PubCrawl

A peer-reviewed pub crawl through the literature.

Your AI can write code, summarise documents, and hold a conversation. But ask it about a specific paper or a drug’s prescribing information and you'll get a confident guess — or a hallucinated DOI that goes nowhere. PubCrawl fixes that.

Real papers. Real data. Real citations.

View on GitHub → Install →
The Problem

Your AI tools don’t have access to PubMed.

When you ask an AI about a paper, it’s usually working from memory — training data that’s months or years out of date. It’ll confidently cite papers that don’t exist.

Most AI tools can’t search by MeSH term, filter by study type, or tell you what was published last week. For anyone working in healthcare, pharma, or biomedical research, that’s not good enough. PubCrawl gives reliable PubMed access to all of them.

× Hallucinated doi:10.1016/j.cell.2024.08.142
✓ Verified doi:10.1016/j.media.2023.102802
What PubCrawl Does

Thirteen tools. One install. Literature, labels, and trials.

PubCrawl is an MCP server — a plugin that gives AI assistants real-time access to PubMed, FDA drug labelling, UK medicines data, and ClinicalTrials.gov. Install it once, and your AI can:

Search PubMed
Full query support with filters for date range, article type, and sort order. PubMed’s own search syntax works too — Boolean operators, field tags, all of it.
Pull Structured Abstracts
Background, methods, results, and conclusions parsed into sections. Plus MeSH terms, keywords, and links to the full text where available.
Retrieve Full Text
For open-access papers in PubMed Central, extract the complete article: sections, figure captions, table descriptions, and reference counts.
Find Related Papers
Surface related work using NCBI’s citation-aware relevance scoring. Follow threads through the literature without manual searching.
Format Citations
APA, Vancouver, Harvard, or BibTeX — generated from real metadata, not guesswork.
Track What’s New
Find recent papers in any topic area, with optional filtering to high-impact journals. Stay current without setting up alerts.
Resolve Drug Names
Convert a brand name to its generic — or a generic to its US brand names — with drug class and common indications. Deterministic, via RxNorm. No guesswork.
US Prescribing Information
Pull FDA-approved USPI sections via openFDA, cited to DailyMed — indications, dosing, warnings, contraindications, and more.
UK SmPC Data
Retrieve Summary of Product Characteristics from the eMC — the UK equivalent of US prescribing information, with numbered SmPC sections.
Compare US/UK Labels
Side-by-side comparison of FDA and MHRA/EMA labelling for the same drug. Spot regulatory differences in indications, warnings, and dosing.
Search by Indication
Find drugs approved for a medical condition via OpenFDA, then cross-reference UK availability on the eMC. One query, both markets.
Search Clinical Trials
Query ClinicalTrials.gov by condition, intervention, phase, and recruitment status. Find what’s recruiting, completed, or in progress.
Get Trial Details
Full study detail by NCT ID — eligibility criteria, study design, arms, outcomes, locations, and linked PubMed publications.
How It Works

Plug in and go.

PubCrawl uses the Model Context Protocol — an open standard created by Anthropic that lets AI applications call external tools securely. Think of it as a USB port for AI: plug in PubCrawl, and any MCP-compatible client gains biomedical literature and drug data superpowers.

1 You ask your AI something about the literature
2 The AI recognises it needs PubMed and calls PubCrawl
3 PubCrawl queries NCBI, OpenFDA, DailyMed, the eMC, RxNorm, or ClinicalTrials.gov in real time
4 Results come back structured and verified — real PMIDs, real DOIs, real data
5 Your AI uses the results to give you a grounded, accurate answer

No API keys required. No subscription. Just install and go.

See It in Action

From question to verified citations.

Built With

Solid foundations.

  • TypeScript · Type-safe and maintainable
  • MCP SDK · Anthropic’s official protocol
  • NCBI E-utilities · PubMed’s own API
  • OpenFDA · FDA drug labelling search
  • DailyMed · US structured product labels
  • eMC · UK medicines compendium
  • RxNorm · Drug name normalisation
  • ClinicalTrials.gov · Clinical trials registry
  • fast-xml-parser · PubMed XML handling
  • LRU Cache · Smart rate-limit-aware caching
Who It’s For

Built for people who need the literature to be right.

Medical Writers
Verify references instantly. Pull structured abstracts without leaving your writing environment. Generate properly formatted citations in any style.
Researchers
Explore the literature conversationally. Follow citation trails. Stay on top of new publications in your field.
Healthcare AI Teams
Ground your models in real evidence. Build retrieval-augmented workflows that cite actual papers.
Regulatory & Pharma Teams
Compare US and UK drug labelling side by side. Search by indication across FDA and eMC. Ground regulatory intelligence in real prescribing data.
Anyone Tired of Hallucinated Citations
If you’ve ever clicked a DOI that led nowhere, PubCrawl is for you.
Install

Up and running in minutes.

Works with Claude Desktop, Cursor, and any MCP-compatible client.

terminal 
npm install -g @pharmatools/pubcrawl

Then add to your claude_desktop_config.json:

claude_desktop_config.json 
{
  "mcpServers": {
    "pubcrawl": {
      "command": "pubcrawl"
    }
  }
}

Restart Claude Desktop. Full setup instructions on GitHub.

npm Package → GitHub Repository →
About

A PharmaTools.AI project.

PubCrawl is part of a growing suite of tools built at the intersection of AI and healthcare. It’s open source, free to use, and built by someone who’s spent 25 years turning complex medical information into something useful.

Because the literature, the label, and the trial are too important to hallucinate.

PubCrawl is not affiliated with NCBI, the National Library of Medicine, the FDA, the MHRA, or ClinicalTrials.gov. PubMed® is a registered trademark of the U.S. National Library of Medicine. DailyMed, OpenFDA, and ClinicalTrials.gov are services of the U.S. National Library of Medicine and the FDA respectively.