Text that tries to steer an AI model: instruction overrides (“ignore previous instructions”), data exfiltration patterns, and hidden or disguised characters (zero-width characters, soft hyphens, bidi controls, Unicode tag characters, homoglyph words). Each result shows the ruleset version used.

What text should I scan?

Anything an agent or LLM will read that you didn’t write: web pages, retrieved RAG chunks, emails, support tickets, uploaded documents converted to text, or tool output.

Yes. The free preview needs no signup or wallet, is rate-limited per IP and scans up to 2,000 characters per run.

How do I scan longer text or call it from an agent?

The paid InjectScan call ($0.003 per call in USDC over x402) scans up to 100,000 characters and can also scan a URL or MCP tool definitions. Its remote MCP endpoint has a free tier of 30 calls a day.

Does a clean result mean the text is safe?

No. It means no rule matched. Rules catch known patterns; keep treating untrusted text as data, not instructions.

Prompt injection scanner

Paste text an AI model is about to read: a web page, a retrieved RAG chunk, an email or a document. We flag instruction overrides, data exfiltration requests and hidden characters. Free, no signup; up to 2,000 characters per scan.

What is prompt injection?

Prompt injection is text that tries to give an AI model new instructions from inside the data it reads. An agent that browses, retrieves documents or reads email can meet it anywhere, often hidden in an HTML comment, white-on-white text or invisible Unicode characters. Scanning untrusted text before it reaches the model catches the known patterns.

The scan runs on InjectScan, our prompt-injection scanner API, using its free preview. In a pipeline, call the full API ($0.003 per call in USDC over x402) or its remote MCP server; see the prompt injection scanner API guide. InjectScan is built and operated by the youraitools.co team.

Checking an MCP server instead? Use the MCP security scanner for tools/list JSON. More checks in x402 security & verification APIs. Questions? Ask in the community.

FAQ

What does it check?: Text that tries to steer an AI model: instruction overrides (“ignore previous instructions”), data exfiltration patterns, and hidden or disguised characters (zero-width characters, soft hyphens, bidi controls, Unicode tag characters, homoglyph words). Each result shows the ruleset version used.
What text should I scan?: Anything an agent or LLM will read that you didn’t write: web pages, retrieved RAG chunks, emails, support tickets, uploaded documents converted to text, or tool output.
Is it free?: Yes. The free preview needs no signup or wallet, is rate-limited per IP and scans up to 2,000 characters per run.
How do I scan longer text or call it from an agent?: The paid InjectScan call ($0.003 per call in USDC over x402) scans up to 100,000 characters and can also scan a URL or MCP tool definitions. Its remote MCP endpoint has a free tier of 30 calls a day.
Does a clean result mean the text is safe?: No. It means no rule matched. Rules catch known patterns; keep treating untrusted text as data, not instructions.