TokenPad
Safety

Secret Scanner for Prompts

Catch the API key in the config you were about to paste into a chat.

Your input

Paste a config file, environment dump, stack trace or log line.

217 characters5 lines0 tokensor drop a file

Secret ScannerExact
0Secrets found
Secrets found0no known formats matched
Patterns checked9API keys, tokens, private keys, URLs
Characters scanned217in your browser only
Token cost of this result
Output tokens0
As input$0.00
× 100K requests$0.00

Everything on this page runs in your browser. Nothing you paste is transmitted, because there is no server here to transmit it to.

Result
 

The most common way secrets leak

Not a breach. An engineer debugging a deployment pastes a config file, an environment dump or a stack trace into a model to ask what is wrong, and the credentials go with it.

It is mundane, it happens constantly, and it is entirely preventable by checking first. That is the whole purpose of this page.

If something has already been sent

Rotate it. Immediately, and without debating how likely exposure is.

A credential that has left your machine should be treated as compromised: it is in request logs, possibly retained for abuse monitoring, and possibly in your own application logs and error tracker as well. Rotation is cheap. The alternative is not.

What it detects

Structured credential formats with recognisable prefixes or shapes — provider API keys, GitHub and Slack tokens, AWS access key identifiers, private key blocks, JWTs and credentials embedded in connection URLs.

It will not catch a bare random string with no distinguishing structure, because nothing distinguishes it from any other random string. If secrets routinely pass through your prompts, the fix is architectural rather than a regular expression.

Frequently asked questions

Is my text sent anywhere to be scanned?
No. A secret scanner that uploaded your text in order to scan it would be actively harmful. Everything runs as regular expressions in your browser.
Providers say they do not train on API data. Is that enough?
No. Retention for abuse monitoring, your own application logs, your error tracker, and contractual obligations to your customers are all separate concerns from training. Data you never sent is subject to none of them.
What about passwords in prose?
Not detectable by pattern. A password written in a sentence looks like a word. This finds structured credentials, which is where the automatable risk lives.

More safety tools