BioMCP: Connecting AI to 15+ Biomedical Databases

By Prahlad Menon Published 2026-03-04 4 min read

If you’ve ever wished you could ask Claude or another AI assistant to look up a gene variant in ClinVar, find recruiting clinical trials, or search PubMed—all within a single conversation—BioMCP makes that possible. It’s an open-source Model Context Protocol (MCP) server that connects AI assistants directly to over 15 biomedical databases through a unified command grammar.

What Is BioMCP?

BioMCP is a single-binary CLI and MCP server developed by GenomOncology. It provides a consistent interface for querying diverse biomedical data sources—from gene databases to clinical trial registries to drug safety reports. The project is MIT-licensed, written in Rust, and works with Claude Desktop, Cursor, and any MCP-compatible client.

The key insight behind BioMCP is that biomedical research requires consulting multiple databases that each have their own APIs, authentication schemes, and query languages. BioMCP normalizes all of this into a simple grammar:

search <entity> [filters]    # Discovery
get <entity> <id> [sections] # Focused detail
<entity> <helper> <id>       # Cross-entity pivots

Supported Databases and Entities

BioMCP connects to an impressive array of authoritative sources across 12 entity types:

Entity	Data Sources
Gene	MyGene.info, UniProt, Reactome, QuickGO, STRING, CIViC
Variant	MyVariant.info, ClinVar, gnomAD, CIViC, OncoKB, cBioPortal, GWAS Catalog, AlphaGenome
Article	PubMed, PubTator3, Europe PMC
Trial	ClinicalTrials.gov, NCI CTS API
Drug	MyChem.info, ChEMBL, OpenTargets, Drugs@FDA, CIViC
Disease	Monarch Initiative, MONDO, CIViC, OpenTargets
Pathway	Reactome, g:Profiler
Protein	UniProt, InterPro, STRING, PDB/AlphaFold
Adverse Event	OpenFDA (FAERS, MAUDE, Recalls)
PGx	CPIC, PharmGKB
GWAS	GWAS Catalog
Phenotype	Monarch Initiative (HPO)

Most queries work without any API keys. Optional keys for NCBI, OpenFDA, OncoKB, and AlphaGenome improve rate limits or enable additional features.

Installation and Setup

Getting started takes about 30 seconds:

# Install the binary
curl -fsSL https://raw.githubusercontent.com/genomoncology/biomcp/main/install.sh | bash

# Verify it works
biomcp health --apis-only

To use BioMCP with Claude Desktop or Cursor, add this to your MCP configuration:

{
  "mcpServers": {
    "biomcp": {
      "command": "biomcp",
      "args": ["serve"]
    }
  }
}

BioMCP also includes 14 “skills”—guided investigation workflows—that you can install into your agent directory:

biomcp skill install ~/.claude --force

These skills provide structured prompts for common research tasks like variant-to-treatment workflows, rare disease investigation, and pharmacogenomics analysis.

Practical Examples

Here’s what makes BioMCP genuinely useful. Instead of manually navigating multiple websites, you can query directly:

Find clinical trials for a specific mutation:

biomcp variant trials "BRAF V600E" --limit 5

Get drug safety data:

biomcp drug adverse-events pembrolizumab
biomcp search adverse-event -d pembrolizumab

Research a gene with pathway context:

biomcp get gene BRAF pathways interactions
biomcp gene articles BRCA1

Search for recruiting trials by condition:

biomcp search trial -c melanoma -s recruiting

Gene set enrichment analysis:

biomcp enrich BRAF,KRAS,NRAS --limit 10

Pharmacogenomics recommendations:

biomcp get pgx CYP2D6 recommendations

The cross-entity helpers are particularly powerful—they let you pivot between related data without rebuilding queries. Start with a disease, find associated genes, then pivot to clinical trials.

Built-in Skills for Complex Workflows

BioMCP bundles 14 guided investigation workflows that demonstrate best practices for common research scenarios:

Variant to Treatment — Map variants to evidence-based treatment options
Trial Searching — Patient matching against recruiting trials
Rare Disease — Evidence gathering and trial strategy for rare conditions
Hereditary Cancer — Comprehensive syndrome workup
Pharmacogenomics — Gene-drug interactions and dosing guidance
Literature Synthesis — Evidence synthesis with automatic cross-entity verification

View available skills with biomcp skill list and read any skill with biomcp skill show 03.

Why This Matters

For researchers, BioMCP transforms the AI assistant from a general-purpose chatbot into a biomedical research partner with direct database access. Instead of relying on the AI’s training data (which may be outdated), you get live queries against authoritative sources.

For developers building biomedical AI applications, BioMCP provides a well-designed foundation. The unified grammar, caching layer (at ~/.cache/biomcp), and built-in rate limiting mean you don’t need to implement the plumbing for each data source yourself.

The progressive disclosure model is also worth noting—get commands support selectable sections, so you can request just what you need (summary, pathways, interactions) rather than pulling everything.

Getting Started

Install BioMCP: curl -fsSL https://raw.githubusercontent.com/genomoncology/biomcp/main/install.sh | bash
Run biomcp list to see all available entities and commands
Try biomcp get gene BRAF for a quick test
Configure your MCP client to use biomcp serve

The full documentation includes detailed guides for each entity type, troubleshooting help, and API reference material.

References: