UniProt Protein Research — tool reference

Server URL: https://toolcargo.com/mcp/uniprot · Scope: connector:uniprot. Activate Site Audit free and connect with ToolCargo OAuth or a scoped API key. No UniProt account, provider key or paid upstream. Each successful page or accession lookup uses one shared-plan call.

uniprot_gene_search

{ "gene": "HLA-A", "organismId": 9606, "reviewed": "reviewed", "limit": 5 }

Choose a literal gene symbol and an exact organism taxonomy ID. organismId selects that organism, not all taxonomic descendants. Reviewed, unreviewed or any-status records are explicit choices; reviewed is the default. No free-form query, wildcards, similarity search or sequence input.

Gene matching may be case-insensitive. Preserve returned source spelling and inspect the selected record rather than assuming one gene produces one protein. Search returns up to ten canonical-record candidates. Use the returned continuation token with unchanged gene, organism, reviewed choice and limit. Paging is explicit; a source-release change requires a fresh search. Tokens and totals do not establish a permanently pinned snapshot.

uniprot_protein_details

{ "accession": "P01308" }

Use an exact uppercase primary UniProt accession, not an entry name, URL, secondary identifier or isoform suffix. Returned primary identity must match. Inactive or deleted records remain explicit. Redirects are not followed to substitute another record; source not-found states do not establish that a protein cannot exist.

Names, evidence and length

Inspect protein names with their source naming category, gene names, organism identity, reviewed class and protein-existence classification. Reviewed means manual UniProt annotation; unreviewed means computer annotation. Neither is a clinical approval or general quality score. Protein-existence evidence does not establish sequence correctness.

Only canonical sequence-length metadata is returned, with precursor context where available. No amino-acid sequence residues. For example, the P01308 insulin record describes a canonical precursor length of 110; that is not the length of a mature insulin chain or every isoform. Function and disease narratives, features, structures and linked files are excluded.

Versions and source dates

Keep UniProt release, release date, API deployment date, entry version, sequence version and annotation-update date distinct. Retrieval time is when ToolCargo read the response. Release metadata does not make separately requested pages or records an atomic snapshot; do not infer immutable content or a source update time from a retrieval timestamp.

Access, attribution and bounds

Source: UniProt Consortium, UniProtKB. Copyrightable database material uses CC BY 4.0. Preserve credit, source links, license and the notice that ToolCargo selects, bounds and restructures the metadata. Linked resources and patent or other rights remain separate; no endorsement or clinical suitability is implied.

UniProt provides the source API freely. ToolCargo fees concern hosted access and processing, with no added restrictions on the underlying licensed material. Public gene terms and accessions reach UniProt; ToolCargo account identifiers and credentials do not.

Fixed UniProt REST endpoints and selected JSON fields, one bounded source request per operation, shared provider pacing and source Retry-After cooldown. No automatic paging, retries, redirects, result cache, stale fallback, bulk streams, sequences or arbitrary URLs. Source text is inert metadata, not instructions for your agent.

Primary references: API queries · Accession identities · Protein existence · Canonical records and isoforms · License and disclaimer.