What gets flagged
- Keys in RPC URL paths (Infura-style /v3/…, Alchemy-style /v2/…, long opaque path segments).
- Credentials in URLs (user:password@host) and secret-looking query parameters (apikey, token, key, signature…).
- Authorization, X-API-Key, Cookie and similar header values, and Bearer tokens.
- JWTs, values of JSON fields named apiKey, secret, password, privateKey, mnemonic and similar.
- 32-byte hex labelled as a private key, PEM private-key blocks, and 12–24-word runs shaped like seed phrases.
- Optional, off by default: emails, IPv4 addresses and long opaque tokens.
Verified output
Each chosen value is replaced everywhere it appears with a stable placeholder such as <REDACTED:provider-key#1>, so the structure of a log stays readable. After replacement the output is searched for every chosen original value; if any remains, copying and downloading are blocked.
If you pasted a real private key or seed phrase
This site never needs one. Treat it as exposed: move funds with a wallet you trust and stop using that key. Redacting a copy does not undo exposure elsewhere.
Questions
Is anything I paste stored?
No. The redaction workspace keeps text only in page memory. Reloading the page clears it.
Why is a transaction hash not flagged?
Hashes, addresses and calldata are public on-chain data. 64-hex strings are flagged only when labelled as a key.
Can I rely on this to catch every secret?
No tool can. It catches common formats and shows you what it found; read the output before you share it.