From pixels to protection
OCR reads the page; detection locates sensitive spans; rendering applies the replacement and checks the output.
DOCUMENT REDACTION · FOR PEOPLE & AI
Turn scanned pages into safer documents.
OCR connects detected information to its position on the page. See a real scanned agreement before you upload your own.
Blackout or placeholdersWeb, API & MCP
Your next document, ready to share.Try it
Hide detected information behind a solid black mask.
Drop your document here.PDF · up to 20 MiB
Private processing · No third-party LLM services
HTTPS uploads — read data handling details
BUILT FOR THE DOCUMENTS YOU ACTUALLY USEPDF PDFs & scansW Word documentsX Excel spreadsheetsTXT Text & Markdown
THE RESULT, NOT JUST THE PROMISE
Same document.
Fewer private details.
See how sensitive information changes while the rest of the document stays useful.
Original documentagreement-scan.pdf
Processed copy● Blackout mode
Recorded engine outputs · fictional documents · click to enlarge
Download these examples →English test notes →Download original
01 / UPLOAD
Bring your document.
A scan, a contract, a spreadsheet. Choose a file and how you want sensitive information replaced.
02 / PROCESS
Let the busywork happen.
Detection and redaction run automatically. Follow the progress while your document is processed.
03 / DOWNLOAD
Take your copy with you.
Download the processed file. Check it is suitable for your purpose before sharing it.
MADE FOR THIS FORMAT
English and Chinese
Both languages have been tested on real documents. Complex tables, faint print and handwriting can still be missed; equal accuracy across languages is not established.
Speed with context
A one-page English scanned agreement took about 25 seconds in a 29 October 2026 public test. Page count, image complexity and queue length change the wait.
Before you upload
How is the file handled?
Files are processed on our managed servers, without sending document content to third-party LLM APIs. Completed task data is automatically cleaned up after 24 hours; active tasks follow the retention policy.
Can I recover the original information?
Blackout and placeholders do not include a restoration map. Keep your original file separately.
Will every sensitive detail be found?
Automatic detection can miss information. Inspect the output before sharing, especially handwriting, poor scans and unusual layouts.
Security & retention · Help & troubleshooting · Read the test evidence
THE REDACTION LAYER FOR YOUR STACK
Built for developers.
Readable by agents.
The same file-processing capability through REST and MCP. Clear inputs, asynchronous jobs, and an authenticated result.
REST APIMCPMarkdown docs
Start building ↗# Your configured service URL and Bearer key curl -X POST "$REDACTTOOL_BASE_URL/v1/files" \ -H "Authorization: Bearer $REDACTTOOL_API_KEY" \ -F "file=@client-notes.docx" # Then create a redaction job, poll, and download. # Complete examples in the quickstart →
01 / INTEGRATEUpload → job → result ↗A full Python example with bounded polling and authenticated download.02 / CONNECTGive your agent a tool ↗Transport, authentication, tool schemas, and the file upload prerequisite.03 / UNDERSTANDDocs your LLM can read ↗Plain Markdown, an API schema, and an honest capability contract.