biomcp MCP — Use Cases, Install & Live Demo

Why use it

Key features

11+ entity types covering genes, variants, articles, trials, drugs, diseases, pathways, proteins, adverse events, pharmacogenomics, GWAS
Deduplication across overlapping sources (PubTator3)
Optional API keys unlock higher rate limits and premium data (NCBI, OncoKB, Semantic Scholar, NCI, OpenFDA)
Install via curl script, PyPI (biomcp-cli), or remote HTTP
Zero-key operation possible — unauthenticated public APIs work by default

Live Demo

What it looks like in practice

biomcp.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "biomcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/genomoncology/biomcp"
      ],
      "_inferred": true
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "biomcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/genomoncology/biomcp"
      ],
      "_inferred": true
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "biomcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/genomoncology/biomcp"
      ],
      "_inferred": true
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "biomcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/genomoncology/biomcp"
      ],
      "_inferred": true
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "biomcp",
      "command": "TODO",
      "args": [
        "See README: https://github.com/genomoncology/biomcp"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "biomcp": {
      "command": {
        "path": "TODO",
        "args": [
          "See README: https://github.com/genomoncology/biomcp"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add biomcp -- TODO 'See README: https://github.com/genomoncology/biomcp'

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use biomcp

Triage a patient's gene variant for clinical relevance

👤 Genetic counselors, clinical researchers ⏱ ~15 min advanced

When to use: A VCF annotation returned a variant you've never seen — is it actionable?

Prerequisites

biomcp installed — curl -fsSL https://biomcp.org/install.sh | bash

Flow

Look up the variant

Look up BRCA2 c.7934del. ClinVar significance, gnomAD frequency, and any CIViC evidence.✓ Copied

→ Consolidated card from multiple sources
Pull supporting papers

Find the 3 most cited PubMed articles on this variant.✓ Copied

→ Paper list with PMIDs
Check for trials

Any open clinical trials recruiting for BRCA2 mutation carriers?✓ Copied

→ Trial list from ClinicalTrials.gov

Outcome: A multi-source picture of a variant with evidence pointers.

Pitfalls

Not a substitute for clinical review — Always have a qualified clinician review — this is research aid, not diagnosis
Some sources disagree on variant classification — Surface disagreements explicitly; don't average them

Conduct a targeted literature review

👤 PhD students, MDs, industry researchers ⏱ ~20 min intermediate

When to use: You need a bibliography on a specific gene-disease-drug angle.

Flow

Search articles

Search PubMed + Europe PMC for 'KRAS G12C AND adagrasib' 2022-2026. Dedupe across sources.✓ Copied

→ Deduplicated paper list
Summarize

For the top 10 by citation count, give me title + one-line take-home.✓ Copied

→ Annotated bibliography

Outcome: A defensible starting bibliography.

Combine with: filesystem

Check adverse-event profile for a drug

👤 Pharmacy, safety teams ⏱ ~20 min intermediate

When to use: Before recommending a drug, check OpenFDA / Drugs@FDA signals.

Flow

Query drug

For pembrolizumab: top adverse events per OpenFDA FAERS, last 5 years, broken down by seriousness.✓ Copied

→ Ranked AE table
Context

What's the expected AE profile per the FDA label vs FAERS real-world?✓ Copied

→ Label vs real-world comparison

Outcome: Grounded safety picture combining label and post-market data.

Find clinical trials matching a patient profile

👤 Clinicians, patient advocates ⏱ ~25 min advanced

When to use: A patient exhausted standard options and you want to scan for trial matches.

Flow

Match trials

Find Phase 2/3 recruiting trials for stage IV NSCLC with EGFR exon 20 insertion, US sites.✓ Copied

→ Trial list with NCT IDs
Summarize eligibility

For the top 5, summarize key inclusion/exclusion criteria.✓ Copied

→ Per-trial eligibility card

Outcome: A prioritized trial match list for clinical discussion.

Combinations

Pair with other MCPs for X10 leverage

biomcp + filesystem

Save evidence memos as local markdown

For variant BRCA2 c.7934del, generate a one-page evidence memo and save to ~/research/brca2-7934.md.✓ Copied

biomcp + documentation-server

Build a personal literature library across biomcp queries

Every paper I star via biomcp, also ingest into documentation-server for future RAG queries.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
gene	symbol or id, sources?: str[]	Gene-level lookup	multiple upstream calls
variant	hgvs\|rsid\|position	Variant interpretation	multiple upstream calls
article	query or pmid	Literature search	PubMed/Europe PMC/S2 calls
trial	query/condition/phase/status	Clinical trial match	CT.gov + NCI CTS calls
drug	name or id	Drug profile	multiple calls
disease	name or MONDO id	Disease context	multiple calls
adverse_event	drug or reaction	Safety analysis	OpenFDA calls

Cost & Limits

What this costs to run

API quota: Public APIs default. Optional keys raise limits: NCBI_API_KEY, S2_API_KEY, ONCOKB_TOKEN, OPENFDA_API_KEY, NCI_API_KEY
Tokens per call: Consolidated entity responses 1-5k tokens
Monetary: Free; some premium sources (OncoKB clinical) require license even with key
Tip: For heavy literature work, get NCBI_API_KEY (free) — triples PubMed rate.

Security

Permissions, secrets, blast radius

Credential storage: Env vars per provider key

Data egress: Calls to biomedical public APIs; queries themselves can be sensitive (patient variants etc.)

Queries can include PHI-adjacent data — confirm your deployment's data handling before running on real patient info.
Not a medical device; outputs are research aids only.

Troubleshooting

Common errors and fixes

Rate limited on PubMed

Set NCBI_API_KEY for higher limits.

Verify: Check NCBI API Key page

OncoKB returns 403

OncoKB requires a token; register at oncokb.org and set ONCOKB_TOKEN.

Verify: curl -H 'Authorization: Bearer $ONCOKB_TOKEN' https://www.oncokb.org/api/v1/utils/numbers

No results for a known gene symbol

Some symbols have aliases; try both (e.g. ERBB2 vs HER2).

Verify: Search on genecards.org

Alternatives

biomcp vs others

Alternative	When to use it instead	Tradeoff
Individual databases via direct REST	You only need one source	No deduplication or cross-linking
OpenTargets Platform directly	You primarily care about target-disease evidence	Narrower coverage

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills