mcpcute

MCP aggregator - aggregate multiple MCPs behind a single interface to reduce context pollution for AI agents.

Instead of exposing 20+ MCP tools directly to your AI agent, mcpcute provides just 3 tools:

1. search_tools - Search/list available tools across all aggregated MCPs (returns names only)
2. get_tool_details - Get detailed schema and description for a specific tool
3. execute_tool - Execute a tool with the given arguments

Installation

bun install

Configuration

Create a mcpcute.config.json file (or set MCPCUTE_CONFIG env var):

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
    },
    "fetch": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-fetch"]
    }
  }
}

See mcpcute.config.example.json for a full example.

Usage

As an MCP server

bun run start

In Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "mcpcute": {
      "command": "bun",
      "args": ["run", "/path/to/mcpcute/src/index.ts"],
      "env": {
        "MCPCUTE_CONFIG": "/path/to/mcpcute.config.json"
      }
    }
  }
}

How it works

1. mcpcute connects to all configured MCP servers on startup
2. It caches the list of available tools from each server
3. When an AI agent needs a tool, it first searches with search_tools
4. Then it gets the schema with get_tool_details
5. Finally, it executes with execute_tool

This reduces the initial context from potentially hundreds of tool schemas to just 3 simple tools.

License

MIT