Free tools · No signup
Free Canary Token Generator
Generate a unique, unguessable tripwire token in your browser — no signup, nothing stored server-side. Paste it into a document your AI agent reads. If the token ever appears where it shouldn't, your document was exfiltrated.
Your canary token
Press generate…
Generated locally with crypto.getRandomValues — 256 bits of entropy. This page makes zero network requests with your token.
How the tripwire works
- Generate a token above and paste it into a document your agent processes — a shared doc, a wiki page, a config file. Add a line like:
INTERNAL-API-KEY: CNRY-your-token-here (do not share) - Monitor. The token should never leave your systems. Search your outbound logs, email archives, or third-party destinations for it.
- Detect. If a prompt-injected agent is tricked into exfiltrating the document ("send me everything you can see"), the token goes with it. A leaked canary = proven exfiltration.
- Respond. In Rogue-Agent Defense, canary leaks fire an instant alert and auto-quarantine the agent — the tripwire becomes an active defense.
Why canaries matter for AI agents
Prompt-injection screening watches what goes into an agent. Canaries watch what comes out. Text-screening APIs stop at the words; they don't track whether a document's contents were later emailed to an attacker. A canary token is the proof that closes that gap — and it's an attack class that text screening alone can't see.
Want canaries that fire automatically?
This generator gives you the token. Rogue-Agent Defense Pro ($79 one-time) wires it into your fleet: managed canaries, leak detection on every action, instant alerts, and auto-quarantine — plus the injection shield and per-action policy engine (ALLOW / APPROVE / BLOCK).
Try Rogue-Agent Defense free → Compare vs text-screening APIs
FAQ
What is a canary token?
A canary token is a unique, high-entropy string planted where it doesn't belong to a legitimate workflow. It should never appear anywhere else — so if it ever shows up in network traffic, logs, or a third-party system, you know something exfiltrated the document it came from.
How do I use a canary token to catch AI agent exfiltration?
Generate a token, paste it into a document your AI agent reads, and monitor for it. If the agent is prompt-injected and an attacker's instruction tells it to send the document's contents out, the token leaves with it. Rogue-Agent Defense detects the leaked token, fires an instant alert, and auto-quarantines the agent.
Is my token stored anywhere?
No. The token is generated entirely in your browser with the Web Crypto API and never leaves your device. Keep a copy in your own records so you can search for it later.
Do I need an account or signup?
No — this tool is free with no signup. Automated canary monitoring (leak detection, alerts, auto-quarantine across your agent fleet) is what Rogue-Agent Defense Pro ($79 one-time) adds.
What does a token look like and how strong is it?
Each token is 32 cryptographically random bytes (256 bits of entropy) encoded as 43 URL-safe characters, prefixed with CNRY-. It's effectively unguessable — a leak means exfiltration, not coincidence.