PaperRadar 🛰️📄

An MCP server that pulls the latest papers in your field into Claude (or any MCP client).

Ask your assistant:

"What's new in cyanobacteria pigment research this week? Summarize the top 5."

…and it fetches recent papers (journals and preprints) from Europe PMC, newest first, so the model can summarize / compare them. The reasoning runs on your Claude/ChatGPT — PaperRadar just fetches and structures the data. No API key. No account. Free.

Built in a weekend (~1.5 days) as a #weekendbuild.

What it gives Claude

One tool:

• search_recent_papers(query, days=7, limit=20) — latest papers matching query, first-published within the last days. Returns title, authors, date, source, doi, url, abstract.

Add to Claude

With uv — runs straight from this repo, no PyPI needed:

Claude Code:

claude mcp add paper-radar -- uvx --from git+https://github.com/nagidev-lab/paper-radar paper-radar

Claude Desktop — add to claude_desktop_config.json:

{ "mcpServers": { "paper-radar": {
  "command": "uvx",
  "args": ["--from", "git+https://github.com/nagidev-lab/paper-radar", "paper-radar"]
} } }

From source (no uv):

git clone https://github.com/nagidev-lab/paper-radar && cd paper-radar
python -m venv .venv && .venv/bin/pip install -e .
claude mcp add paper-radar -- /full/path/to/paper-radar/.venv/bin/paper-radar

Then just ask: "what's new in <topic> this week?"

(A PyPI release is coming, which will shorten this to uvx paper-radar.)

How it works

• Data: Europe PMC REST API (covers PubMed + preprints such as bioRxiv). No auth.
• Date filter + newest-first sort done server-side; the model does the summarizing/comparing.
• Titles/abstracts are stripped of HTML entities before returning.

Roadmap (maybe)

• more sources (arXiv, direct bioRxiv), per-paper detail, saved topics, a hosted version.

---

MIT. A weekend build — feedback welcome.