Academic Paper Search: OpenAlex, Crossref, arXiv, PubMed

Search scholarly papers across OpenAlex, Crossref, arXiv and PubMed in one call. One row per work with title, authors, year, venue, DOI, abstract (where the licence allows), citation count and a free full-text link. $0.004 per 25 works. No login, no API key needed, no CAPTCHA.

Built for people and AI agents that need literature search as a single tool call: "find the most cited papers on retrieval augmented generation since 2023", "look up these 40 DOIs", "what has been published on CRISPR base editing this year, open access only".

Use cases

• Literature reviews and related-work sections: collect candidate papers on a topic from four indexes at once, already deduplicated by DOI and title.
• AI research agents and RAG pipelines: structured paper metadata with abstracts and open-access links, ready to feed into a retrieval index or a summariser.
• Citation tracking and bibliometrics: citation counts per work, with the source of each count stated.
• Reading lists and newsletters: sortBy: newest for the latest papers on a topic, sortBy: citations for the classics.
• DOI enrichment: turn a list of DOIs into titles, authors, venues, citation counts and open-access links.

Where the data comes from

Source	What it adds	Terms
OpenAlex	Abstracts, citation counts, open-access links, work type	Data is CC0
Crossref	DOI metadata: titles, authors, venue, year, citation counts	Bibliographic facts from the public REST API
arXiv	Preprints with abstracts and PDF links	Metadata via the arXiv API, CC0
PubMed	Biomedical records through NCBI E-utilities, PubMed Central links	NCBI usage policy: identified client, at most 3 requests a second
OpenCitations	Citation counts for works whose source gives none	Open citation index

Only each provider's documented public API is called, and requests are paced to the provider's own limits: one arXiv request every 3 seconds, under 3 NCBI requests a second, polite-pool identification (a contact address) on OpenAlex and Crossref, and any Crawl-delay a host publishes.

Authors are the names as printed in the bibliographic record. No emails, affiliations or ORCID iDs are returned. Abstracts come from OpenAlex and arXiv only. Crossref and PubMed abstracts are left out because publishers hold their copyright; the sourceUrl points to the record where the abstract is shown.

Input

Field	Meaning
queries	Topics or keywords, one per line. arXiv field prefixes (ti:, au:) and PubMed syntax work in those sources.
dois	Specific works by DOI or doi.org link, looked up in OpenAlex, then arXiv (for arXiv DOIs) and Crossref.
citationsFor	Citation-graph mode: DOIs whose references (works they cite) and/or citing works are returned as ordinary rows with relation (reference or citing) and seedDoi. OpenAlex first (referenced_works, cites: filter), OpenCitations when OpenAlex has no answer.
citationDirection	both (default), references or cited_by. maxResults caps each direction per DOI; citing works come most cited first.
sources	Any of openalex, crossref, arxiv, pubmed. Default: all four.
maxResults	Works per topic after merging (default 25, up to 500).
yearFrom, yearTo	Publication year range.
sortBy	relevance (default), citations or newest. arXiv and PubMed have no citation sort and fall back to relevance.
openAccessOnly	Keep only works with a free full-text link.
addCitationCounts	Fill missing citation counts from OpenCitations (up to 150 lookups per run). Default on.
openalexApiKey	Optional. A free OpenAlex key raises OpenAlex's daily budget. Without one, OpenAlex can stop answering when the shared free budget is used up, and the other sources still run.

Leave queries, dois and citationsFor empty to run a built-in example: 5 works on large language model agents (a few seconds, 5 x $0.00016).

{
  "queries": ["retrieval augmented generation", "crispr base editing"],
  "yearFrom": 2023,
  "sortBy": "citations",
  "maxResults": 25
}

Output

{
  "title": "Large language models encode clinical knowledge",
  "authors": ["Karan Singhal", "Shekoofeh Azizi"],
  "year": 2023,
  "venue": "Nature",
  "doi": "10.1038/s41586-023-06291-2",
  "abstract": "Large language models (LLMs) have demonstrated impressive capabilities...",
  "citationCount": 4091,
  "citationSource": "openalex",
  "openAccessUrl": "https://www.nature.com/articles/s41586-023-06291-2.pdf",
  "isOpenAccess": true,
  "workType": "article",
  "source": "openalex",
  "sources": ["openalex", "crossref"],
  "sourceId": "W4384071683",
  "sourceUrl": "https://openalex.org/W4384071683",
  "query": "retrieval augmented generation",
  "rank": 5,
  "scrapedAt": "2026-11-01T16:00:00+00:00"
}

Results from the sources are mixed in round-robin order, each source in its own ranking. A work found in several sources is one row, with the gaps (abstract, open-access link, citation count) filled from the others and sources listing all of them.

Pricing

Pay per event: $0.004 per 25 works ($0.00016 per work), charged for each work returned. Duplicates across sources are merged first and billed once. Worked examples:

• 25 works on one topic: $0.004.
• 4 topics at 50 works each (200 works): $0.032.
• 100 DOIs looked up: $0.016.
• Citation graph of one paper, 25 references and 25 citing works: $0.008.
• The empty-input example (5 works): $0.0008.

Set a maximum cost per run in the Apify run options; the Actor stops at that limit and says so in its status message.

Limits

• Each source returns its own ranking; Crossref matching is looser than OpenAlex's, so on narrow topics a few Crossref rows can be only loosely related. Use sources to leave a source out.
• A source that is down or out of budget is skipped and named in the run's status message; the others still return rows.
• OpenCitations counts only citations in its own index, so the count of an arXiv or PubMed work can be lower than Google Scholar's.
• Citation-graph rows are billed like any work, once per run: a work linked to two seed DOIs appears under the first. Reference lists are as complete as the OpenAlex and OpenCitations indexes.
• At most 500 works per topic per run.

FAQ

Do I need API keys? No. An optional OpenAlex key avoids the shared free daily budget.

Can I get full text? The openAccessUrl links to a free copy when one is known. The Actor does not download or store the papers.

Is the same paper billed twice? No. Works are merged by DOI and by title across sources and across the topics of one run, and billed once.

Why is there no abstract on some rows? Abstracts are returned only where the licence allows (OpenAlex, arXiv). Crossref and PubMed abstracts are publisher copyright.

Responsible use

The data is public bibliographic metadata from open APIs. You are responsible for how you use it, including the copyright of the papers behind the links and any personal-data rules that apply to author names in your jurisdiction.

Use the output for literature review, research monitoring and citation work. It is not for building outreach, marketing or recruiting lists of researchers: the Actor returns author names only as printed in the record, with no emails, affiliations or ORCID iDs.

Removal requests

If you are named in a record and want your name left out of this Actor's output, email datagleaner@gmail.com with the DOI or title. We answer within 30 days, record each request with its date and answer, and add a filter for it in the next release. The Actor keeps no copy of the data, and it cannot change the source records at OpenAlex, Crossref, arXiv or PubMed; ask those services to correct a record at the source.