AGENTIC MEMORY

Skills

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

← Back to Skills

paper-lookup

Searches 18 scholarly APIs for papers, preprints, citations, open-access full text, repository records, and journal OA status, and returns results with reproducible provenance. Covers PubMed, PMC, Europe PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, Unpaywall, OpenCitations, PubTator3, Zenodo, Figshare, ROR, BioStudies, and DOAJ. Use when searching for papers, citations, DOI/PMID/arXiv lookups, abstracts, full text, open-access PDFs, preprints, citation graphs, author publications, biomedical entity annotations, deposited records (Zenodo, Figshare, BioStudies), institution ROR IDs, or any scholarly literature query. Triggers on mentions of any supported database or requests like "find papers on X", "look up this DOI", "who cites this paper", or "get me the PDF".

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.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--paper-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 · 19

references/pubtator.md

PubTator3

NCBI text-mined annotations on PubMed abstracts (and some PMC full text): genes, diseases, chemicals, species, variants, and cell lines, plus typed relations. Use it when the user wants entities in a paper or papers that mention a normalized entity (@CHEMICAL_remdesivir), not when they want a citation list.

Europe PMC has a separate Annotations API. PubMed search does not normalize "remdesivir" to a concept ID. PubTator3 APIs are not the old PubTator / CBBresearch endpoints.

All figures below verified 2026-09-10.

Base URL

https://www.ncbi.nlm.nih.gov/research/pubtator3-api

Docs: https://www.ncbi.nlm.nih.gov/research/pubtator3/api

Authentication

None.

Rate Limits

No more than 3 requests per second. Serialize. For bulk annotation dumps use the FTP site (https://ftp.ncbi.nlm.nih.gov/pub/lu/PubTator3/) rather than the API.

Key Endpoints

1. Resolve a mention to an entity ID

GET /entity/autocomplete/?query={text}&concept={type}&limit={n}

concept is optional (chemical, disease, gene, species, variant, cellline).

GET /entity/autocomplete/?query=remdesivir&limit=3

Returns a JSON array. First hit (verified):

{
  "_id": "@CHEMICAL_remdesivir",
  "biotype": "chemical",
  "db_id": "C000606551",
  "db": "ncbi_mesh",
  "name": "remdesivir"
}

Search with _id (@CHEMICAL_remdesivir), not the display name. Nearby synonyms (@CHEMICAL_GS_441524_triphosphate) are different entities.

2. Search papers by text, entity, or relation

GET /search/?text={query}&page={n}

text may be free text, an @TYPE_name entity ID, a boolean combination, or a relation:

@CHEMICAL_Doxorubicin AND @DISEASE_Neoplasms
relations:ANY|@CHEMICAL_Doxorubicin|@DISEASE_Neoplasms
relations:ANY|@CHEMICAL_Doxorubicin|DISEASE
GET /search/?text=@CHEMICAL_remdesivir

Verified: count 23295, page_size 10, current 1, total_pages 2330, results length 10. First result pmid is an integer (37711410); _id is a string. Page with page (1-based). There is no cursor.

Do not use this as a general PubMed replacement. Rank is entity-centric.

3. Export annotations for PMIDs

GET /publications/export/{format}?pmids={id,id}&full={true|false}

format is pubtator, biocxml, or biocjson. full=true (PMC full text) works only for biocxml / biocjson.

GET /publications/export/biocjson?pmids=29355051

The JSON is not a bare BioC document. It is:

{ "PubTator3": [ { "_id": "29355051|None", "id": "29355051", "passages": [...], "relations": [...] } ] }

Verified on PMID 29355051: one document, two passages, first passage has 5 annotations. An annotation looks like:

{
  "infons": {
    "type": "Species",
    "database": "ncbi_taxonomy",
    "normalized_id": 112863,
    "biotype": "species"
  },
  "text": "Lycium barbarum",
  "locations": [{"offset": 14, "length": 15}]
}

Read PubTator3[0].passages[].annotations. A 200 with "PubTator3": [] is "no documents," not a transport success you can ignore.

4. Related entities

GET /relations?e1={entityId}&type={relation}&e2={entity_type}

Relation types include treat, cause, interact, associate, positive_correlate, negative_correlate, prevent, inhibit, stimulate, drug_interact, cotreat, convert, and compare. e1 must be an autocomplete _id; type and e2 are optional. Related entity types include gene, disease, chemical, and variant.

Typical Workflow

  1. Autocomplete the user's string → @CHEMICAL_… / @DISEASE_….
  2. Search with that ID (and a relation if they asked "what does X treat?").
  3. Export biocjson for the PMIDs you will report, and list the annotations.
  4. For the paper itself (abstract, OA PDF), go to PubMed / Europe PMC / Unpaywall. PubTator full-text annotations cover a subset; check returned passages before claiming full-text coverage.

Failure Modes

| What you did | What happens | What to do | |---|---|---| | Parsed export JSON as BioC root | No passages at the top level | Descend into PubTator3 | | Searched remdesivir and treated hits as exact-chemical papers | Keyword search, not the concept | Autocomplete, then search @CHEMICAL_remdesivir | | Called the old pubtator-api or CBBresearch URL | May still 200, but the contract changed | Use pubtator3-api | | full=true with format=pubtator | No full text | Use biocjson or biocxml | | Parallel fan-out | Easy to exceed 3 req/s | Serialize |

Official sources reviewed 2026-09-30

  • https://www.ncbi.nlm.nih.gov/research/pubtator3/api