Labsco
MCP SERVER

Medical MCP

by JamesANZ

FDA, WHO, PubMed, RxNorm and Scholar lookups as tools, with no API key anywhere in the setup.

Biomedical: Genomics & Clinical Data
Summary
Every answer traces back to a source you can name.

The value is not that a model can talk about medicine — it already will. It is that a drug answer comes from the FDA record, a prevalence figure from WHO, and a claim from a PubMed citation with a PMID you can open. The paediatric tools are the unusual part: dosing, AAP guidelines and child health indicators are separated out rather than buried in general search.

What it is

A local MCP server that queries named medical sources directly — the FDA drug database, WHO's Global Health Observatory, PubMed, RxNorm, Google Scholar and the American Academy of Pediatrics. Queries run in real time and nothing is stored.

What you get
  • Drugs: `search-drugs` against the FDA database by brand or generic name, `get-drug-details` by NDC code, `search-drug-nomenclature` for standardised RxNorm names
  • Population data: `get-health-statistics` for WHO indicators such as life expectancy, mortality and disease prevalence, by country
  • Literature: `search-medical-literature` over PubMed, `get-article-details` by PMID, `search-google-scholar`, `search-medical-databases` across PubMed, Scholar, Cochrane and ClinicalTrials.gov, and `search-medical-journals` for NEJM, JAMA, Lancet, BMJ and Nature Medicine
  • Clinical practice: `search-clinical-guidelines`
  • Paediatrics as a first-class area: `search-pediatric-guidelines`, `search-aap-guidelines`, `search-pediatric-literature`, `search-pediatric-drugs` for pediatric labelling and dosing, and `get-child-health-statistics`
  • A cache with per-source lifetimes — FDA 24 hours, PubMed 1 hour, WHO 7 days, RxNorm 30 days — readable through `get-cache-stats` and tunable with `CACHE_ENABLED`, `CACHE_MAX_SIZE` and the `CACHE_TTL_*` variables
Requirements

Node.js 18+ and npm; `npm install -g medical-mcp`, a one-click Cursor link, or a build from source. The package is `medical-mcp`. No keys and no accounts — every source is reached without credentials. One asymmetry worth knowing: Google Scholar is reached by rate-limited web scraping while the other sources use official APIs, so Scholar is the part most likely to fail. It is a reference tool, not medical advice.

Setup effort

One command — npm install -g medical-mcp