DocCloak
Use AI without leaking client data.
DocCloak is an open-source document anonymizer that strips personally identifiable information (PII) before you share documents with AI services - and restores the original names in the AI's response. Everything runs in your browser. No server, no API calls, no data leaves your machine.
This page requires JavaScript to run. DocCloak performs all PII detection locally in your browser using WebAssembly, so JavaScript and a modern browser are required.
Who it's for
- Lawyers and legal teams - redact client names from contracts before asking AI to review clauses
- Consultants - anonymize company data in reports before generating AI summaries
- Healthcare professionals - strip patient identifiers from notes before using AI for research
- HR departments - remove employee PII from documents before AI-assisted policy drafting
- Anyone who uses AI tools but handles sensitive data they can't afford to leak
How it works
- Paste your document or upload a .doc/.docx/.pdf file or an image (.png, .jpg) - text is extracted locally with OCR
- Redact - DocCloak detects names, emails, phone numbers, addresses, and other PII using a local ML model and regex patterns
- Copy the anonymized text into any AI service (ChatGPT, Claude, Gemini, etc.) - or download the redacted document
- Restore - paste the AI's response back into DocCloak to replace placeholders with the original names
Features
- Runs locally - ML models run in-browser via ONNX Runtime WebAssembly, zero network requests while a text is anonymized (opening a PDF or an image loads the app's own PDF or OCR files; documents are never sent)
- 14 entity types: persons, emails, phones, SSNs, credit cards, dates, currencies, IP addresses, IBANs, addresses, companies, custom labels
- Supports .doc, .docx and .pdf documents with formatting preserved (the redacted PDF keeps its text layer; pages that cannot be edited safely are rasterized)
- OCR for images and screenshots (.png, .jpg) - text is extracted locally with Tesseract WebAssembly, with redacted image export
- Regex pattern detection for 19 regions: the US, China, Japan and 16 European countries
- 8 UI languages: English, Polish, German, French, Spanish, Portuguese, Swedish, Norwegian
- Open source under AGPL-3.0 - every line of code is auditable on GitHub
View source on GitHub