Bring your library
Zotero and Mendeley sync via OAuth. Drop a folder of PDFs. Drag from your file system. Whatever fits your existing workflow: we import, we don't replace.
Lit reviews that survive peer review from sources you control.
A source-grounded workspace for researchers: bring your papers, ask real questions, get answers that cite back to the page. Zotero and Mendeley connect directly, so your library is one click away.
No credit card. Free tier covers your first review.
Lit review isn't a search problem. It's a synthesis problem.
You don't need a better search engine. You need a workspace where 50 PDFs become an evidence table, where every claim links straight to the page it came from, where the draft you're writing pulls from the corpus you uploaded.
Most AI for research stops at the answer. The hard part is what comes next: the methods section, the comparison table, the cited paragraph in your draft.
Citation discipline, built into the engine.
Every claim our model makes is anchored to a specific page or section of a specific paper. Click the citation, the document opens at that location. No paraphrased footnotes. No "as the authors discuss" hand-waves.
How researchers use it
Built for citation-heavy work.
Focus on the synthesis. Let docAnalyzer handle the bookkeeping.
Ask once, across the corpus
The Individual workflow runs the same question against every paper in your dataset and returns per-paper answers. The Summarizer creates uniform digests across N papers in one batch.
Draft in canvas mode
Co-work lets you draft your review alongside the AI in a rich-text editor. The AI sees what you're writing and helps refine, with citations staying attached as you write.
What researchers actually push on.
AI summaries are never specific enough for academic work.
Push the context-adherence dial to high. At that setting, the model stays inside your dataset. Every claim is a clickable reference back to the page. If it can't cite, it doesn't ship.
I can't have my unpublished work used to train models.
Per-tenant isolation. No training on customer content. Your library is yours alone: the model sees it only for your queries, never for training.
Reference management: I'm not switching from Zotero.
You don't have to. OAuth-connect Zotero or Mendeley and pull from your existing collections. The library you've spent years curating stays exactly where it is.
My corpus is bigger than any one AI can read at once. Won't it miss things?
docAnalyzer doesn't run a single retrieval and hope for the best. It searches your corpus in rounds: pulls passages, reads them, decides whether the answer holds up against the sources, and runs more searches if it doesn't. A library of hundreds of papers or a stack of 300-page theses works the same way, and every cited claim still opens at the exact page.