Crossref Scraper - Academic Papers & DOI Lookup

Search scholarly works by keyword or look up a list of DOIs, and get clean, structured metadata: title, journal, publisher, publication date, citation count, licenses and full-text links. The Actor reads the public Crossref REST API, so it needs no browser, no proxy, no API key and no login.

Use it to:

• build a reading list on a topic, sorted by relevance, citations or date,
• enrich a list of DOIs from a spreadsheet or reference manager with journal, publisher and citation data,
• feed a literature-review pipeline, RAG system or AI agent (n8n, Make, MCP) with a legal, stable source of paper metadata,
• run a weekly search for the newest papers on a topic and send the results to a webhook.

What you get

• One row per work with 20 fields: DOI, link, title, journal or book title, publisher, work type, publication date, citation count, reference count, number of authors, ISSN, ISBN, volume, issue, pages, license URLs, full-text links and funder names.
• Filters: publication year range and work type (journal article, conference paper, book chapter, book, preprint, dataset, report, dissertation).
• Sorting: by relevance, most cited, or newest.
• Export as JSON, CSV, Excel, XML or HTML from the Apify dataset, or read it through the Apify API.
• Run summary: the RUN_SUMMARY key in the key-value store lists requests, results and the status of every input.

How to use it

1. Search a topic

{
  "searchQueries": ["large language models"],
  "maxItemsPerQuery": 50,
  "fromYear": 2023,
  "workType": "journal-article",
  "sortBy": "mostCited"
}

2. Look up DOIs

{
  "dois": ["10.1038/nature14539", "https://doi.org/10.1126/science.ade2574"]
}

DOIs can be bare, start with doi:, or be full doi.org links. Each DOI returns one row.

3. Weekly new papers

Create an Apify Schedule that runs the Actor with "sortBy": "newest" and "fromYear" set to the current year, then connect a webhook or integration to receive the dataset. The Actor does not remember earlier runs, so remove duplicates by doi on your side.

Input

Field	Description
searchQueries	Keywords, titles, journals or years to search for. Each query is run separately.
dois	DOIs or DOI links to look up.
maxItemsPerQuery	Results per search query, 1 to 1000 (default 50). A DOI always gives one result.
fromYear, untilYear	Only works published in this year range. Searches only.
workType	any or one work type. Searches only.
sortBy	relevance (default), mostCited or newest. Searches only.

Provide at least one search query or DOI. Duplicate queries and DOIs are merged, and values that are not DOIs are skipped with a warning.

Output

Every result is one dataset item. The columns of the default table view are title, journal, published, citations, type, DOI and link.

{
  "doi": "10.1126/science.ade2574",
  "url": "https://doi.org/10.1126/science.ade2574",
  "title": "Evolutionary-scale prediction of atomic-level protein structure with a language model",
  "containerTitle": "Science",
  "publisher": "American Association for the Advancement of Science (AAAS)",
  "type": "journal-article",
  "publishedDate": "2023-03-17",
  "citationCount": 6228,
  "authorCount": 15,
  "inputType": "searchQuery",
  "input": "large language models"
}

The full item has 20 fields; see sample-output.json for a real example.

Field notes

• publishedDate is YYYY, YYYY-MM or YYYY-MM-DD, depending on what the publisher deposited. It is the earlier of the print and online dates, so for works at the edge of your year filter it can fall one year before fromYear.
• citationCount is the number of citations known to Crossref, not Google Scholar or Scopus.
• newest ignores records with future dates: unless you set an earlier untilYear, results are capped at today.
• fullTextLinks and licenseUrls are the links publishers deposit. Many are for text mining and may need a subscription or a license.
• Abstracts, author names, ORCID iDs and affiliations are deliberately not returned; only authorCount is. See the Legal section.
• inputType and input tell you which query or DOI produced a row.

Pricing

The Actor uses pay-per-event pricing: you pay per result, meaning one saved dataset item. Empty searches, DOIs that do not exist and failed inputs are not saved and not charged. You can test with a small maxItemsPerQuery. If you set a maximum cost per run, the Actor stops cleanly when the limit is reached and keeps what it already saved. The current price is shown on the Store page.

Speed and limits

• The Actor sends about one request per second, one at a time, as asked by the rate-limit headers of the public Crossref API. A request returns up to 100 results, so 1,000 results take roughly 10 requests, about half a minute.
• On HTTP 429 or temporary errors it waits and retries up to three times. On HTTP 403 it stops instead of trying to get around the block.
• Up to 1,000 results per query. This is a search tool, not a full database dump.
• Only Crossref metadata is returned. Items that are not registered with Crossref (for example many arXiv-only papers) are not found.

Legal

This is an unofficial tool. It is not affiliated with or endorsed by Crossref and does not use Crossref logos or trademarks.

The Actor only calls the public Crossref REST API, which according to the Crossref documentation can be used by anyone: "Anyone can access the REST API, no signup or registration is required". Crossref also states that "The majority of metadata is considered to be 'facts' which are not copyrightable and are thus in the public domain (CC0)", with the exception of abstracts, which are subject to publisher or author copyright. For that reason the Actor does not return abstracts, and it does not return author names, ORCID iDs or affiliations (personal data). It only fetches the queries and DOIs you provide, with no bulk crawl mode, and it identifies itself with a User-Agent that names this Actor.

You are responsible for using the data lawfully, including applicable laws and the license of each publisher (see licenseUrls) before you use any full-text link. This section is not legal advice.

FAQ

Do I need a Crossref account or API key?

No. The Actor uses the public API pool without any key.

Why are some results not about my topic?

Crossref ranks by bibliographic match on titles, journals and authors, not by meaning. Use more specific phrases, a year range or a work type to narrow results.

Can I get abstracts or author names?

No. Abstracts are subject to copyright and names are personal data, so the Actor returns only metadata about the work. To read a paper, open the url field.

Can I get a larger result set?

Set maxItemsPerQuery up to 1000 per query and add more queries, or split a topic by year with fromYear and untilYear.