home / mcp / pubmed-mcp mcp server

PubMed-MCP MCP Server

Provides an MCP server to query PubMed via Entrez and return article data in JSON.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "emi-dm-pubmed-mcp": {
      "command": "python",
      "args": [
        "pubmed_server.py"
      ]
    }
  }
}

PubMed-MCP is an MCP server that lets you search PubMed articles through the NCBI Entrez API and receive detailed results in JSON. It’s useful for programmatic access to PubMed data from downstream applications, dashboards, or research workflows.

How to use

You interact with the PubMed MCP server using an MCP client. Start the server locally, then send search requests to retrieve matching PubMed articles. Each result includes key fields such as PMID, title, authors, abstract, journal, and publication date. You can control the maximum number of results returned per query to suit your needs.

How to install

Prerequisites: you need Python 3.11+ installed on your system.

1. Clone the project and navigate into it.

2. Install dependencies via the runtime runner. The server is designed to be started with the standard Python invocation.

3. Configure Entrez email in your server code before running. This email is required by the NCBI Entrez API.

4. Start the server.

Configuration and notes

To run the PubMed MCP server locally, you execute the following command from your project directory.

python pubmed_server.py

Set Entrez.email in your code to your actual email address as soon as you start the server.

Available tools

search_pubmed

Search PubMed for articles matching a query and return a list of article objects containing PMID, title, authors, abstract, journal, publication year, month, and a URL to PubMed.