Protein

Protein MCP Server provides fast, multi-protocol access to protein structural data, enabling researchers to search, retrieve, and download protein structures with ease. It supports streamlined core tools and multiple transport modes to fit your workflow.

How to use

You interact with Protein MCP Server through an MCP client or integrated tools to search for protein structures, fetch data packages, and download structure files. Start by running the local server in stdio mode for fast, script-friendly use, or switch to HTTP or SSE modes when you need remote access or real-time streaming. Use the search tool to discover structures by keywords or PDB IDs, fetch complete data packages for a structure, and download structure files in your preferred format.

How to install

Prerequisites you need on your machine are Python 3.10+ and a runtime like UVX for running MCP servers. You also need a compatible shell to execute commands.

# Option 1: run directly with the runtime (recommended for local development)
uvx protein-mcp

# Option 2: run in HTTP mode (exposes a remote API on port 8080)
uvx protein-mcp --transport http --port 8080

# Option 3: run in SSE mode (real-time streaming on port 8080)
uvx protein-mcp --transport sse --port 8080

# View all available options
uvx protein-mcp --help

Additional sections

Configuration and runtime options let you tailor the server to your workflow. You can run a local stdio-based MCP server, or enable HTTP or SSE access for integration with clients and dashboards.

Configuration and environment

When running in stdio mode, the server can be started with the core command uvx and the argument protein-mcp. You can also configure environment variables to adjust logging, cache behavior, and API timeouts.

Example runtime configuration

{
  "mcpServers": {
    "protein-mcp": {
      "command": "uvx",
      "args": ["protein-mcp"],
      "env": {
        "PROTEIN_MCP_LOG_LEVEL": "debug",
        "RCSB_API_TIMEOUT": "60",
        "PROTEIN_MCP_CACHE_DIR": "~/.cache/protein-mcp",
        "PROTEIN_MCP_CACHE_DURATION": "3600",
        "PROTEIN_MCP_MAX_RETRIES": "3"
      }
    }
  }
}

Available tools

find_protein_structures_tool

Smart search tool that supports keyword, category, pdb_id, and result limits to discover protein structures.

get_protein_data_tool

Data retrieval tool that can fetch basic information, sequences, structures, and all data types for a given PDB ID and optional chain.

download_structure_tool

File download tool that supports PDB, mmCIF, and MMTF formats and can save files locally or return content.