MCPfinder acts as an "App Store for agents," letting a language model search for and install new MCP server capabilities on demand instead of being limited to a fixed, static toolset. It queries a registry of several hundred verified MCP servers and can add or remove server configurations directly in the client application's local configuration.

Use it when an AI agent hits a task it can't complete with its current tools - for example, needing code generation, data analysis, or specialized domain knowledge - and you want it to autonomously discover and install the right MCP server rather than you manually searching and configuring one. Most registry servers run locally via npx or uvx with no environment variables required. Do not use it if you need strict, locked-down control over exactly which tools an agent can access - by design, this lets the agent modify its own local MCP configuration.

• search_mcp_servers: search the MCPfinder registry by keywords or tags.
• get_mcp_server_details: retrieve detailed information about a specific server, including its manifest and installation instructions.
• add_mcp_server_config: add or update an MCP server's configuration in the client application's local configuration.
• remove_mcp_server_config: remove a specific MCP server configuration from the local configuration.

Supports both Node.js (npx) and Python (uvx) packaged servers, and both stdio and HTTP/SSE transports for its own connection.

Interactive setup:

npx -y @mcpfinder/server install

Or connect to the hosted cloud registry over HTTP/SSE:

claude mcp add --transport http mcpfinder https://mcpfinder.dev/mcp

Or run from source:

git clone https://github.com/mcpfinder/server
node index.js

Manual stdio configuration:

{
  "mcpServers": {
    "mcpfinder": {
      "command": "npx",
      "args": ["-y", "@mcpfinder/server"]
    }
  }
}

Or HTTP/SSE:

{
  "mcpServers": {
    "mcpfinder": {
      "url": "https://mcpfinder.dev/mcp",
      "transport": "http"
    }
  }
}

Optional environment variables: MCPFINDER_API_URL (registry API base URL) and MCP_PORT (listen port in HTTP mode).

Agent builders and power users who want their AI assistant to autonomously expand its own toolset mid-conversation instead of manually researching and configuring new MCP servers ahead of time.