BioMCPMCP Server

BioMCP is a single binary that puts roughly 30 trusted biomedical sources behind one consistent command grammar, and doubles as an MCP server so agents such as Claude Code, Claude Desktop and Codex can query them directly. It covers PubMed and PubTator3, Europe PMC, ClinicalTrials.gov, ClinVar, OncoKB, Reactome and more — sources that normally each demand their own API, identifier scheme and search conventions.

The design goal is to stop rebuilding a workflow for every database. search article fans out across PubTator3 and Europe PMC, deduplicates PMID, PMCID and DOI identifiers, and can add a Semantic Scholar leg when the filters allow it. From any result you can pivot straight into the next built-in view — gene, variant, drug, disease, pathway, protein or article — instead of hand-rebuilding filters. article citations, article references, article recommendations and article entities turn one known paper into a broader evidence map, and biomcp batch bundles up to ten focused get calls into a single command.

Beyond live lookups, the study commands run local analytics over downloaded cBioPortal-style datasets — query, cohort, survival, comparison and co-occurrence workflows — and render results as native terminal, SVG or PNG charts. biomcp skill list ships worked examples as guided playbooks, and biomcp skill install ~/.claude drops those investigation workflows into an agent directory.

Install via the biomcp.org script, Homebrew, Docker (ghcr.io/genomoncology/biomcp) or PyPI. Note the packaging caveat the project flags itself: install biomcp-cli, not the unrelated biomcp PyPI package. It registers with Claude Code as a plugin, with Codex via codex mcp add, and serves remote deployments over HTTP. MIT licensed and maintained by GenomOncology.