MCP-MCP

MCP-MCP is an MCP server that acts as a discovery and provisioning directory for other Model Context Protocol servers. It connects AI assistants directly to an aggregated catalog of thousands of community and official MCP tools sourced from curated repositories, including the official modelcontextprotocol collection, Punkpeye, and Appcypher lists. Developers, autonomous agents, and AI practitioners use MCP-MCP to locate specialized tools dynamically when an assistant encounters an unhandled capability request, such as fetching real-time market data, scraping web content, or checking domain availability. Instead of requiring users to manually research repositories and configure tools upfront, MCP-MCP enables language models to perform semantic searches over precomputed embeddings to identify relevant servers with sub-second response times. The underlying registry deduplicates entries and updates continuously from upstream sources. MCP-MCP supports standard STDIO transport for desktop environments as well as HTTP transport for local server workflows, functioning as a central tool registry that helps assistants find and inspect integrations on demand.

Category: AI & LLM Tooling

Tags: discovery, package-management, provisioning, registry

Visit MCP-MCP

How to install and configure MCP-MCP

1. Ensure Python 3.13+ and uv are installed on your machine. 2. Open your Claude Desktop configuration file (~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows). 3. Add mcp-mcp to the mcpServers object: json { "mcpServers": { "mcp-mcp": { "command": "uvx", "args": ["mcp-mcp"] } } } 4. Alternatively, for Claude Code, run: bash claude mcp add mcp-mcp uvx mcp-mcp 5. Restart your client to activate MCP-MCP.

What you can do with MCP-MCP

• Searching for live data tools: Querying the database to locate MCP servers providing weather forecasts, financial market feeds, or cryptocurrency prices. - Identifying web scraping integrations: Finding specialized servers that extract structured content, crawl documentation websites, or bypass web scraping barriers for AI agents. - Discovering domain and networking utilities: Finding tools capable of checking domain availability, querying DNS records, or running network diagnostics on behalf of user prompts. - Exploring community developer tooling: Finding code intelligence, Git management, and database manipulation MCP servers published across open-source GitHub collections without leaving the chat interface.

Key facts

• https://github.com/wojtyniak/mcp-mcp
• AI & LLM Tooling, Developer Tools & Code Intelligence
• discovery, package-management, provisioning, registry

Part of MCP Servers

How do I install MCP-MCP?

You can run MCP-MCP directly using uvx without a manual installation step. Configure it inside your Claude Desktop configuration file under mcpServers with the command uvx and args set to mcp-mcp, or register it in Claude Code using the claude mcp add command.

What can MCP-MCP do?

MCP-MCP acts as a directory meta-server that allows AI assistants to search and locate MCP tools from thousands of curated sources. It provides semantic search powered by precomputed embeddings so assistants can find relevant servers for scraping, financial data, or developer tasks instantly.

Which MCP clients work with MCP-MCP?

MCP-MCP works with Claude Desktop and Claude Code using standard STDIO transport. It can also run via HTTP transport using the http flag for custom MCP client setups, local network setups, or standalone development environments on port 8000.

Where does MCP-MCP get its server catalog?

The catalog aggregates servers from multiple curated repositories, including the official Model Context Protocol servers repository, Punkpeye's awesome-mcp-servers list, and Appcypher's awesome-mcp-servers collection. The database deduplicates entries across sources and automatically refreshes every three hours to include newly released community tools.

Is MCP-MCP open source?

Yes, MCP-MCP is open-source software distributed under the MIT License. The complete source code, test suite, and release history are publicly available on GitHub, allowing developers to inspect, modify, and contribute back to the registry.