Pattern detection
Flags instruction overrides, fake system markers, exfiltration and jailbreak phrasing.
Free tool · Beta
Scan text for hidden instructions before your LLM reads it.
Paste any untrusted text and scan it for prompt-injection patterns and hidden characters before you feed it to an LLM.
Free beta · No signup · Runs in your browser
What it is
Scan text for hidden instructions before your LLM reads it.
AntiPrompt is a free scanner that checks untrusted text — web pages, emails, documents, tool output — for prompt-injection patterns and invisible characters before it reaches your model. It runs entirely in your browser.
Features
What it does
Hidden characters
Catches zero-width, bidi and Unicode-tag characters that hide instructions.
Clean copy
One click to copy the text with hidden characters stripped out.
Private
Nothing is uploaded. All scanning happens locally.
How it works
Three steps, no setup
- 1
Paste the text
Any untrusted content headed for your LLM.
- 2
Read the findings
See flagged patterns and hidden characters by severity.
- 3
Clean or reject
Copy a cleaned version or decide not to pass it on.
Try it now · Free beta
Try AntiPrompt
Use it right here. Nothing you enter leaves your browser.
Scan result
Paste text to scan
A heuristic scanner that flags known patterns and hidden characters. It can miss novel attacks and flag harmless text. A clean result is not a guarantee; always treat external content as untrusted. Runs entirely in your browser.
FAQ
Questions, answered
Is it free?
Yes. AntiPrompt is free to use while it’s in beta, with no account required.
Is my data stored anywhere?
No. Everything runs in your browser. Nothing you enter is sent to a server, and it clears when you reload.
Can it catch every injection?
No. It is a heuristic that matches known patterns and hidden characters. It can miss novel attacks and sometimes flags harmless text. A clean result is not a guarantee.