Ask a research question. Cognify searches the published literature, reads the papers and writes up what they found. Every claim links to its source, and every verdict is calculated from the evidence.
Does creatine improve short-term memory in healthy adults?
Leaning supported
62% confidenceCalculated from the evidence table, weighted by study strength.
Try it free
Drop a paper, report or contract. Answers cite the page they came from.
Free plan · 500 MB storage
Built on open scholarly data and proven infrastructure
How a run works
Example run
0 found · 0 kept
Leaning supported
0% confidenceIllustration with example data. Scroll to play it back.
Research
Each one runs on real papers from OpenAlex, Crossref, arXiv and Europe PMC, with DOIs, abstracts and citation counts. A run keeps going if you close the tab, and every run is saved with its sources.
“What does the research say about remote work and productivity?”
Plans the searches, screens every result for relevance, extracts what each included paper found and writes a review grouped by theme. It ends with where the papers disagree and where the evidence runs out.
“Does creatine improve short-term memory?”
Tests one claim. It searches for evidence on both sides on purpose, classifies each paper as supporting, contradicting or mixed, and calculates a verdict with a confidence score.
“How large is the effect of tutoring on maths scores?”
Pulls reported effect sizes into a comparison table and pools them with random effects. You get the pooled estimate, I² heterogeneity and a prediction interval, with each study's weight shown.
“What is known about microplastics in drinking water?”
Searches, reads and judges its own coverage, then searches again to fill the gaps it found. When something can't be established from the papers, it says so.
Workspace
A research project keeps everything you learn in one place and links it together, so findings build up over months instead of disappearing into chat history.
Each project starts with a research question and the hypotheses you want to test.
Save papers from a run or a search. Cognify extracts each paper's method, data, findings and limitations, and scores how reproducible it is.
Claims are pulled from papers with the sentence that states them. Write your own, and weigh every paper in the project for or against it.
Find what the papers leave unsettled, each gap tied to the papers that imply it, and turn any gap into an experiment design you can edit.
Draft by section with citations to your papers. Cognify checks that each cited paper supports the sentence citing it, and flags claims with no source.
How it stays reliable
Evidence verdicts, pooled estimates and reproducibility scores are computed in code from the evidence table. The model explains the result; it doesn't decide it.
A claim is only kept if the model can quote the sentence that states it. Every finding links to its paper and passage.
If an index is down or a step can't finish, the run says so. A rate limit never turns into a confident "no relevant papers found".
Row-level security in Postgres means each request can only reach rows you own or were invited to.
Your documents
Upload papers, reports or contracts and ask questions about them. Answers name the page they used, and clicking it opens the page. Search across your whole library by meaning, highlight as you read, and share files, tasks and meetings with your team.
500 MB of storage on the free plan.
Free tools
Seventeen everyday tools. Most run entirely in your browser, and the rest process your file in memory and discard it as soon as the download starts.
OpenAlex, Crossref, arXiv and Europe PMC. Together they index well over 200 million scholarly works. Cognify reads titles and abstracts, and links you to the full text wherever it's open access.
Cognify only cites papers it actually retrieved from those indexes, each with a link you can follow. Claims must quote the passage that supports them, and the manuscript checker tests whether a cited paper really backs the sentence citing it.
Usually five to ten minutes, depending on the pipeline and how many papers it reads. You can close the tab; the run continues and is saved.
No. It is a fast, transparent first pass that shows its sources and its limits. For a formal systematic review, use it to scope the literature and find gaps, then follow your field's protocol.
Only you, unless you share a teamspace. Every table is protected by Postgres row-level security, so a request can only return rows that belong to your account.
Yes, from Settings → Billing, at any time. You keep premium access until the end of the period you've paid for.
Free to start, no card required. Your first literature review takes about ten minutes.