Competitive landscape — bio-MCP servers (April 2026)

Last full review: 2026-04-26. Maintenance procedure: COMPETITIVE_LANDSCAPE_MAINTENANCE.md.

This page records what uniprot-mcp does that other Model Context Protocol servers in the biomedical space do not, as observed on 2026-04-26 in the reviewed set documented here. The bio-MCP ecosystem is moving fast; the table below will be updated when changes are detected. If you are the author of one of these servers and an entry is wrong or out of date, please file an issue and I will correct it.

The point of this page is not to disparage any other server — most are excellent at what they do, with broader scope or larger user bases than uniprot-mcp. The point is to be specific about the intersection of features the regulated-bio-pharma niche cares about, so an adopter can pick the right tool for their workflow.

Survey

Server	Domain	Tools	License	Last push (2026-04-26)	Per-response SHA-256	Verify primitive	Release pinning	SLSA / Sigstore / SBOM	.well-known/mcp.json
smaniches/uniprot-mcp (this)	UniProt + AlphaFold + ClinVar	41	Apache-2.0	2026-04-26	Yes	Yes (uniprot_provenance_verify)	Yes (--pin-release)	Yes (release.yml ships SLSA + Sigstore + CycloneDX)	Yes
genomoncology/biomcp	meta-router (gene/variant/trial/drug/protein/disease/...)	~30+ over 40+ upstreams	MIT	2026-04-26	No (installer SHA-256 only)	No	No	No	No
Augmented-Nature/UniProt-MCP-Server	UniProt	26	Custom (NOASSERTION)	2025-12-21	No	No	No	No	No
TakumiY235/uniprot-mcp-server	UniProt	~5	none	2025-03-11	No	No	No	No	No
BioContext/UniProt-MCP	UniProt	small	—	—	No	No	No	No	No
QuentinCody/uniprot-mcp-server	UniProt	small	—	—	No	No	No	No	No
Augmented-Nature/AlphaFold-MCP-Server	AlphaFold DB	22	MIT	2025-12-21	No	No	No	No	No
longevity-genie/biothings-mcp	mygene/myvariant/mychem	~15	MIT	2025-11-03	No	No	No	No	No
longevity-genie/gget-mcp	gget wrapper (BLAST/AlphaFold/Enrichr)	~15	MIT	2026	No	No	No	No	No
cyanheads/pubmed-mcp-server	NCBI E-utilities	9	Apache-2.0	2026-04-25	No	No	No	No	No
openags/paper-search-mcp	arXiv/PubMed/bioRxiv	~10	MIT	2026-04-21	No	No	No	No	No
JamesANZ/medical-mcp	FDA/WHO/PubMed/RxNorm	~12	MIT	2026-02-18	No	No	No	No	No
Cicatriiz/healthcare-mcp	FDA/PubMed/medRxiv/ICD-10	~15	MIT	2025-08-16	No	No	No	No	No
Proprius-Labs/pocketscout-mcp	UniProt+PDB+AlphaFold+ChEMBL+PubMed druggability	small	MIT	2026-03-13	No	No	No	No	No
donbr/lifesciences-research	OpenTargets/ChEMBL/UniProt agent wrappers	small	MIT	2026-02-26	No	No	No	No	No

ClinVar-specific MCP and a ChEMBL-only MCP were not found as standalone servers — only as facets of meta-routers.

What uniprot-mcp does that no server in the reviewed set does (as of 2026-04-26)

1. Per-response SHA-256 + canonicalised Provenance footer on every tool result (release tag, retrieval timestamp, resolved URL, body digest). No other bio-MCP in the reviewed set attaches a body digest. BioMCP attaches an installer SHA-256 only.
2. uniprot_provenance_verify re-fetch primitive with five enumerated verdicts (verified, release_drift, hash_drift, release_and_hash_drift, url_unreachable) and per-verdict advice. GitHub code search across repos finds zero other implementations of provenance_verify / hash_drift / release_drift semantics.
3. UNIPROT_PIN_RELEASE / --pin-release=YYYY_MM release pinning that raises on drift inside the running server. Not present in any competitor README.
4. uniprot_target_dossier nine-section composition (identity / function / chemistry / structure / drug-target / disease / variants / functional annotations / cross-refs) in one call.
5. uniprot_lookup_variant (HGVS) + uniprot_features_at_position + uniprot_get_alphafold_confidence (pLDDT bands) + uniprot_resolve_clinvar, all chained with the same provenance envelope.
6. UNIPROT_MCP_CACHE_DIR offline replay (uniprot_replay_from_cache) for air-gapped re-runs.
7. SLSA build provenance + Sigstore keyless signatures + CycloneDX SBOM on every release artefact. PyPI Trusted Publishing (OIDC) — no long-lived API tokens. Not present in any competitor scanned.
8. Pre-registered benchmark with SHA-256 commitments on main (tests/benchmark/expected.hashes.jsonl) — reviewer cannot rewrite "correct" answers post-hoc. 30 prompts; v1.1.0 measured run at 30/30.

Honest weaknesses vs alternatives

• Scope: BioMCP is bigger surface area (13 entities, 40+ upstreams). uniprot-mcp is single-database (UniProt) plus two cross-origin enrichments (AlphaFold DB, NCBI eutils for ClinVar).
• Tool overlap on basics: search / get-entry / sequence / GO / cross-refs are commodity; ~25 of the 41 tools are commodity wrappers and ~16 tools are the differentiated layer (provenance, verify, dossier, position-aware features, biomedical features, cache).
• Distribution presence: BioMCP is in the Anthropic Connectors Directory; Augmented-Nature is on Smithery. As of 2026-04-26 uniprot-mcp is on PyPI but not yet listed in either directory.
• Adoption: BioMCP has 497 GitHub stars; uniprot-mcp has 1. The Zenodo DOI argument is artefact merit, not adoption.

Adjacencies the same author could pursue

(documented for future planning, not commitments)

1. clinvar-mcp — same provenance/verify discipline, ClinVar release ID + per-record digest. Currently only available as a sub-route inside biothings-mcp / BioMCP.
2. pdb-mcp — direct RCSB with mmCIF / structure-factor digest. Currently only available as part of BioMCP and Augmented-Nature.
3. ensembl-mcp — gget wraps it, but no dedicated server.
4. reactome-mcp / kegg-mcp — pathway resources, no dedicated server.

Method

The survey was performed via:

• GitHub code search: bio mcp, uniprot mcp, bioinformatics mcp, provenance_verify, hash_drift, release_drift.
• MCP Registry (registry.modelcontextprotocol.io) browse.
• Smithery (smithery.ai) bio category browse.
• Anthropic Connectors Directory.
• PyPI search for *-mcp packages.

Each candidate's README was read for: per-response digest behaviour, verify-style primitives, release-pinning semantics, supply-chain attestations. Where the README was ambiguous, the source was inspected.

If you find a server I missed, please file an issue. I will update this page and adjust the README's claim accordingly. See COMPETITIVE_LANDSCAPE_MAINTENANCE.md for the full update procedure.