Zenodo MCP provides two complementary implementations to interact with Zenodo records via the Model Context Protocol. The MCP API implementation exposes a FastAPI-based service that can be integrated with LangChain, LangGraph, and OpenAI-compatible clients, enabling search, retrieval, metadata access, citations, and file downloads as MCP tools. The MCP SDK Core offers a Python-based MCP server designed for direct integration with development environments like Cursor IDE, providing MCP-compliant access to Zenodo data without the need for extra middleware. Together, they let you either embed Zenodo capabilities into AI workflows (via the API) or run a lightweight, environment-embedded MCP service (via the SDK Core).

Prerequisites:

• Python 3.9+ and pip
• Git
• (Optional) Node.js if you plan to use additional tooling, but not required for the Python implementations

Install and set up both implementations:

1. Clone the repository

   git clone https://github.com/yourusername/zenodo-mcp.git

2. Install and run the MCP SDK Core (Cursor IDE integration)

   cd zenodo-mcp/mcp_sdk_core python -m venv venv source venv/bin/activate # on Windows use venv\Scripts\activate pip install -r requirements.txt

   Follow the SDK Core README for Cursor-specific configuration (mcp.json) and integration steps.

3. Install and run the MCP API (LangChain/LangGraph/OpenAI-compatible API)

   cd zenodo-mcp/mcp_api python -m venv venv source venv/bin/activate pip install -r requirements.txt

   Set up environment variables (create .env from .env.example and provide your Zenodo API token)

   cp .env.example .env nano .env # or your preferred editor

   Run the API server

   uvicorn server.main:app --host 0.0.0.0 --port 8000

Notes:

• If you intend to expose the API to other services, you may run behind a reverse proxy or containerize the application.
• Ensure your Zenodo API token has appropriate scopes for the operations you plan to perform.

Tips and common issues:

• Ensure Python virtual environments are activated when installing dependencies for each component.
• If you encounter port conflicts, change the port in the uvicorn command (e.g., --port 8001).
• For the API, store sensitive tokens in a dedicated environment file (.env) and do not commit it to version control.
• The SDK Core is intended for Cursor IDE integration; refer to the mcp_sdk_core/README.md for Cursor-specific configuration nuances.
• If you must run in Docker, convert the python-based components into appropriate container images and adjust the mcp_config accordingly (docker command format).