PubMed Search & Paper Analysis
JackKuo666LABSCO SUMMARY
Finding the paper and reading it well are separate problems; this covers both.
Search and metadata are the easy half. The prompt generators are the part that changes the output — a structured comparison prompt built from two IDs produces a far more useful read than "compare these papers". One caveat stated plainly by the author: trend analysis is not real data yet, so do not plan around it.
What it is
An MCP server over the arXiv paper database, built with FastMCP. It searches, looks up detail, and also generates the prompts that help a model summarise or compare the papers it found.
What you get
- Keyword search across arXiv with sorting by relevance or most recent
- Latest papers by category — give an arXiv category code such as
cs.AIormath.APand get the newest work in that field - Paper detail by arXiv ID: title, authors, abstract, categories, DOI, PDF link and more
- Author search for everything a given researcher has published
- Prompt generators — a summarisation prompt for one paper, and a structured comparison prompt from two paper IDs
- Trend analysis over recent papers in a category is present but experimental, and currently uses mock data
Requirements
Python 3.11+ and uv. Install from PyPI with uv pip install arxiv-paper-mcp, then run arxiv-paper-mcp or register uv tool run arxiv-paper-mcp in your client's MCP settings; Smithery can install it for Claude Desktop. No account or API key — arXiv is open. Package version 0.1.2.
Setup effort
One command — uv pip install arxiv-paper-mcp
USE CASES
01
Catching up on a field
Pull the most recent papers in the category code you follow, sorted by date.
02
Deciding whether a paper is worth reading
Fetch its abstract, categories and PDF link, then generate a summarisation prompt for the model to work from.
03
Comparing two approaches
Give both arXiv IDs and get a structured comparison prompt rather than an unguided summary.
04
Following a researcher's work
Search by author and read the metadata of everything they have published.
BEFORE YOU INSTALL
What to know first
Requires
- uv on PATH
- Python 3.11+
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
uv pip install arxiv-paper-mcpMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"arXivPaper": {
"command": "uv",
"args": [
"tool",
"run",
"arxiv-paper-mcp"
]
}
}
}
