PubMed¶
"PubMed" is an umbrella label for BioMCP's PMID-centric literature workflow, so it is the starting point for most biomedical literature work: researchers get a shared identifier system, durable abstracts, and the fastest path from a gene, disease, or drug question to the papers that matter. If you want an MCP-friendly literature workflow that still speaks the language of PMIDs, this is the page to start with.
In BioMCP, PubMed is both a direct article-search source and part of the
default compatible article federation. search article --source pubmed uses
BioMCP's PubMed ESearch/ESummary loop directly, while the default --source
all route combines PubTator3, Europe PMC, and PubMed when the selected
filters are PubMed-compatible. variant articles --strategy union also sends a
bounded strict request with quoted gene and variant Title/Abstract clauses,
then retains the discovery federation; --debug-plan exposes the versioned
strict request. Direct PubMed search and the compatible federated PubMed leg
share the variant-article invocation's 60-second deadline; completed ESearch/
ESummary exchanges remain usable when later provider work times out. They
clean question-format unfielded article terms before ESearch; BioMCP keeps the
raw gene, disease, drug, or keyword wording in
markdown and JSON query echoes, and other article sources keep their existing
query behavior. The opt-in indexing section uses PubMed citation EFetch XML for associated author affiliations, ORCID, and structured MeSH headings; all includes it while ordinary detail/search/batch do not. Full-text resolution uses Europe PMC, NCBI E-utilities, PMC OA, NCBI ID Converter, PMC HTML, and opt-in Semantic Scholar PDF metadata; full text and PDFs remain governed by article-level licenses. Article JSON records the full-text ladder as not_requested, data, confirmed empty, or unavailable; a later successful source wins, but a healthy miss cannot erase an earlier source failure. Markdown and _meta.section_sources project the same outcome.
Semantic Scholar TLDR, citation, reference, and recommendation helpers belong
on the Semantic Scholar page because they come from a
different provider surface.
What BioMCP exposes¶
| Command | What BioMCP gets from this source | Integration note |
|---|---|---|
search article |
PMID-ranked literature search results with typed filters | Direct --source pubmed route plus default compatible federation with PubTator3 and Europe PMC |
get article <id> |
Article summary card with identifiers, journal, and abstract context | Uses Europe PMC metadata with BioMCP normalization |
get article <id> annotations |
PubTator entity annotations for a paper | PubTator3-only section |
get article <id> indexing |
Associated citation authors/affiliations/ORCID and structured MeSH headings | Opt-in PubMed citation XML; explicit status separates available-empty from unavailable; included by all |
get article <id> fulltext |
Open-access full-text handoff with saved Markdown path and rendered references when available | Uses Europe PMC, NCBI E-utilities, PMC OA, PMC HTML, and opt-in Semantic Scholar PDF fallbacks; NCBI ID Converter bridges PMID/DOI identifiers to PMCID before the PMCID-dependent source attempts |
article entities <pmid> |
Entity-grouped follow-up view for a PMID | Derived from PubTator3 annotation output |
Example commands¶
Returns an article table with PMID and title columns for a fast literature scan.
Returns an article card with PMID, journal, and summary metadata.
Returns a PubTator annotation section with entity groups and counts.
Use biomcp get article 22663011 indexing for PubMed citation indexing metadata
that preserves author-affiliation associations and MeSH descriptor/qualifier
flags.
Returns an entity-grouped follow-up view with separate genes, diseases, and drugs sections.
Returns a full-text section with a Saved to: cache path.
XML, PMC HTML, or explicitly opted-in PDF sources can resolve. JATS Markdown can
render references, figure captions, supplementary-material metadata, and complex-table
omission markers. Semantic Scholar PDF is attempted only when the caller passes --pdf.
PMC OA resolves a versioned S3 prefix listing to its flat <version>/<version>.json
metadata object, then downloads only metadata-declared XML and media objects from
https://pmc-oa-opendata.s3.amazonaws.com. The retired FTP/archive route is not
used: NLM removes its legacy files in August 2026. Scheme, origin, port, DNS
answers, and every redirect are checked before contact without exposing a rejected
URL in the public error. Each object is capped at 8 MiB; media is capped at 256
objects and 64 MiB aggregate payload. Route-resolution failures identify the
PMC OA package route; resource failures are sanitized as source unavailable.
API access¶
Optional NCBI_API_KEY for higher NCBI throughput. Set it through the API Keys guide and create one in My NCBI.
Official source¶
PubMed is the official NLM literature search surface most researchers already anchor on.