posebusters

This MCP server exposes PoseBusters pose-validation capabilities via a standard MCP interface. It accepts ligand and protein files (typically .sdf for ligands and .pdb for proteins) and returns structured results that indicate pose plausibility, chemical validity, and geometric checks. The server implements GET /mcp/context for discovery and POST /mcp/predict (multipart/form-data) for executing actions such as validate_pose and redocking_validation. To use it in an MCP-enabled workflow or UI, query the context to understand available inputs and outputs, then submit your files along with the chosen action. Results are provided in a schema-friendly JSON format suitable for integration into dashboards, agents, or pipelines.

Prerequisites:\n- Docker (or a container runtime that supports docker run) installed on your machine or server.\n- Internet access to pull the posebusters-mcp-server image, or a local image built from source.\n\nStep 1: Pull the Docker image (recommended)\nbash\ndocker pull posebusters-mcp-server:latest\n\nStep 2: Run the container locally (exposes MCP port 7860 by default)\nbash\ndocker run -p 7860:7860 posebusters-mcp-server:latest\n\nStep 3: Verify the API is reachable\nbash\ncurl -s http://localhost:7860/mcp/context | jq\n\nIf you prefer building and running directly from source, follow the repository’s standard Docker build workflow to create an image named posebusters-mcp-server and run it similarly.

Notes and tips:\n- The server accepts multipart/form-data POST requests to /mcp/predict with actions like validate_pose and redocking_validation. Provide ligand_input and protein_input files; optionally include crystal_input for redocking_validation.\n- The MCP context endpoint (/mcp/context) describes expected inputs, outputs, and schemas. Use it to drive UI generation in MCP-enabled clients.\n- Ensure that the container has access to the input files you upload (bind mounts or in-container file paths when hosting via a platform).\n- If you encounter port conflicts, map to a different host port (e.g., -p 8080:7860).\n- Runtime dependencies inside the container include Gradio/FastAPI and the PoseBusters CLI (bust). The documentation and exact schema definitions live in the repository (schema.py).