Karakeep (Hoarder) MCP Server

This project provides a Mind Control Panel (MCP) compatible server that allows interactions (searching and creating bookmarks) with the Karakeep bookmarking service via its API.

It's designed to be used by Large Language Models (LLMs) or other clients capable of making HTTP requests according to the MCP protocol defined by services like Glama.ai.

Features

• Search Bookmarks: Find existing bookmarks in Karakeep based on a query string.
• Create Bookmarks: Add new bookmarks to Karakeep with a URL and optional title/description.
• MCP Compatible Endpoint: Exposes a single /mcp endpoint for actions.
• Environment Variable Configuration: Securely configure your Hoarder API key.

Prerequisites

• Python 3.8+
• A Karakeep account and API Key (See Karakeep API Docs)
• Access to a hosting platform (e.g., Render, Heroku, Fly.io, VPS) or Docker for deployment.

Setup and Local Development

1. Clone the repository:
   git clone <your-repository-url> cd hoarder-mcp-server
2. Create a virtual environment (recommended):
   python -m venv venv source venv/bin/activate # On Windows use `venv\Scripts\activate`
3. Install dependencies:
   pip install -r requirements.txt
4. Configure Environment Variables:
   • Create a file named .env in the project root.
   • Add your Hoarder API key to it:
     HOARDER_API_KEY="YOUR_HOARDER_API_KEY_HERE"
   • Important: Make sure .env is listed in your .gitignore file to avoid committing secrets.
5. Run the server locally:
   uvicorn main:app --reload --port 8000
   The server will be available at http://127.0.0.1:8000. The --reload flag automatically restarts the server when code changes are detected.

API Usage

The server exposes a single POST endpoint: /mcp.

Request Body (JSON):