protein_structure_MCP_server

The Protein Structure Server is an MCP (Model Context Protocol) service that helps you search for human proteins by name, retrieve UniProt accession numbers, and obtain detailed protein structure information by integrating UniProt metadata with AlphaFold predictions. It exposes three core tools: search_proteins, get_protein_structure, and get_uniprot_id. You can query these tools through MCP-compatible clients (like Claude Desktop or the MCP Inspector) to discover proteins, convert names to UniProt IDs, and fetch AlphaFold-linked structure data.

To use the server with Claude Desktop, configure the embedded MCP server by pointing to your local project directory with uv. The server will expose the MCP tools you can call from Claude to search for proteins, fetch structural data, and access AlphaFold links. The available tools let you search by protein name (prioritizing human proteins), convert names to UniProt IDs, and retrieve AlphaFold structure information, providing direct links to 3D structure predictions.

For testing and debugging, you can also run the MCP Inspector to interact with the server in a web UI. This helps verify that search_proteins, get_protein_structure, and get_uniprot_id return the expected data before integrating with Claude Desktop.

Prerequisites:

• Python 3.13 or higher
• uv package manager
• Node.js

Step-by-step installation:

1. Clone the repository:

git clone https://github.com/soffelcopter/protein_structure_MCP_server.git
cd protein_structure

2. Install and set up uv (if not installed):

# uv is the recommended package manager as per the project docs
# install instructions may vary by system; ensure it's available on your PATH

3. Install dependencies (via uv):

uv sync

4. Configure your local path in Claude Desktop or prepare to run the MCP Inspector with the server folder. You can start the inspector with:

# Navigate to your server directory
cd protein_structure

# Start the MCP Inspector (example, assumes npm is available for the inspector package)
npx @modelcontextprotocol/inspector uv run python protein_structure_server.py

5. Run the server locally:

# Ensure you point to your actual project path when configuring Claude Desktop
uv run --directory /absolute/path/to/your/protein_structure python protein_structure_server.py

Notes:

• Ensure Python 3.13+ is installed and available as python in your PATH.
• Replace the placeholder path with the actual absolute path to your project when configuring clients.

Tips and common issues:

• Ensure the directory path in the mcp_config matches your local project location exactly.
• When using Claude Desktop, restart the app after adding the server configuration.
• The server relies on UniProt REST API and the AlphaFold Database API; network access is required.
• If you encounter import or dependency errors, re-run uv sync to resolve package prerequisites.
• For debugging, the MCP Inspector provides a convenient UI to test search_proteins, get_protein_structure, and get_uniprot_id before integration.
• Consider setting environment variables for API endpoints or rate limiting if you deploy in a restricted network.
• If the AlphaFold links or UniProt data fail to load, check the external service status and adjust timeouts or retries accordingly.