PubMed articles search — abstracts, authors, MeSH (E-utilities)

One row per PubMed article from NCBI's public E-utilities (esearch + efetch, db=pubmed). Search with PubMed's own query syntax and a publication-date range; get PMID, title, abstract, authors, journal, volume/issue/pages, publication date, DOI, PMC id, MeSH terms, keywords, publication types, language and links — normalised, one flat row per article, ready for CSV or JSON.

The default input ("GLP-1 weight loss", last 365 days, 100 rows) succeeds within 5 minutes and returns rows — Apify auto-tests it daily.

Data source and status

Publisher	National Library of Medicine (NLM), National Center for Biotechnology Information (NCBI)
API	https://eutils.ncbi.nlm.nih.gov/entrez/eutils/ — public; no key needed (an optional free NCBI API key raises the rate limit). Documentation: https://www.ncbi.nlm.nih.gov/books/NBK25499/
What this Actor reads	esearch.fcgi (db=pubmed, term, retmax, sort, datetype=pdat, mindate/maxdate, JSON) for the PMIDs, then efetch.fcgi (retmode=xml, 100 PMIDs per request) for the records; every request carries tool=retrainmap and email=<contact> as NCBI's usage policy asks
Refresh cadence	PubMed is updated daily, Tuesday–Saturday (new citations appear each morning US Eastern). The Actor reads live on every run; fetched_at on each row is the fetch time
Terms / attribution	PubMed data are provided by NLM (https://www.nlm.nih.gov/web_policies.html); the E-utilities usage policy (https://www.ncbi.nlm.nih.gov/books/NBK25497/) asks for ≤ 3 requests/second without a key (10 with one), identification via tool/email, and that large jobs run on weekends or 9 pm–5 am US Eastern on weekdays. Abstracts remain copyrighted by their publishers; please cite PubMed/NLM as the source
Known caveats	esearch returns at most 10,000 ids per query — narrow the date range for more; abstract is empty when PubMed holds none (many older or non-English records) and is cut at 5,000 characters (abstract_truncated says so); mesh_terms and publication_types are empty until NLM indexes the citation (weeks to months after publication); pub_date follows the journal's PubDate and may be a year only or a range string like 2025 Jan-Feb; NCBI Bookshelf entries in the results are skipped and counted in the run summary

Honesty note: the rows are the PubMed XML's own elements flattened — nothing is inferred, summarised or classified by this Actor, and a run stops with an error rather than guess when the API's shape changes.

Identification: every request carries NCBI's tool and email parameters, a product-token User-Agent and the operator's contact address in the standard From: header (RFC 9110 §10.1.2). An API key you supply is sent only to eutils.ncbi.nlm.nih.gov and never logged.

Input

Field	Type	Meaning
query	string	PubMed search syntax: plain words, semaglutide[Title], diabetes[MeSH Terms], randomized controlled trial[Publication Type], Smith J[Author], AND/OR/NOT
date_from / date_to	YYYY-MM-DD	Publication date, inclusive. Both empty = the last 365 days; 1900-01-01 for no lower bound
sort	pub_date / relevance / most_recent	Default pub_date (newest first)
include_abstracts	boolean	Default true
max_records	integer	Default 200 (prefilled 100), at most 10,000
ncbi_api_key	secret string	Optional; 10 requests/second instead of 3
contact_email	string	NCBI email parameter and From: header

Example — semaglutide randomised controlled trials published in 2025, without abstracts:

{ "query": "semaglutide[Title] AND randomized controlled trial[Publication Type]", "date_from": "2025-01-01", "date_to": "2025-12-31", "include_abstracts": false, "max_records": 25 }

Output (dataset row)

pmid, title, abstract (labelled sections joined as LABEL: text, ≤ 5,000 chars), abstract_truncated, authors (ForeName LastName or collective name), journal, journal_abbrev, volume, issue, pages, pub_date, pub_year, epub_date, doi, pmc_id, mesh_terms, mesh_major, keywords, publication_types, language, pubmed_url, pmc_url, doi_url, fetched_at, source.

A run summary (filters, PubMed's own query translation and match count, efetch batches, skipped Bookshelf entries, esearch warnings, requests, whether the pay-per-event budget stopped the run) is stored as RUN_SUMMARY in the run's key-value store.

Pricing (pay per event)

Event	Price
run-start — once per run	$0.10
record — per row written	$0.005

The default 100-article pull costs $0.60; 1,000 articles cost $5.10. Rows stop when your run's maximum charge is reached; the run summary says so.

Operator

Steelyard Ventures LLC (RetrainMap) — info@steelyardclinical.com. Not affiliated with NLM, NCBI or the NIH. The Actor writes only to its own dataset and key-value store; it stores no credentials and sends nothing else.