DECIMER MCP Server

Connects Claude to DECIMER's chemical structure recognition engine via a FastAPI backend. Exposes two tools: server_health to ping the API and analyze_chemical_image to send a base64-encoded image and get back SMILES notation plus classifier confidence scores. You run the DECIMER API separately (Docker image or source available), then point this MCP adapter at it. Useful when you're working with chemical diagrams in papers or lab notebooks and need programmatic access to structure data. Returns structured JSON with the SMILES string, classifier decision, and fallback reasons when the image isn't recognized as a chemical structure. The server is a thin transport layer, not a standalone model runtime.