About DocuScan
Meet the founder
Creator and lead developer
NAME builds privacy-first legal technology with a simple conviction: understanding what you are signing shouldn't require a law degree, or a credit card. After watching friends and freelancers get burned by clauses buried on page nine, he built DocuScan to put a lawyer's first-pass instincts into everyone's hands, entirely free. Every line of DocuScan is open source, the analysis runs on your own device rather than on a server, and every feature ships to every user. That transparency isn't a business model; it's the point.
Under the hood
A real neural network, running entirely on your machine.
Your PDF, DOCX, or TXT is converted to text and split into overlapping windows so no clause slips between the cracks.
A RoBERTa transformer, fine-tuned on thousands of expert-annotated legal contracts, scans each window for 41 clause categories, from non-competes to uncapped liability, via fast ONNX inference on your CPU.
Detected clauses are grouped by severity and rolled into one clear safety score, with every flag linked back to the exact text that triggered it.
DocuScan's intelligence is powered by open research, and we are deeply grateful to its creators. Our bundled model is roberta-base-on-cuad by Mohammed Rakib and collaborators (Nawar, Rakib, Hai and Haq, LREC 2022, MIT license), thank you for making world-class contract understanding freely available. The model was trained on CUAD, the Contract Understanding Atticus Dataset, expertly annotated by The Atticus Project (Hendrycks et al., NeurIPS 2021, CC BY 4.0). The underlying RoBERTa encoder was released by Facebook AI (Liu et al., 2019). DocuScan simply would not exist without their work.
Analyze your first contract in seconds, free and on your own device.