Academic Paper Search API: Crossref, arXiv & DOI Lookup

Use when: an AI agent or research pipeline needs real, checkable citations: find papers on a topic (journals, books, preprints), or verify that a DOI exists and get its title, authors, year, venue and citation count, instead of trusting a model's memory.

Try this input

{"requests": [{"query": "large language model hallucination", "from_year": 2024, "sort": "cited", "max": 3}, {"doi": "10.1038/nature14539"}, {"doi": "10.48550/arXiv.1706.03762"}]}

Sample output (real run, 30 Oct 2026, trimmed)

{"query": "large language model hallucination", "count": 6, "results": [
  {"title": "A Survey on Evaluation of Large Language Models", "year": 2024, "venue": "ACM Transactions on Intelligent Systems and Technology", "doi": "10.1145/3641289", "citations": 3001, "via": "Crossref"},
  {"title": "A Survey on Hallucination in Large Language Models: Principles, Taxonomy, ...", "year": 2025, "doi": "10.1145/3703155", "citations": 2389, "via": "Crossref"},
  {"title": "Hallucination Detection with Small Language Models", "year": 2025, "venue": "arXiv", "doi": "10.48550/arXiv.2506.22486", "pdf_url": "https://arxiv.org/pdf/2506.22486...", "via": "arXiv"}], "error": null}
{"query": "10.1038/nature14539", "count": 1, "results": [{"title": "Deep learning", "authors": ["Yann LeCun", "Yoshua Bengio", "Geoffrey Hinton"], "year": 2015, "venue": "Nature",
  "type": "journal-article", "citations": 78904, "publisher": "Springer Science and Business Media LLC", "pdf_url": "http://www.nature.com/articles/nature14539.pdf", "via": "Crossref"}], "error": null}
{"query": "10.48550/arXiv.1706.03762", "count": 1, "results": [{"title": "Attention Is All You Need", "authors": ["Ashish Vaswani", "Noam Shazeer", "..."], "year": 2017, "publisher": "arXiv", "type": "article", "abstract": "The dominant sequence transduction models...", "via": "DOI registry (DataCite/other)"}], "error": null}

Options (per item in requests):

• query: topic, title or keywords (a DOI pasted in a query is detected and looked up).
• doi: look up one DOI (Crossref first, then doi.org for DataCite/other registries such as arXiv and Zenodo).
• max: results per source, default 10, max 50 (Crossref up to max, arXiv up to 10; combined list capped at 2 × max).
• from_year: only works published from this year (Crossref).
• sort: relevance (default), newest or cited (Crossref).
• include_arxiv: default true; set false for Crossref only.

Each paper has title, authors (first 15), year, date, venue, doi, url, type, citations, publisher, abstract (when deposited, up to 1500 chars), pdf_url, license and via. Up to 300 items per run; no API key needed.

Pricing: pay per event, $0.003 per paper search (one topic search or DOI lookup). Rows with errors (e.g. DOI not found, empty query) are not charged.

Limits & failure modes

• from_year and sort apply to Crossref only; arXiv results are appended by relevance and may be older or unsorted.
• Citation counts are Crossref's is-referenced-by-count (lower than Google Scholar); arXiv and DataCite records have none.
• Many Crossref records have no abstract; publishers decide what they deposit.
• Crossref relevance is keyword-based, so broad queries can return loosely related highly cited papers. Use specific phrases.
• Crossref or arXiv rate limits can occasionally make a row fail with "source unavailable" (not charged); retry it.
• Metadata from Crossref REST API and arXiv API (CC0). Thank you to arXiv for use of its open access interoperability. Abstracts and full texts belong to their authors and publishers.

Last tested: 30 Oct 2026.

Use from AI agents (MCP)

• Apify MCP server: add https://mcp.apify.com/?tools=tudorlane/academic-paper-search to Claude, Cursor, ChatGPT, VS Code or any MCP client.
• REST: POST https://api.apify.com/v2/acts/tudorlane~academic-paper-search/run-sync-get-dataset-items?token=YOUR_TOKEN with the input JSON.