BioMCP is a single-binary CLI and MCP server designed to query a wide range of biomedical databases from a unified interface. It provides a compact markdown output format across 12 entities (gene, variant, article, trial, drug, disease, pathway, protein, adverse-event, pgx, gwas, phenotype) drawing data from more than 15 data sources. The server exposes a set of core commands for discovery and detailed retrieval, including search, get, enrich, and batch operations, along with cross-entity pivots to navigate between related results. You can also run guided workflows via the built-in skills to assist investigations, and you can deploy a scalable HTTP endpoint for multi-worker usage with a shared limiter budget. The toolchain is designed for researchers and clinicians who need rapid access to integrated biomedical evidence from multiple knowledge bases through a consistent command syntax.

To use BioMCP, start the server with biomcp serve, then interact with the CLI to perform discovery and retrieval. Examples include biomcp health to verify API connectivity, biomcp list to enumerate entities and commands, and biomcp get gene BRAF to fetch gene-level details. You can also perform cross-entity pivots like biomcp variant trials "BRAF V600E" or biomcp pathway drugs R-HSA-5673001 to explore linked results across data sources. For enrichment workflows, run biomcp enrich BRAF,KRAS,NRAS to perform gene-set enrichment analyses, or leverage the 14 built-in guided investigations via biomcp skill list and biomcp skill show <id>.

Operationally, BioMCP supports per-source API keys and rate-limiting considerations. Optional API keys can improve throughput (for PubTator, NCBI, OpenFDA, OncoKB, AlphaGenome, and others). The server can also be deployed in multi-worker setups by exposing a single HTTP endpoint (serve-http) so multiple workers share the same rate limiter budget, e.g., biomcp serve-http --host 0.0.0.0 --port 8080.

Prerequisites:

• A working development environment with either Rust/Cargo, Python, or Node tooling depending on installation path
• Internet access to fetch packages and install dependencies

Installation options (choose one):

1. PyPI tool installation (Python)

• Prereq: Python and pip installed
• Command: uv tool install biomcp-cli

  or: pip install biomcp-cli

• This installs the biomcp binary on your PATH

2. Binary install (system-wide script)

• Prereq: curl
• Command: curl -fsSL https://biomcp.org/install.sh | bash
• This installs the biomcp binary and places it on your PATH

3. Install skills (optional)

• Command: biomcp skill install ~/.claude --force
• This installs guided investigation workflows into your agent directory

4. Build from source (for developers)

• Prereq: Rust toolchain (cargo)
• Command: cargo build --release --locked
• The biomcp binary will be produced in target/release/ and can be run directly

MCP clients configuration example (for reference):

• The following JSON enables biomcp as an MCP client/server: { "mcpServers": { "biomcp": { "command": "biomcp", "args": ["serve"] } } }

Notes on usage:

• After installation, you can verify health with biomcp health and list capabilities with biomcp list.
• For a multi-worker HTTP deployment, run biomcp serve-http to expose an endpoint that multiple workers can share.

Tips and known considerations:

• API keys: You may set environment variables to improve rate limits and access to data sources (e.g., NCBI_API_KEY for PubTator/NCBI IDs, OPENFDA_API_KEY for OpenFDA, ONCOKB_TOKEN for OncoKB, ALPHAGENOME_API_KEY for AlphaGenome).
• Multi-source results: BioMCP aggregates data from many sources; expect slight variations in entity naming and IDs across providers. Use cross-entity pivots to navigate related evidence.
• Performance: For high concurrency, consider a shared serve-http endpoint to allow multiple workers to reuse a single limiter budget.
• Commands overview:
  • biomcp health: check API connectivity
  • biomcp list: enumerate entities and commands
  • biomcp get <entity> <id> [sections]: retrieve focused details with selectable sections
  • biomcp search <entity> [filters]: discovery queries
  • biomcp enrich <genes> --limit N: gene-set enrichment
  • biomcp batch <entity> <ids>: parallel gets
  • biomcp skill list/show: view and run guided workflows
• If you encounter issues, consult the official docs: biomcp.org and the Troubleshooting guide in the docs directory.