Leak Watch
Leak Watch is autonomous OSINT leak monitoring — continuous discovery of your organisation’s exposed secrets and data across public sources. There is nothing to configure: no API keys, no provider setup, no sources to wire up. Register targets and (optionally) add watchlist items, and Pencheff hunts for you.
How it works
Leak Watch runs two parallel, autonomous hunters. Their results are deduped into a single alerts list.
| Track | What it does |
|---|---|
| GitHub code search (deterministic) | Searches public code for exposed secrets and credentials — e.g. an AWS key committed to a public CI workflow or a .env file. |
| OSINT agent (LLM) | Autonomously crawls public paste sites and the web for exposures of your identifiers. |
Both tracks feed the same alerts table, deduped by location (source +
source_ref) so a leak found twice isn’t reported twice.
What it covers
Leak Watch hunts for two kinds of identifiers:
- Every registered target, automatically. Domains, emails, repo URLs, and cloud account IDs are derived from every target in your workspace and hunted with no extra setup.
- Watchlist. Add anything you want tracked that isn’t already a target —
a domain, an email, an email domain, an IP or IP range, or a keyword (e.g.
acme.com,[email protected],github.com/acme/api).
Alerts & severity
Each alert carries:
- Severity — how serious the exposure is.
- Location (
source_ref) — exactly where the leak was found (e.g. the file/commit in a public repo, or the paste/page URL). - Evidence — the matched secret is redacted and encrypted at rest; you see enough to confirm the finding without the raw secret sitting in plain text.
Triage an alert to triaged, resolved, or false_positive as you work
through it.
Cadence
- The GitHub collector runs hourly against every workspace with watch items.
- The OSINT agent runs daily, and on demand.
- A workspace-wide Scan now button triggers both tracks immediately.
Plans
Leak Watch is available on Pro and Team/Enterprise plans.
Leak Watch vs. cloud scanning
Leak Watch and cloud scanning catch different things:
- Leak Watch finds exposed secrets in public text — a credential committed to a public repo, dropped in a paste, or indexed somewhere on the open web.
- Cloud security scans find misconfigurations in your infrastructure — a public S3 bucket, an over-permissive IAM policy — by inspecting your cloud accounts directly, not public text.
If you suspect a bucket or IAM role is misconfigured rather than a secret being leaked, run a cloud scan against that target instead.
MCP
Leak Watch can also be triggered from the MCP server via the scan_leaks
tool — see MCP — drive from your IDE for how to connect an
IDE agent.