Grain MCP Server

A Model Context Protocol server for Grain, a service that records and transcribes meetings.

This server provides integration with Grain through MCP, allowing users to access Grain functionality without needing the enterprise API (which is only available on enterprise subscriptions) or paid integrations like Zapier. The service is based on Playwright for browser automation.

Installation

Manual Installation

Add the Grain MCP server configuration to your MCP client:

{
  "mcpServers": {
    "grain_uvx": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/eadm/grain-mcp-server",
        "grain-mcp-server",
        "--user-data-dir",
        "<absolute-path-to-browser-session-data>"
      ]
    }
  }
}

Replace <absolute-path-to-browser-session-data> with the absolute path where you want to store the browser session data. On first MCP usage, you will need to login to Grain via the browser.

Related MCP server: Low Cost Browsing MCP Server

Components

Tools

1. get_all_meetings: Retrieve all meetings from Grain

   • Returns a list of dictionaries containing meeting information, including:

     • id (string): Meeting ID

     • title (string): Meeting title

     • url (string): URL to access the meeting

     • date (string): Meeting date in ISO format

2. download_meeting_transcript: Download a meeting transcript

   • Required inputs:

     • absolute_save_path (string): The file path where the transcript will be saved

     • meeting_id (string): The unique identifier of the meeting

     • transcription_type (string): The format of the transcript file ("vtt" or "srt")

   • Returns:

     • bool: True if the download was successful, False otherwise

Usage Examples

Some example prompts you can use with your MCP client to interact with Grain:

1. "Show me all my recent meetings" → execute the get_all_meetings tool to retrieve a list of all meetings stored in Grain

2. "Download the transcript for my last team meeting" → use get_all_meetings to find the meeting, then use download_meeting_transcript to download its transcript

Development

1. Install dependencies:

uv sync

2. Run the application:

uv run grain-mcp-server

3. Run with debug mode:

uv run grain-mcp-server --debug

4. Run tests:

uv run pytest