recommendations
• Tracks all changes

Example

You say:

"Create a todo app with React and FastAPI"

CoordMCP automatically:

1. Discovers or creates the project
2. Registers your AI agent
3. Locks files before editing
4. Records "Use React" and "Use FastAPI" decisions
5. Tracks all created/modified files
6. Unlocks files when done

Next session: Your AI remembers you're using React and FastAPI.

Quick Start

Install

pip install coordmcp
coordmcp --version

Configure Your Agent

Option 1: Using coordmcp CLI (recommended)

For most agents, add to your config file:

{
  "mcpServers": {
    "coordmcp": {
      "command": "coordmcp",
      "args": [],
      "env": {
        "COORDMCP_LOG_LEVEL": "INFO"
      }
    }
  }
}

Option 2: Using Python module

{
  "mcpServers": {
    "coordmcp": {
      "command": "python",
      "args": ["-m", "coordmcp"],
      "env": {
        "COORDMCP_LOG_LEVEL": "INFO"
      }
    }
  }
}

See integrations for specific setup instructions for each agent.

Test It

Restart your AI agent and say:

"What CoordMCP tools are available?"

Documentation

Audience	Start Here
End Users	User Guide
Developers	API Reference
Contributors	Contributor Guide
Architecture Decisions	ADRs

User Guide

• What is CoordMCP? - Overview and features
• Installation - Install and configure
• How It Works - Behind the scenes

Integrations

• OpenCode
• Cursor
• Claude Code
• Windsurf
• Antigravity

Developer Guide

• API Reference - All 52 tools
• Data Models - Data structures
• Examples - Usage examples

Contributor Guide

• Architecture - System design
• Development Setup - Dev environment
• Testing - Run and write tests
• Extending - Add new features

Reference

• Troubleshooting - Common issues
• Configuration - All options
• Architecture Decision Records - Design decisions

Features

Long-Term Memory

Your AI agent remembers decisions across sessions. If you chose React last week, it knows this week.

Multi-Agent Coordination

Multiple AI agents can work on the same project without conflicts through file locking.

Architecture Guidance

Design pattern recommendations without expensive LLM calls. 9 patterns available: MVC, Repository, Service, Factory, Observer, Adapter, Strategy, Decorator, CRUD.

Task Management

Create, assign, and track tasks across agents. Support for task dependencies, priorities, and completion tracking.

Agent Messaging

Enable communication between agents with direct messages and broadcast capabilities.

Health Dashboard

Monitor project health with comprehensive dashboards showing task progress, agent activity, and actionable recommendations.

Zero LLM Costs

All architectural analysis is rule-based - no external API calls needed.

Development

git clone https://github.com/yourusername/coordmcp.git
cd coordmcp
pip install -e ".[dev]"
python -m pytest src/tests/ -v

License

MIT License - see LICENSE.