RCSB PDB

You have an MCP Server that lets an AI assistant directly access and retrieve detailed data from the RCSB PDB. This makes it easy to ask questions about PDB entries, molecules, sequences, experimental data, and computed structure models, and get structured results you can use in research chats, analyses, or reports.

How to use

You can connect an AI interface to the PDB Explorer MCP to query PDB data and retrieve precise details. Two common ways are through a Cloudflare AI Playground or via the Claude Desktop app. Once connected, you can ask natural-language questions like the title and experimental method for a PDB entry, details about molecules, or summaries of computed structure models.

How to install

Prerequisites you need before running or connecting to an MCP server:

Install Node.js and npm on your machine.

If you plan to run a local MCP server, clone the project repository, install dependencies, build, and start the server using the standard workflow.

How to install (continued)

# Example workflow for a local MCP server
# 1) Install dependencies
npm install

# 2) Build the project (if a build step exists)
npm run build

# 3) Start the server (adjust if your project uses a different start script)
npm run start

Additional sections

Connecting to a deployed PDB Explorer server requires the server URL. For example, you may have a deployed server URL like the following, which you would use when configuring your MCP client:

https://rcsb-pdb-mcp-server.quentincody.workers.dev/sse

If you are configuring an editor or agent to run the MCP locally via a tool, you can use a configuration snippet that points the tool to the deployed server. The snippet below shows the required format and how to reference your server URL. Replace YOUR_RCSB_PDB_SERVER_URL_HERE with the actual deployed URL.

{
  "mcpServers": {
    "rcsb-pdb": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://rcsb-pdb-mcp-server.quentincody.workers.dev/sse"
      ]
    }
  }
}

Security and notes

Keep access to the MCP server restricted to trusted AI interfaces and approved clients. Do not expose the server URL in public or untrusted environments. If your deployment requires authentication, add the necessary security controls at the server or gateway level.