MCP-MCP Server

MCP-MCP is a Meta-MCP Server that acts as a centralized discovery and provisioning hub for MCP servers. It helps you quickly find and use MCP tools from a large, up-to-date database, so your AI assistants can access the capabilities they need without manual tool hunting or complex setup.

How to use

You interact with MCP-MCP through your MCP client to discover and select MCP servers that match a task. Start a search by asking for a capability, such as weather data, domain checks, or stock information. MCP-MCP returns a list of relevant MCP servers from its aggregated database, helping you pick the right tool and provisioning it for use in your workflow.

How to install

Prerequisites you need before installing MCP-MCP are Python 3.13 or newer and the uv package manager. You may optionally use direnv for automatic environment setup and just for convenient development commands.

Install and run MCP-MCP locally using the following steps.

# Clone the project
git clone https://github.com/wojtyniak/mcp-mcp.git
cd mcp-mcp

# Install dependencies and prepare the environment
uv sync

# Run unit tests (optional during development)
uv run pytest

# Start the MCP-MCP server (stdio mode by default for local clients)
uv run main.py

If you prefer to test directly without local development tooling, you can also run the package using the provided development commands, which include an HTTP mode alternative and auto-reload options.

# HTTP mode (for MCP clients that communicate over HTTP)
uv run main.py --http
# Start in stdio mode for MCP clients like Claude Desktop
uv run main.py

Configuration and usage notes

MCP-MCP provides configuration snippets for clients that can load MCP servers. A typical client configuration for Claude Desktop includes a reference to MCP-MCP as the available MCP server. You can specify how to invoke MCP-MCP from your client, including the transport method and startup commands.

Security and reliability notes

MCP-MCP emphasizes origin validation and secure handling of MCP server data. The database is designed to deduplicate entries across sources to avoid conflicts and ensure consistent results. Regular updates occur to keep the server list fresh and reliable.

Troubleshooting

If you encounter startup issues, verify that you have Python 3.13+ and the uv tool installed. Check that you are running the correct transport mode (stdio for local clients or HTTP for web-based clients) and confirm that the server is reachable on the expected host and port. Review any error messages for hints about missing dependencies or misconfigurations in your MCP client setup.

Notes

MCP-MCP aggregates MCP servers from multiple curated sources and performs intelligent deduplication to present a clean, comprehensive catalog. The system is designed to be updated regularly—every few hours—to reflect new server additions and changes in existing endpoints.

Development and testing (glossary for end users)

For developers or advanced users who want to contribute or validate the onboarding flow, you can run unit and integration tests locally. The project supports a test suite that exercises startup performance and the discovery of thousands of MCP servers.