Image Reading (OCR)

Image Reading (OCR)

Security work is full of things that live in pictures — a screenshot of a request, a scanned report, a photo of a whiteboard, a PDF of findings. Image Reading lets you hand any of those to XHack and have the AI pull the text out and act on it immediately.

Read an image in Chat

1. Open Chat in the web app.
2. Attach an image or PDF — click the attach button, paste from your clipboard, or drag the file in.
3. Ask your question. XHack reads the file first, then answers with its contents in context.

Read an image in the Agent

Feed the desktop Agent a screenshot mid-engagement and it uses what it reads without breaking stride — same attach / paste / drag flow.

Accurate, and it respects the target

Reading is tuned for precision — exact text, structure preserved — because a mistyped token changes everything. Extracted text is also anti-injection hardened: content lifted out of an image is treated as data, never as a command, so a hostile screenshot can't hijack the AI.

Your images stay yours

The reading step is light on storage. In the web app the image is read and used, not warehoused; in the Agent, attached images are kept locally on your machine and removed when you delete the chat.

Use your own vision model

Prefer a specific model for reading? Pair Image Reading with Bring Your Own Key and point your own vision-capable model at the task.