BioMCP

BioMCP is a single MCP server that wraps PubMed, ClinicalTrials.gov, UniProt, and a handful of other biomedical databases behind a unified query interface. Instead of juggling API keys and learning five different query syntaxes, you ask Claude to search PubMed for papers on BRCA1 mutations or pull clinical trials for a specific drug, and BioMCP handles the translation. It's built by GenomOncology, so the domain knowledge shows: queries feel tuned for real research workflows, not generic web scraping. I'd reach for this when I'm doing literature reviews or need to cross-reference a gene variant across databases. The standout feature is how it handles follow-up questions. You can ask for papers, then drill into one, then pivot to UniProt for protein details, all without switching tools. The grammar is consistent enough that Claude rarely misunderstands what you want, which matters when you're chaining queries. Quirks: it's read-only, so you can't submit data or save searches server-side. The tool also doesn't cache results aggressively, so repeated queries hit the upstream APIs each time. If you're doing bulk analysis, you'll want to export results and work locally. The install is trivial (one uv command, no API keys), but the lack of rate-limiting visibility means you might hit PubMed's throttle without warning. Who shouldn't bother: if you're not working in biomedicine or adjacent fields, this is too specialised. It won't help with general research, and the databases it covers are niche outside life sciences. Also, if you need write access or want to integrate with lab systems, this isn't the tool. It's a query layer, not a workflow engine. Who should: researchers doing exploratory work, bioinformaticians who want Claude as a research assistant, or anyone tired of tabbing between PubMed, ClinicalTrials, and UniProt. The zero-config setup and unified grammar make it worth installing even if you only use it occasionally.