mcp-pubmed

This MCP server provides PubMed data to your LLM workflows. It exposes a set of tooling endpoints that allow you to search PubMed, fetch detailed records, extract key insights, and manage full-text availability and EndNote exports. Typical usage involves running the server locally or in your environment, then configuring your MCP client (e.g., VS Code extension, Cherry Studio, Claude Desktop/Code) to point at the local node process. The included tools include pubmed_search for querying literature, pubmed_get_details for retrieving detailed records, pubmed_extract_key_info for structured extraction, and pubmed_batch_query to optimize batch requests. You can also check full-text status, download Open Access PDFs when available, and manage EndNote RIS/BibTeX exports. The server supports proxy configuration and adjustable abstract/full-text modes to balance context length with retrieval depth.

Prerequisites:

• Node.js v18.0.0+ installed
• npm installed
• Access to PubMed API key and a registered email

Steps:

1. Clone the repository and navigate to the project

git clone https://github.com/your-repo/mcp-pubmed-server.git
cd mcp-pubmed-server

2. Install dependencies

npm install

3. Create and configure environment variables

cp .env.example .env

Edit .env to include:

PUBMED_API_KEY=your-pubmed-api-key
PUBMED_EMAIL=your-email@example.com
ABSTRACT_MODE=deep
FULLTEXT_MODE=enabled
PROXY_ENABLED=disabled
HTTP_PROXY=
HTTPS_PROXY=
ENDNOTE_EXPORT=enabled

4. Run the MCP server (as a local MCP server)

node src/index.js

5. Verify the server starts and outputs a ready message such as:

PubMed Data Server v2.0 running on stdio

6. Optional: configure your MCP client adapters (VS Code, Cherry Studio, Claude) with the provided templates in the README to connect to this server.

Tips and notes:

• The server exposes PubMed-related tools via JSON interfaces; ensure your client config uses the same command and environment as shown in the README examples.
• If you enable FULLTEXT_MODE, the server will attempt to detect and download OA full texts where available; ensure your network allows access.
• Use ABSTRACT_MODE to balance context length; quick mode reduces the abstract length for faster queries, while deep mode provides longer abstracts when context window permits.
• PROXY_ENABLED and related HTTP_PROXY/HTTPS_PROXY settings help in restricted networks; set appropriate values if you’re behind a corporate proxy.
• EndNote export (RIS/BibTeX) can be disabled by setting ENDNOTE_EXPORT to disabled if you don’t need exports.
• For MCP client integrations, you can copy the server configuration into your client’s mcpServers section to enable seamless cross-tool usage.