mcp-mcp

What is MCP-MCP?

MCP-MCP is a Meta-MCP Server designed to facilitate the discovery and provisioning of MCP (Model Context Protocol) servers for AI assistants. It acts as a "phone book" for MCP servers, aggregating information from various curated sources to provide a centralized database of over a thousand servers. This allows AI assistants to find and utilize specific tools and capabilities they need, without manual searching or complex setup.

How to use MCP-MCP?

MCP-MCP can be installed and run via uvx (recommended) or configured with pipx. Once installed, it needs to be added to the configuration files of AI assistants like Claude Desktop or Claude Code. After configuration, users can ask their AI assistant to discover MCP servers using natural language queries, such as "Find me an MCP server for weather data" or "Search for MCP servers related to stock market data."

Key Features of MCP-MCP

• Multi-source discovery: Aggregates MCP server lists from multiple curated sources.
• Semantic search: Utilizes precomputed embeddings for efficient and relevant search results.
• Configuration string generation: Simplifies the process of configuring AI assistants to use discovered servers.
• FastMCP integration: Ensures quick startup and access to a large database of servers.
• Easy installation: Can be installed and run with a single command using uvx.

Use Cases of MCP-MCP

• AI assistant tool discovery: Enables AI assistants to dynamically find and integrate new tools and capabilities based on their needs.
• Automated tool provisioning: Streamlines the process of setting up and using external tools for AI tasks.
• Enhanced AI capabilities: Allows AI assistants to access a wider range of real-time data and specialized functionalities (e.g., checking domain availability, getting weather data, web scraping).
• Developer productivity: Simplifies the process for developers to integrate and test MCP servers with AI models.

FAQ about MCP-MCP

Q: What problem does MCP-MCP solve? A: MCP-MCP addresses the challenge of AI assistants needing specific tools but struggling with discovery, setup, and integration. It provides a centralized, searchable directory of MCP servers, eliminating the need for manual research and complex configurations.

Q: How many MCP servers can MCP-MCP discover? A: MCP-MCP aggregates information from multiple sources, providing access to over a thousand MCP servers.

Q: Is MCP-MCP difficult to set up? A: No, MCP-MCP is designed for easy installation using uvx and straightforward configuration with AI assistants like Claude Desktop and Claude Code.

Q: Can I contribute to MCP-MCP? A: Yes, contributions are welcome. The project provides clear guidelines for development, testing, and contributing via pull requests.

MCP-MCP: Meta-MCP Server

MCP-MCP is a Meta-MCP Server that acts as a tool discovery and provisioning service for the Model Context Protocol (MCP). When an AI assistant needs a capability that isn't currently available, it can ask MCP-MCP to discover and suggest appropriate MCP servers from a comprehensive database of over a thousand servers aggregated from multiple curated sources.

Think of it as a "phone book" for MCP servers - one tool to find all other tools.

Motivation

Agents Just Wanna Have Tools

• Agents know what they need: AI assistants can clearly articulate requirements like "check domain availability" or "get weather data"
• Web search isn't always enough: Generic search results don't always provide realtime data
• CLI tools require setup: Many tools need complex installation, configuration, and API keys - agents have to repeat this setup every single time they need to complete a task
• MCP servers are scattered: Great tools exist but discovering them requires manual research across GitHub, forums, and documentation

Why make agents (and users) hunt for tools when we can bring the tools to them?

Quick Start

Install via uvx (Recommended)

uvx mcp-mcp

This installs and runs the MCP-MCP server directly via uvx.

Claude Desktop Configuration

Add MCP-MCP to your Claude Desktop configuration file:

Configuration File Location:

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

Configuration:

{
  "mcpServers": {
    "mcp-mcp": {
      "command": "uvx",
      "args": ["mcp-mcp"]
    }
  }
}

Alternative with pipx:

{
  "mcpServers": {
    "mcp-mcp": {
      "command": "mcp-mcp"
    }
  }
}

Claude Code Configuration

Add MCP-MCP to your Claude Code configuration file:

claude mcp add mcp-mcp uvx mcp-mcp

Usage Examples

Once configured, you can ask Claude Desktop to discover MCP servers using natural language:

• "Find me an MCP server for weather data"
• "I need a server for checking domain availability"
• "Search for MCP servers related to stock market data"
• "What MCP servers are available for web scraping?"

Development

Prerequisites

• Python 3.13+
• uv package manager
• direnv (optional, for automatic environment setup)

Setup

# Clone the repository
git clone https://github.com/your-username/mcp-mcp.git
cd mcp-mcp

# Install dependencies
uv sync

# Run tests
uv run pytest

# Run the server
uv run main.py

Development Mode

For development and testing, use HTTP transport (easier to stop with Ctrl+C):

# HTTP mode (accessible at http://localhost:8000)
uv run main.py --http

# Custom host/port
uv run main.py --http --host 0.0.0.0 --port 3000

# STDIO mode (for MCP clients like Claude Desktop)
uv run main.py  # Note: To stop STDIO mode, use Ctrl+D (EOF), not Ctrl+C

Building

# Build package
uv build

# Test local installation
uvx --from ./dist/mcp_mcp-0.1.0-py3-none-any.whl mcp-mcp

Command Line Options

mcp-mcp --help

Option	Description	Default
--transport {stdio,http}	Transport method	stdio
--http	Use HTTP transport	-
--host HOST	Host for HTTP transport	localhost
--port PORT	Port for HTTP transport	8000

Testing

# Run all unit tests (fast, no network)
uv run pytest

# Run specific module tests
uv run pytest db/ -v

# Run GitHub integration tests (optional, requires network)
MCP_MCP_TEST_GITHUB_INTEGRATION=1 uv run pytest db/test_precomputed_data_workflow.py::test_real_github_download -v -s

# Run with coverage
uv run pytest --cov=db

Integration Tests: Set MCP_MCP_TEST_GITHUB_INTEGRATION=1 to test real GitHub downloads and verify the complete first-user onboarding experience. These tests ensure users get fast startup (< 5 seconds) with 1296+ servers.

Roadmap

Current Status: PoC Complete

• ✅ Multi-source discovery (3 curated MCP server lists)
• ✅ Semantic search with precomputed embeddings
• ✅ Configuration string generation
• ✅ FastMCP integration

Next: MVP Foundation

• Multi-source discovery (GitHub API)
• Docker integration for server execution
• MCP protocol proxy
• Server lifecycle management

Future Enhancements

• Private registry support
• Dependency resolution
• Performance monitoring
• Web UI for server management

Contributing

We welcome contributions! Please see our development setup and:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development Guidelines

• Follow Python 3.13+ best practices
• Add tests for new functionality
• Update documentation as needed
• Use semantic commit messages

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Model Context Protocol team at Anthropic
• Open source MCP server maintainers and contributors
• MCP Server Lists:
  • Official MCP Servers
  • Punkpeye's Awesome MCP Servers
  • Appcypher's Awesome MCP Servers

---

Made with ❤️ for the MCP ecosystem