BioMCP (Typescript)
About
Agent-first rewrite of genomeoncology's BioMCP in TypeScript to provide next-gen biomedical data access for agents.
Details
- Author
- yeyuan98
- Categories
- Database, Other
Jump to
Setup
Install BioMCP (Typescript) in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/yeyuan98/biomcp-ts
Follow the installation instructions in the repository README, then restart your MCP client.
A high-performance MCP server that gives LLMs access to 25 biomedical tools federated across 50+ upstream APIs — genes, variants, drugs, diseases, literature, clinical trials, and structural biology in a single integration.
Adapted from theBioMCP Rustwith agent-first development approach and enhancements. Kudos to the original authors.
- 25 toolsacross 8 domains — search, retrieve, and cross-reference biomedical entities
- 50+ upstream sources— MyGene, MyVariant, MyChem, MyDisease, ClinVar, gnomAD, UniProt, Reactome, OpenTargets, CIViC, OncoKB, DisGeNET, GTEx, STRING, DGIdb, ClinicalTrials.gov, PubMed, EuropePMC, Semantic Scholar, PubTator, LitSense, Monarch Initiative, OpenFDA, NIH Reporter, AlphaGenome, and more
- Section-based fetching—entityGet(id, sections)fans out to multiple sources with per-section timeouts and graceful degradation (failed sections return{ _error }instead of crashing)
- Federated article search— queries 5 literature backends simultaneously with PMID/PMCID/DOI deduplication
- Zero-config startup— works out of the box; optional API keys unlock higher rate limits and premium data
- 350+ unit tests(mocked) +90+ integration tests(live APIs via in-process MCP client, with automatic retry on 429 rate limits)
git clone <repo-url> && cd biomcp-ts make install build
Add to your Claude Desktopclaude_desktop_config.json:
{ "mcpServers": { "biomcp": { "command": "npx", "args": ["biomcp"] } } }
{ "mcpServers": { "biomcp": { "command": "node", "args": ["/path/to/biomcp-ts/dist/bundle.js"] } } }
BioMCP speaks standard MCP overstdio. Point any MCP client at thebiomcpbinary ornode dist/bundle.js.
- Fast(default): Europe PMC, Semantic Scholar, Crossref (~4s)
- Returns: Citations with title, authors, journal, year when available
- Note: Crossref requires DOI; Europe PMC auto-resolves DOI/PMCID to PMID
- Adds: PubMed (PMID only) and OpenCitations (DOI only)
- Use for: Comprehensive citation analysis
Automatic Fallback:Fast mode automatically queries PubMed when other providers return counts but no items (requires PMID).
make # Show available targets make install # Install dependencies make build # Compile and bundle into dist/bundle.js make typecheck # Type-check without emitting make test # Run unit tests (fast, mocked) make test-integration # Run integration tests (live APIs, ~60s) make test-all # Run all tests make clean # Remove build artifacts
After building, you can test the MCP server locally via npx:
make build # Creates dist/bundle.js npx . # Runs the bundled MCP server
This is the recommended workflow for local development and testing.
All keys are optional. BioMCP works without any keys — they unlock higher rate limits and additional data sources.
Access and interact with Allen Institute for Neural Dynamics (AIND) metadata directly within your IDE.
A high-performance JavaScript server for the Alliance of Genome Resources (AGR) MCP.
Access the AlphaFold Protein Structure Database for protein structure prediction and analysis.
Interface with Biomart, a biological data query tool, using the pybiomart Python package.
Perform complex queries on the DANDI Archive, a platform for neurophysiology data.
Interact with DROMA drug-omics association analysis databases using natural language.
A bridge to the Drug Gene Interaction Database (DGIdb) API, enabling AI clients to query drug-gene interaction data.
Query the Materials Project database using the mp_api client. Requires an MP_API_KEY environment variable.
Access PubMed articles through the Entrez API.
Access NCBI's suite of APIs, including E-utilities, BLAST, PubChem, and PMC services.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
