Search UniProtKB by protein function, fetch curated records, map IDs across databases, and pull reference proteomes, taxonomy, and sequences via MCP. STDIO or Streamable HTTP.
Five tools with strong naming (verb_noun pattern), comprehensive descriptions (150 - 400 chars), and detailed parameter schemas. All parameters have types and descriptions. Output schemas are documented inline in descriptions. Error handling is present but generic. Tool composition is sound, each tool has one clear responsibility and outputs chain into downstream tools (e.g., uniprot_search_proteins.results[].accession → uniprot_get_entry). Descriptions are LLM-optimized and include dependency hints ('Accessions come from uniprot_search_proteins.results[].accession or uniprot_map_ids'). Minor gaps: no explicit enum constraints for database names in uniprot_map_ids (from_db/to_db accept free-form strings), and error recovery guidance is minimal. No security concerns (read-only tools, no secrets in params).
Fetch full curated UniProtKB entries by accession in one batch (up to 20). Each entry carries function, catalytic activity, cofactors, subcellular location, disease involvement, PTMs, natural variants, isoforms, domains, GO terms, keywords, and cross-references. Partial failures do not abort the batch — resolved entries land in succeeded[] and unknown/withdrawn accessions in failed[]. Pass fields to trim the upstream projection. A single oversized record returns kind: "outline" (a section listing with byte sizes) instead of overflowing context — re-call the same accession with sections:[...] (e.g. ["disease","variants"]) to pull only those. This tool does not search: accessions come from uniprot_search_proteins.results[].accession or uniprot_map_ids.
Fetch the reference proteome for an organism by UPID (e.g. "UP000005640") or NCBI taxon ID (e.g. 9606) — provide exactly one. Returns metadata inline: proteome type, total protein count, BUSCO completeness (score, complete/fragmented/missing counts, lineage dataset), and the genome assembly accession. The protein set is opt-in via include_proteins (it is large — human is ~147,506) and returns a capped page with a forward cursor; narrow it with the query filter (UniProtKB Lucene syntax) for a subset. Resolve an organism name to a taxon ID first with uniprot_get_taxonomy.
Fetch the canonical amino-acid sequence (FASTA) for a UniProtKB accession, with length and the parsed header. Set include_isoforms to also return the alternatively-spliced isoform sequences. This is the cheap sequence-only path — for the full functional record use uniprot_get_entry. Accessions come from uniprot_search_proteins or uniprot_map_ids; strip any "-N" isoform suffix (P04637-2 to P04637) before calling.
uniprot_map_ids from_db and to_db parameters lack enum constraints. Free-form strings invite hallucinated database names (e.g., 'UniProt_Swiss' instead of 'UniProtKB-Swiss-Prot'). Should declare valid values as enums: ['Gene_Name', 'UniProtKB_AC-ID', 'Ensembl', 'PDB', 'RefSeq_Nucleotide', 'RefSeq_Protein', 'ChEMBL', 'GeneID'] for from_db and ['UniProtKB-Swiss-Prot', 'UniProtKB', 'UniProtKB_AC-ID'] for to_db.
Error handling is minimal. Tools return generic API errors without recovery guidance. E.g., if uniprot_get_entry receives an invalid accession, the response should suggest 'Try uniprot_search_proteins() to find valid accessions' rather than a bare error code.
uniprot_get_proteome query parameter accepts 'UniProtKB Lucene syntax' but does not document valid field names or operators. LLMs cannot construct valid queries without examples or a reference. Should include: 'Valid fields: reviewed, keyword, organism, length. Operators: AND, OR, NOT. Example: "reviewed:true AND keyword:KW-0067"'.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 68 | 2026-07-28+ | v2 |
Resolve a taxonomy record by NCBI taxon ID (e.g. 9606) or scientific name (e.g. "Homo sapiens") — provide exactly one. Returns the scientific and common name, mnemonic, rank, parent, and the full lineage. Set include_children to also fetch immediate child taxa (a separate lookup — not inline on the record). Use this to turn an organism name into the taxon ID that uniprot_search_proteins (organism_id) and uniprot_get_proteome (taxon_id) expect.
Translate identifiers across databases via UniProt's ID-mapping service — gene names to accessions, accession to PDB / Ensembl / RefSeq / ChEMBL / GeneID, and back. The job runs asynchronously; this tool submits it and polls within a budget. A running job returns status "running" with a ticket; pass that ticket alone to poll the same job. A completed call returns status "finished" with one results page; when continuation is present, pass it alone to fetch the next completed page without re-submitting or polling the job. A gene name often maps to one reviewed Swiss-Prot accession plus dozens of unreviewed TrEMBL ones, so target UniProtKB-Swiss-Prot (reviewed only) for the usual intent, or UniProtKB / UniProtKB_AC-ID to include TrEMBL. Pair a gene-symbol from_db with tax_id to disambiguate species. Chain the resulting accessions into uniprot_get_entry.
uniprot_map_ids has complex state management (ticket, continuation, submission inputs). Description does not clearly explain the state machine: when to pass ticket alone vs. continuation alone vs. new submission. LLMs may pass conflicting parameters.
No tool explicitly documents rate limits, timeouts, or batch size caps. uniprot_get_entry accepts 'up to 20' accessions but does not specify what happens if 21 are passed (truncation, error, partial success). Should state: 'If >20 accessions provided, returns error: "Max 20 accessions per call. Received N."'