What Is BioMCP?¶
BioMCP is a biomedical command-line interface and MCP server designed for practical research and clinical-informatics workflows.
It combines two ideas:
- A stable command grammar for entity retrieval and search.
- A protocol surface for agent runtimes through MCP tools and resources.
Core goal¶
The project goal is not to mirror every field from every API. The goal is to deliver dependable, compact answers that can be expanded on demand.
The command grammar¶
BioMCP uses one consistent grammar:
search <entity> [filters]get <entity> <id> [section...]
This grammar is intentionally shared across entities so users and agents do not need a different mental model per endpoint.
Entities¶
BioMCP covers entities across clinical, research, and regulatory domains:
Core clinical entities: gene, variant, trial, article, drug, disease
Extended entities: pathway, protein, adverse-event, PGx (pharmacogenomics)
Discovery entities: GWAS, phenotype search, biomarker, organization, intervention
Why this matters for agents¶
Agent runtimes often fail on data tooling for three reasons:
- command discovery is inconsistent,
- output is too verbose or unstable,
- source provenance is unclear.
BioMCP addresses these with:
- predictable commands,
- compact markdown-first defaults,
- explicit source mapping in docs.
What BioMCP is not¶
BioMCP is not a local biomedical warehouse. It is also not a replacement for formal clinical interpretation systems.
It is a workflow interface that normalizes retrieval from trusted public APIs.
Progressive detail¶
BioMCP starts concise, then expands by section.
Examples:
biomcp get gene BRAF
biomcp get gene BRAF pathways
biomcp get article 22663011
biomcp get article 22663011 fulltext
The first command gives orientation. The second command gives focused depth.
MCP mode¶
biomcp serve runs a stdio MCP server.
The server advertises:
- tools: one execution tool (
shell) - resources: curated markdown help and pattern documents
This keeps client integration simple while preserving discoverability.
Where BioMCP fits¶
BioMCP works well for:
- trial searching triage,
- variant-to-literature pivots,
- disease normalization and downstream lookups,
- adverse-event and recall quick checks.
Design priorities¶
The design priorities are:
- deterministic command behavior,
- explicit error messages,
- low context overhead,
- documented API provenance,
- CI-backed contract checks.
Next reading¶
- Concept: Progressive disclosure
- User guides:
docs/user-guide/ - Data provenance:
docs/reference/data-sources.md - MCP behavior:
docs/reference/mcp-server.md