Turn a paper's reference list into a .bib file, as printed
Upload a paper and CruxDoc reads its reference list end to end, structures every cited work as authors, title, year and venue, and downloads it as BibTeX. Nothing is invented and nothing is auto-completed: what the paper prints is what you get, with the gaps visible.
Drop a document to get started
PDF, Word, PowerPoint, text, images, audio or video. Up to 20 MB and 50 pages. No account needed for your first one.
Processing continues if you close the tab.
Without an account: 1 document, up to 50 pages, kept for 24 hours. Sign up any time and it comes with you.
How it works
Upload the paper
PDF, Word, or a scan read with OCR.
Run Reference list
Every cited work, structured as printed. Per-chapter lists included.
Download the .bib
One click. Import into LaTeX, Zotero, Mendeley, or anything that reads BibTeX.
Check what is flagged
Entries that could not be matched to the page are labelled, never repaired.
Nothing invented, nothing completed
A bibliography tool that guesses a missing year or a plausible venue produces citations that look right and are wrong, which is worse than a blank. CruxDoc structures only what the paper prints and string-matches each entry against the paper's text on the server. An entry that fails the check is labelled, so the gap is yours to fill on purpose.
The same engine builds the paper snapshot and the literature matrix, so the papers you cite and the papers you compared are the same uploads.
Frequently asked questions
Q.01What does the BibTeX extractor produce?
A .bib file with one entry per cited work: authors, title, year and venue as they are printed in the paper, ready for LaTeX, Zotero, or any reference manager that imports BibTeX.
Q.02Does it look up DOIs or fill in missing fields?
No, deliberately. The extractor structures what the paper prints and nothing more. A reference that is incomplete in the paper is incomplete in the .bib file, where you can see it, rather than completed with a guess that looks authoritative.
Q.03What about garbled or unusual references?
Each entry is checked against the paper text before it reaches you; an entry that cannot be matched is labelled instead of silently dropped or repaired. Scanned papers go through OCR and low-confidence pages are flagged, so you know which entries to check by eye.
Q.04Which formats work?
PDF, Word documents, and scanned PDFs or images read with OCR. Reference lists at the end of a paper and per-chapter lists are both read, because the whole document is.
Q.05Is it free?
Your first document needs no account. The reference list tool lives in the Research workspace of a free account, which includes 100 questions a month; each extraction uses one, and the run stays in your history to reopen or download again.
Extract the references from a paper
Drop a document to get started
PDF, Word, PowerPoint, text, images, audio or video. Up to 20 MB and 50 pages. Free. No account needed for your first one.
Processing continues if you close the tab.
No account needed for your first document.