Local by design
Every Keryn surface runs the scan on the machine it is on: your laptop, your CI runner or your browser tab. The engine is the same everywhere, compiled to WebAssembly.
- No telemetry, no usage statistics, no crash reports, no account.
- No rule fetching. The rules ship inside each build; the engine never downloads rules at scan time.
- No probing. On your machine, Keryn classifies observations you hand it (a page snapshot, a list of open ports); it never scans hosts.
The browser surfaces make calls only when you start them:
- fetching a repository, or the repositories of an organisation, that you named, from its host without cookies, or through api.keryn.app without storage if you chose the fast archive in Settings;
- in Keryn Web, a connector call to the one address you configured, when you press Test, Triage or Send. This is the only way findings leave the page, and it sends what the connectors guide lists.
Two parts are hosted, and both are optional. The fast archive streams a public GitHub repository through api.keryn.app to the tab without storing it, only when you choose it in Settings. Keryn Cloud, used only by signing in on the Cloud page, receives an email address and the address you ask it to scan, never code or local results.
The promises are stated on guarantees, and each connector’s page lists the fields it sends: connectors.
The privacy page is the complete list. A test checks it against the code’s own list of allowed network calls, so the two cannot drift apart.