🚀 MCP Project Manager

Hierarchical Task Management via Model Context Protocol (MCP)

Transform any MCP-compatible AI assistant into a powerful project manager with automated task breakdown, dependency tracking, and smart workflow management.

Compatible with: Claude Code • Claude Desktop • Cursor • VS Code + Copilot • Continue.dev • Any MCP Client

📖 Documentation

• 📚 Complete User Guide - Comprehensive setup and usage guide
• ⚙️ Tools Reference - Complete MCP tools documentation
• 🎯 AI Prompts - Ready-to-use prompts for task management

✨ Key Features

🎯 Hierarchical Project Management: Ideas → Epics → Tasks with intelligent decomposition
🤖 AI-Powered Task Creation: Automated breakdown using natural language prompts
📊 Smart Dependency Tracking: Automatic task sequencing and blocker detection
⚡ NPX Ready: Zero-installation deployment with npx mcp-project-manager
🖥️ Interactive CLI Dashboard: Real-time project visualization and navigation
🔄 Status Management: Pending → In-Progress → Done with progress tracking
🛡️ Approval Workflow: User control over all AI actions and modifications

🚀 Quick Start

Prerequisites

• Any MCP-compatible tool (see supported tools below)
• Node.js 18+ (includes NPX automatically)

📁 Data Storage

Important: Understanding where your task data is stored:

• Without MCP_TASK_DATA_DIR: Data stored in current project directory where AI is working
• With MCP_TASK_DATA_DIR: Data stored in specified absolute path

When to use each approach:

✅ Use current directory (no MCP_TASK_DATA_DIR):

• Working on a specific project
• Want task data alongside your code
• Using Cursor, VS Code, or other project-based tools

✅ Use custom directory (MCP_TASK_DATA_DIR):

• Required for Claude Desktop (read-only environment)
• Managing multiple projects centrally
• Want shared task management location

🚀 Installation Methods

Method 1: NPX (Recommended)

Simply use npx mcp-project-manager in your MCP client configuration - NPX will handle downloading and running the server for you.

Method 2: Clone Repository

For development or customization:

git clone https://github.com/croffasia/mcp-project-manager.git
cd mcp-project-manager
npm install
npm run build

Configuration

This MCP server works with any MCP-compatible tool. Here are the most popular options:

Supported Tools

🔷 Claude Code (Recommended)

Add to your Claude Code configuration (CLAUDE.md or project settings):

{
    "mcpServers": {
        "pm": {
            "command": "npx",
            "args": ["mcp-project-manager"]
        }
    }
}

Add "env": {"MCP_TASK_DATA_DIR": "/path/to/your/project"} if you need custom data directory.

🟦 Claude Desktop

MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

⚠️ Required for Claude Desktop: You must specify a writable directory using MCP_TASK_DATA_DIR because Claude Desktop runs in a read-only environment.

Option 1: Project-specific (Recommended) Store task data in your project folder:

{
    "mcpServers": {
        "pm": {
            "command": "npx",
            "args": ["mcp-project-manager"],
            "env": {
                "MCP_TASK_DATA_DIR": "/Users/yourusername/projects/my-app"
            }
        }
    }
}

Option 2: Shared location Store all task data in one central location:

{
    "mcpServers": {
        "pm": {
            "command": "npx",
            "args": ["mcp-project-manager"],
            "env": {
                "MCP_TASK_DATA_DIR": "/Users/yourusername/Documents/task-management"
            }
        }
    }
}

Examples of good directories:

• Project folder: /Users/yourusername/projects/my-app
• Central location: /Users/yourusername/Documents/task-management
• Windows project: C:\\Users\\yourusername\\projects\\my-app
• Windows central: C:\\Users\\yourusername\\Documents\\task-management

🔵 Cursor

Add to Cursor MCP configuration:

Global Configuration: Create ~/.cursor/mcp.json:

{
    "mcpServers": {
        "pm": {
            "command": "npx",
            "args": ["mcp-project-manager"]
        }
    }
}

Project-specific Configuration: Create .cursor/mcp.json in your project:

{
    "mcpServers": {
        "pm": {
            "command": "npx",
            "args": ["mcp-project-manager"],
            "env": {
                "MCP_TASK_DATA_DIR": "/path/to/your/project"
            }
        }
    }
}

🟢 VS Code (Preview)

With GitHub Copilot extension and Agent mode enabled:

Workspace Configuration: Create .vscode/mcp.json in your project:

{
    "servers": {
        "pm": {
            "type": "stdio",
            "command": "npx",
            "args": ["mcp-project-manager"]
        }
    }
}

With custom data directory:

{
    "servers": {
        "pm": {
            "type": "stdio",
            "command": "npx",
            "args": ["mcp-project-manager"],
            "env": {
                "MCP_TASK_DATA_DIR": "/path/to/your/project"
            }
        }
    }
}

🟡 Continue.dev

Note: MCP can only be used in agent mode in Continue.dev.

Create .continue/mcpServers/pm.yaml in your project:

name: MCP Project Manager
version: 0.1.3
schema: v1
mcpServers:
    - name: Project Manager
      command: npx
      args:
          - 'mcp-project-manager'

With custom data directory:

name: MCP Project Manager
version: 0.1.3
schema: v1
mcpServers:
    - name: Project Manager
      command: npx
      args:
          - 'mcp-project-manager'
      env:
          MCP_TASK_DATA_DIR: '/path/to/your/project'

🟣 Other MCP-Compatible Tools

• Windsurf - AI-powered development environment
• Zed - High-performance code editor with AI features
• Cline - AI coding assistant
• Any custom MCP client following the protocol

Local Development Setup

If you cloned the repository (Method 2), update your MCP client config:

For Claude Code/Desktop (using built version):

{
    "mcpServers": {
        "pm": {
            "command": "node",
            "args": ["/absolute/path/to/mcp-project-manager/dist/index.js"]
        }
    }
}

For Cursor (create ~/.cursor/mcp.json or .cursor/mcp.json):

{
    "mcpServers": {
        "pm": {
            "command": "node",
            "args": ["/absolute/path/to/mcp-project-manager/dist/index.js"]
        }
    }
}

For VS Code (create .vscode/mcp.json):

{
    "servers": {
        "pm": {
            "type": "stdio",
            "command": "node",
            "args": ["/absolute/path/to/mcp-project-manager/dist/index.js"]
        }
    }
}

For Continue.dev (create .continue/mcpServers/pm.yaml):

name: MCP Project Manager
version: 0.1.3
schema: v1
mcpServers:
    - name: Project Manager
      command: node
      args:
          - '/absolute/path/to/mcp-project-manager/dist/index.js'

For Developers

Prerequisites

• Node.js 18+
• TypeScript
• MCP SDK knowledge
• Any MCP-compatible development environment

Installation

git clone https://github.com/croffasia/mcp-project-manager.git
cd mcp-project-manager
npm install

Development

npm run dev    # Watch mode
npm run build  # Build for production
npm run test   # Run tests

🖥️ CLI Dashboard

Interactive terminal-based interface for task management with real-time updates and hierarchical navigation. Run with npx mcp-project-manager --dashboard.

Advanced Configuration

Database Storage

Data is stored in SQLite database (database.sqlite) within the data directory.

Global Project Registry

The MCP Project Manager automatically maintains a global registry of all projects where it's used. This registry is stored in:

• MacOS/Linux: ~/.mcp-project-manager/registry.json
• Windows: %USERPROFILE%\.mcp-project-manager\registry.json

The registry tracks:

• Project names and absolute paths
• Database locations for each project
• Usage statistics (last used, usage count)
• Project creation dates

This happens automatically - no configuration needed. The registry enables future features like project switching and usage analytics.

Environment Variables

• MCP_TASK_DATA_DIR - Custom directory for storing task data
  • Required for Claude Desktop (read-only environment)
  • Optional for other tools (defaults to current directory)

🌟 Contributing & Support

🐛 Report issues on our GitHub Issues
🤝 Contribute by submitting pull requests
📖 Share with other developers who use MCP-compatible AI tools

📄 License

MIT © 2025 Andrii Poluosmak

🌟 Found this project useful?

If MCP Project Manager helped improve your AI-powered development workflow, please consider giving it a star! Your support helps us reach more developers and continue improving the project.

Thank you for your support! 🙏