pydata-london-2025

This MCP server is a Python-based service that exposes a Multi-Agent Collaboration (MCP) workflow for analyzing clinical CSV data using the Polars backend. It enables agents to reason over medical data and generate predictions or insights by coordinating AI providers. To run it, install the required Python dependencies, ensure your API keys are set, and start the Polars-backed MCP server. Once running, you can integrate this MCP server with your orchestration tools to query CSV data, trigger AI-powered analyses, and retrieve results from the MCP graph.

The server intentionally supports two data backends: Polars (recommended for performance on large CSV datasets) and a Pandas-based alternative. By launching the Polars variant you gain faster data access and memory efficiency for tabular clinical data, which is useful in healthcare use cases like lab data interpretation, risk scoring, and outcome prediction. When using this MCP server, you interact with the server via its Python entry point and the generated MCP graph to coordinate multiple AI models across providers.

Prerequisites\n- Python 3.8+ (recommended)\n- pip (comes with Python)\n- Access to required API keys (OpenAI, Anthropic) for the IntelliNode/MCP workflow\n- Internet access to install dependencies and download model providers\n\nInstallation steps\n1) Create and activate a virtual environment (recommended)\n - Python 3.x: python -m venv venv\n - On macOS/Linux: source venv/bin/activate\n - On Windows: venv\Scripts\activate\n\n2) Install base IntelliNode package with MCP support\n - pip install "intelli[mcp]"\n\n3) Install project requirements\n - pip install -r requirements.txt\n\n4) Configure environment variables\n - Create a .env file in the project root with:\n OPENAI_API_KEY=your_openai_api_key_here\n ANTHROPIC_API_KEY=your_anthropic_api_key_here\n\n5) Run the MCP server (Polars backend)\n - cd mcp_server\n - python eicu_mcp_server_polars.py\n\n6) (Optional) Run the Pandas-based server instead\n - python eicu_mcp_server.py\n

Tips and caveats:\n- Ensure the required API keys are stored securely (do not commit .env files to version control).\n- The Polars-backed server is optimized for large CSV datasets; if you encounter memory constraints, consider using the Pandas data provider as an alternative.\n- If the server fails to start due to missing dependencies, re-check that requirements.txt is installed and that your Python environment is active.\n- The MCP workflow relies on multiple AI providers; you can customize orchestration by adjusting the MCP graph and provider configurations within the IntelliNode framework.\n- Common issues include API key expiration, quota limits, and version mismatches between dependencies. Refer to the repository's notes and provider docs for troubleshooting.\n- Environment variables can be extended to include additional providers or data sources as needed by your use case.