genomoncology/biomcp
biomcp is an open-source Model Context Protocol (MCP) server that bridges large language models (LLMs) with authoritative biomedical data sources. It implements the MCP specification to provide tools that AI assistants can use to query PubMed, ClinicalTrials.gov, and MyVariant.info in real-time. By running as a local or self-hosted service, biomcp enables secure, low-latency access to biomedical literature, clinical trial registries, and genetic variant annotations without exposing sensitive queries to third-party cloud APIs. The server exposes a set of MCP tools: searching PubMed by keywords or MeSH terms to retrieve citations and abstracts; querying ClinicalTrials.gov for study details based on conditions, interventions, or location; and looking up genetic variants by HGVS notation or dbSNP ID to get functional annotations from MyVariant.info. These tools are designed to be composable, allowing LLMs to chain searches across databases—for example, finding literature on a gene and then checking its known variant pathogenicity. biomcp is built in Python and leverages modern package management (uv) to simplify installation and deployment. It can be run locally via pip, Docker, or integrated into machine learning pipelines. The repository includes extensive documentation, example configurations for popular MCP clients like Claude Desktop and continue.dev, and benchmarks against the BioASQ dataset to validate retrieval accuracy. The project maintains canonical Open Access article fetching to ensure full-text access where permitted. With over 500 GitHub stars and active development (1,300+ commits), biomcp is a community-driven effort to make biomedical knowledge accessible to AI systems. It is designed for extensibility—new data sources or custom tools can be added via a plugin-like architecture—making it suitable for both individual researchers and enterprise-grade biomedical AI applications.
- Category
- Developer Tools
- Pricing
- Open Source
Problem solved
Biomedical researchers, clinicians, and AI developers face a fragmented landscape of databases with incompatible APIs and rate limits. LLMs often generate plausible but incorrect biomedical claims because they lack direct access to up-to-date, evidence-based data. biomcp solves this by providing a unified MCP interface that wraps PubMed, ClinicalTrials.gov, and MyVariant.info into a set of well-documented tools. This allows AI assistants to retrieve factual, real-time information about diseases, drugs, trials, and genetic variants, reducing hallucinations and enabling evidence-backed reasoning in biomedical applications.
Use cases
- Literature search for drug interactions or contraindications to assist in clinical decision support.
- Finding active or completed clinical trials for a specific disease or geographic region to aid patient recruitment or systematic reviews.
- Checking the pathogenicity and population frequency of a genetic variant to support variant interpretation in genomics pipelines.
- Supporting evidence-based diagnosis by cross-referencing symptoms, drugs, and literature from PubMed.
- Accelerating hypothesis generation for drug repurposing by linking disease-gene associations from literature with variant databases.
- Assisting in writing systematic reviews or meta-analyses by automating evidence retrieval across multiple sources.
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
Pros
- Unifies three critical biomedical databases under a single MCP interface.
- Self-hosted, ensuring data privacy and low latency for sensitive biomedical queries.
- Implements the MCP standard, so it works with any compatible AI assistant (Claude, continue.dev, etc.) with minimal configuration.
- Open-source with active development and thorough documentation, including benchmarks.
- Extensible design allows adding new data sources via plugins.
Cons
- Requires self-hosting and basic DevOps knowledge for setup and maintenance.
- Limited to the bundled data sources (PubMed, ClinicalTrials.gov, MyVariant.info) unless extended.
- Rate limits from upstream APIs (e.g., PubMed) may require obtaining API keys for heavy usage.
Official website: https://github.com/genomoncology/biomcp
Canonical LambdaBase page: https://www.lambdabase.com/tools/genomoncology-biomcp