AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

database-lookup

Queries documented public database APIs with explicit endpoints, filters, pagination, and provenance. Used when a scientific, regulatory, financial, or other database-backed fact must be retrieved reproducibly from a named source rather than inferred from general knowledge.

Original source
K-Dense-AI
License
MIT
Pinned revision
5f2bc0341354fd5590aef76a533ada1e1e5b4e62

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--database-lookup at revision 5f2bc0341354fd5590aef76a533ada1e1e5b4e62, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 83

references/ncbi-protein.md

NCBI Protein API Reference

Overview

Protein sequence records (RefSeq, GenBank, UniProt imports) accessible via NCBI E-utilities with db=protein.

Base URL

https://eutils.ncbi.nlm.nih.gov/entrez/eutils/

Authentication

  • API key (recommended): Register at https://www.ncbi.nlm.nih.gov/account/ and append &api_key=YOUR_KEY.
  • Without key: 3 requests/second. With key: 10 requests/second.
  • Provide tool and email parameters for identification.

Key Endpoints

1. ESearch -- Search protein records

GET esearch.fcgi?db=protein&term=QUERY&retmax=N&retmode=json

| Param | Description | |-------|-------------| | term | Search query (Entrez syntax). Fields: [Protein Name], [Organism], [Accession], [Gene Name] | | retmax | Max UIDs per response (default 20); use small batches and history for large sets. | | idtype | Set acc to return accession.version IDs instead of default numeric UIDs. | | retstart | Offset for pagination | | usehistory | y to store results on server (use with large sets) |

Example -- search human insulin:

GET esearch.fcgi?db=protein&term=insulin+AND+homo+sapiens[Organism]&retmax=5&retmode=json

Response (JSON):

{
  "esearchresult": {
    "count": "1523",
    "retmax": "5",
    "idlist": ["116734704", "1240286123", "110591076", ...],
    "querytranslation": "insulin AND \"Homo sapiens\"[Organism]"
  }
}

2. EFetch -- Retrieve protein records

GET efetch.fcgi?db=protein&id=IDS&rettype=TYPE&retmode=MODE

| rettype | retmode | Output | |---------|---------|--------| | fasta | text | FASTA sequence | | gp | text | GenPept flat file | | gp | xml | GenPept XML (INSDSeq) | | acc | text | Accession list | | seqid | text | SeqID list | | ft | text | Feature table |

Example -- fetch FASTA for NP_000198.1 (human insulin):

GET efetch.fcgi?db=protein&id=NP_000198.1&rettype=fasta&retmode=text

Response:

>NP_000198.1 insulin preproprotein [Homo sapiens]
MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAED
LQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN

Example -- fetch GenPept XML for multiple IDs:

GET efetch.fcgi?db=protein&id=NP_000198.1,NP_001278826.1&rettype=gp&retmode=xml

3. ESummary -- Brief record summaries

GET esummary.fcgi?db=protein&id=IDS&retmode=json

Returns: accession, title, organism, length, taxonomy, create/update dates.

4. ELink -- Find related records

GET elink.fcgi?dbfrom=protein&db=gene&id=NP_000198.1&retmode=json

Links protein to gene, nucleotide, structure, taxonomy, etc.

Common Search Patterns

# By accession
term=NP_000198.1[Accession]

# By gene name + organism
term=BRCA1[Gene Name] AND human[Organism]

# RefSeq only
term=insulin AND srcdb_refseq[Properties]

# By sequence length range
term=100:500[Sequence Length] AND kinase[Protein Name]

ELink resolves protein accessions as well as numeric UIDs: the live NP_000198.1 probe resolved UID 4557671 and Gene 3630. Preserve both the accession version and the resolved UID in provenance.

Rate Limits

  • Without API key: 3 requests/second
  • With API key: 10 requests/second
  • Large batch downloads: use usehistory=y with WebEnv/query_key, then fetch in chunks of 500