search_pubmed_key_wordsLABSCO SUMMARY
From a question to the papers, and into one of them.
The two ends of a literature search are usually different tools: a search interface for finding candidates, and something else entirely for reading what you found. Here both are calls in the same conversation, so narrowing and reading happen without moving the citations by hand.
What it is
A research path through PubMed's biomedical literature: keyword and advanced search, metadata for a specific article, an attempt at the full text, and an analysis pass over a paper you have found.
What you get
- Keyword search across PubMed
- Advanced search, for queries that need PubMed's own field syntax
- Detailed metadata for a specific article
- An attempt to download the full-text PDF where one is available
- A structured analysis of a paper, and prompts written for reading them
Requirements
Python 3.10 or newer. No account and no key — PubMed's index is public, though full-text availability depends on the article.
Setup effort
One command — npx -y @smithery/cli install @JackKuo666/pubmed-mcp-server --client claude
USE CASES
01
Starting a literature search from a plain question
Search by keywords and read the hits back with their metadata, then narrow before you commit to reading anything in full.
02
Running a query that needs PubMed's own syntax
Use the advanced search when the question is really a filter — a date range, a publication type, a specific field — rather than a bag of words.
03
Reading the paper, not just the abstract
Try for the full text once a paper looks relevant. Whether a PDF is reachable depends on the article, so this is a request that sometimes comes back empty.
04
Getting a structured read of a study
Run the analysis pass over a paper you have selected, so what comes back follows the same shape each time and two studies can be compared.
05
Checking a citation someone handed you
Look the article up and read back its recorded metadata — the fastest way to catch a reference that does not exist as described.
5 tools
Listed in project docs. Your client may show a different list.
search_pubmed_advancedget_pubmed_article_metadatadownload_pubmed_pdfdeep_paper_analysisBEFORE YOU INSTALL
What to know first
Tools you get
5 tools named in the project's documentation.
Requires
- python on PATH
- Python 3.10+
INSTALLATION
Connect your client
Server setup
Setup
TERMINAL
git clone https://github.com/JackKuo666/PubMed-MCP-Server.git
cd PubMed-MCP-Server
pip install -r requirements.txtnpx
Claude Desktop — Smithery
TERMINAL
npx -y @smithery/cli install @JackKuo666/pubmed-mcp-server --client claudeMCP Clients
Cursor
CURSOR · TERMINAL
npx -y @smithery/cli@latest run @JackKuo666/pubmed-mcp-server --client cursor --config "{}"Windsurf
WINDSURF · TERMINAL
npx -y @smithery/cli@latest install @JackKuo666/pubmed-mcp-server --client windsurf --config "{}"Cline
CLINE · TERMINAL
npx -y @smithery/cli@latest install @JackKuo666/pubmed-mcp-server --client cline --config "{}"Claude Desktop — macOS
CLAUDE DESKTOP — MACOS · JSON
{
"mcpServers": {
"pubmed": {
"command": "python",
"args": [
"<YOUR_PATH_TO_PUB_MED_MCP_SERVER>/pubmed_server.py"
]
}
}
}<YOUR_PATH_TO_PUB_MED_MCP_SERVER> is the full path of the folder the clone created, and use the python you ran pip install -r requirements.txt with.
Claude Desktop — Windows
CLAUDE DESKTOP — WINDOWS · JSON
{
"mcpServers": {
"pubmed": {
"command": "C:\\Users\\<YOUR_USERNAME>\\PATH\\miniconda3\\envs\\mcp_server\\python.exe",
"args": [
"D:\\code\\<YOUR_PATH_TO_PUB_MED_MCP_SERVER>\\pubmed_server.py"
],
"env": {},
"disabled": false,
"autoApprove": []
}
}
}command is the full path of python.exe in the environment you installed requirements.txt into. The path in args, up to \pubmed_server.py, is the full path of the PubMed-MCP-Server folder you cloned.
