mcp -toolhouse

This MCP server connects MCP clients to Toolhouse tools via Groq's API. It enables clients (for example Claude Desktop App users) to access a curated set of tools defined in a Toolhouse bundle, such as web scraping, memory utilities, or email sending, through the MCP interface. To use it, configure your client with the mcp server entry as shown in the repository's configuration example, providing your Toolhouse and Groq API keys and the bundle name. Run the server locally using the uv runtime (as documented in the repository): it will expose MCP endpoints that the client can query for tool actions and tool results. The included inspector can help you debug MCP interactions by connecting to the running server and presenting a debugging UI in a browser.

Once running, the MCP server will route requests from MCP clients to the tools defined in your Toolhouse bundle. Tools such as Scrape the web, Memory, and Send Email can be included in the bundle and made available to clients through the MCP interface. Clients can request tool invocations, pass in prompts, and receive structured tool results that the MCP protocol expects.

Prerequisites:

• Git
• Python 3.x installed on your system
• Access to the uv runtime as described in Toolhouse MCP documentation (the server is designed to run under uv or uv compatible runtimes)

Installation steps:

1. Clone the repository: git clone https://github.com/toolhouse-community/mcp-server-toolhouse.git cd mcp-server-toolhouse

2. Install and configure the uv runtime (follow Toolhouse/MCP docs for your environment). This project is intended to run with uv/uvx. Ensure you can invoke the runtime from your shell.

3. Create or edit your MCP client configuration to reference the server, using the following sample as a guide (adjust paths and keys):

   { "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", "GROQ_API_KEY": "your_groq_api_key", "TOOLHOUSE_BUNDLE_NAME": "a_bundle_name" } } } }

4. Start the server using the configured command in your environment. Example (adjust path accordingly): uv --directory /path/to/this/repo/ run mcp-server-toolhouse

5. If you use Claude or another MCP client, update its config to point to this server and test a few tool invocations.

Note: If you plan to run in ephemeral environments or via uvx, adapt the command accordingly and ensure environment variables are provided in the runtime environment.

Tips and caveats:

• The server relies on environment variables for API keys. If you prefer, you can place TOOLHOUSE_API_KEY, GROQ_API_KEY, and TOOLHOUSE_BUNDLE_NAME in a configuration file or pass them inline where your client/wrapper injects environment data.
• Debugging MCP interactions is easier with the MCP Inspector. You can launch the inspector against the running uv-based server to view requests, responses, and tool invocations.
• Ensure your Toolhouse bundle contains the tools you intend to expose (e.g., web scraping, memory, email) and that the bundle name matches the TOOLHOUSE_BUNDLE_NAME value.
• When configuring your client, point the mcpServers.mcp-server-toolhouse.* fields to the correct directory and server name so MCP clients can discover and query the server properly.
• If you encounter authentication or API access issues, verify the API keys have the necessary permissions and that the bundle referenced by TOOLHOUSE_BUNDLE_NAME exists in Toolhouse.
• This project is described as not yet configured for ephemeral environments like uvx; for local development, use uv as demonstrated.