Toolhouse MCP Server

This MCP server allows you to connect MCP clients with Toolhouse's tools. Built on top of Toolhouse and Groq's API - for fast inference.

The Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools. Whether you’re building an AI-powered IDE, enhancing a chat interface, or creating custom AI workflows, MCP provides a standardized way to connect LLMs with the context they need.

Features

• Allows compatible MCP Clients (i.e Claude Desktop App) to access a vast library of tools to enhance their capabilities

Configuration

Setup Environment Variables

1. Toolhouse API Key: You need a Toolhouse API Key to access the Toolhouse platform.

   • Sign up at Toolhouse and create an account, it's FREE to get started.
   • Obtain your API key from the Toolhouse dashboard.

2. Toolhouse Bundle: You need to create a bundle, essentially a group of tools you can assemble from the ones available on the Toolhouse platform.

   • Navigate to Toolhouse Bundles and create a bundle with any name i.e. mcp-toolhouse
   • Add the tools that you want to use on your client i.e. Scrape the web, Memory, Send Email
   • Ensure your bundle looks good (it auto saves)

3. (Optional) Set these environment variables if you prefer not having them in the configuration:

   export TOOLHOUSE_API_KEY="your_toolhouse_api_key"
   export TOOLHOUSE_BUNDLE_NAME="your_bundle_name"
   

Starting the server

Add this server to your client's configuration. For example on Claude's desktop app navigate to the folder and manually change the settings file called claude_desktop_config.json

On MacOS:

~/Library/Application\ Support/Claude/claude_desktop_config.json

On Windows:

%APPDATA%/Claude/claude_desktop_config.json

Modify the configuration file to look like this:

With UVX

{
  "mcpServers": {
    "mcp-server-toolhouse": {
      "command": "uvx",
      "args": ["mcp-server-toolhouse"],
      "env": {
        "TOOLHOUSE_API_KEY": "your_toolhouse_api_key",
        "TOOLHOUSE_BUNDLE_NAME": "a_bundle_name"
      }
    }
  }
}

With UV

{
  "mcpServers": {
    "mcp-server-toolhouse": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/this/folder/mcp-server-toolhouse",
        "run",
        "mcp-server-toolhouse"
      ],
      "env": {
        "TOOLHOUSE_API_KEY": "your_toolhouse_api_key",
        "TOOLHOUSE_BUNDLE_NAME": "a_bundle_name"
      }
    }
  }
}

Run this project locally

This project is not yet configured for ephemeral environments like uvx. Run the project locally by cloning the repository:

git clone https://github.com/toolhouse-community/mcp-server-toolhouse.git

Add this tool as an MCP server.

On MacOS:

~/Library/Application\ Support/Claude/claude_desktop_config.json

On Windows:

%APPDATA%/Claude/claude_desktop_config.json

Modify the configuration file to include:

"toolhouse": {
    "command": "uv",
    "args": [
        "--directory",
        "/path/to/this/repo/",
        "run",
        "mcp-server-toolhouse"
    ],
    "env": {
        "TOOLHOUSE_API_KEY": "your_toolhouse_api_key",
        "TOOLHOUSE_BUNDLE_NAME": "a_bundle_name"
    }
}

TODO

Future improvements include:

• Adding test coverage for all modules
• Extending API support for enhanced tool configurations

Debugging

Since MCP servers run over stdio, debugging can be challenging. For the best debugging experience, use the MCP Inspector.

Launch the Inspector via npm:

npx @modelcontextprotocol/inspector uv --directory /path/to/toolhouse_mcp run toolhouse-mcp

The Inspector will display a URL to access debugging tools in your browser.