How to redact scanned documents and images with an AI agent Leave feedback

A scan is a picture of a document. The words on it are pixels, so pattern matching has nothing to match — redact_text finds nothing to replace on a scanned page, and finding nothing is not the same as nothing sensitive.

Note

The commands and config snippets on this page are for the .NET build of the server — the only platform available today. Installation and client setup: MCP server for .NET. Other platforms will expose the same tools with their own launch command; everything else on this page applies unchanged.

redact_image_area covers a rectangle with a solid colour box, permanently. Coordinates are pixels from the top-left corner of the page.

How big is scan.png?

get_document_info returns per-page width and height, which turns “the bottom third” into real numbers. Then:

Cover the bottom third of scan.png with a black box.

The tool takes no page argument — only the file and the rectangle — so check where the box landed on every page of the output.

For a specific element — a signature, a face, a header — you need its position. Two practical routes: open the page in any viewer and read the coordinates, or render a preview with a tool that can (render_page on the GroupDocs.Viewer MCP server returns page images inline) and let a vision-capable model estimate the box.

An area redaction that is 20 pixels short is a redaction that failed, and no response text will tell you. Look at the page. This is the one redaction case where verification cannot be automated by re-running a pattern — the check is your eyes, or a preview the agent can see.

Many PDFs are part text, part scan: a born-digital contract with a scanned signature page. Run both tools:

Redact the email addresses in contract.docx, then tell me which saved file to use for the next step.

Because redact_image_area cannot be aimed at one page, treat scanned pages as images that get their own area redaction, and check every page of the final output.

On Linux, including the Docker image (a Linux container), redact_image_area and erase_metadata currently fail on PDF files because the PDF engine’s image handling depends on System.Drawing, which .NET supports only on Windows; both work on Word documents, and redact_text works on PDF. Run PDF area redaction and metadata erasure on Windows with dnx.

This server redacts; it does not recognise text in images. That keeps the behaviour predictable — there is no confidence threshold quietly deciding that a blurry name was not a name. If your workflow needs text out of scans, extract it first with a tool built for it, decide what is sensitive, then redact by area.

One redaction per document in evaluation mode means one box. A page needing three covered regions gets one — and looks processed. get_license_status first.

• Redacting scanned documents when there is no text to find — the same scenario on the GroupDocs blog, with worked examples and the trade-offs.