Catch secrets at chat time, not commit time.
Secret scanning for AI coding agents. Six hooks, fail-closed, zero cloud, zero telemetry. The transcript is the leak surface.
| Docs | https://tokenhunt.dev/docs |
| Changelog | https://tokenhunt.dev/changelog |
| License | BUSL-1.1 · MIT in 2030 · commercial by agreement |
Engine
token-hunt— the CLI (scan / list / close / watch), six hooks with block-and-rotate behavior, the allowlist, and the bash escape-hatch runtime detector no regex can see
Plugins
pat-scanner— the one-command Claude Code plugin: one hook, 13 credential classes, regex matching. The guard, without the workflow.
Commercial
token-hunt-pro— silent redaction (the agent never sees the raw value)- active verification against GitHub / Stripe / Slack, offline ed25519 licensing. Zero phone-home.
- 48 credential rules across 53 provider families, local-only
- BLOCK hooks stop the tool call before it runs, not after it logs
- the escape-hatch class (
${VAR:+SET}${VAR:-UNSET}) is caught by the runtime detector — regex can't see it - everything runs on your machine — zero telemetry
pip install token-hunt
claude plugin install token-hunt@tokenhunt