Keryn Cloud
Everything else on this site describes Keryn running on your machine. Keryn Cloud is the exception, and it is a separate, optional service. You use it only by opening Cloud in Keryn Web (keryn.app, route #/cloud) and signing in. Nothing in the CLI, the language server, the desktop app or a scan in the browser tab talks to it.
What it does
Section titled “What it does”Three scans that look outward, at an address you type, from Keryn’s servers:
| Scan | Looks at | Needs a verified domain? |
|---|---|---|
| Trust site | the public privacy, security, legal and AI pages of a website, scored | no |
| AI surface | which AI services a host exposes: model runtimes, UIs, gateways, vector stores | only for the active port checks |
| MCP server | a remote MCP server’s tools, prompts and resources and their risks | no |
Each result opens in the same results view as a local scan, and the account keeps a history you can delete.
What leaves your machine, and when
Section titled “What leaves your machine, and when”Keryn’s local-first rule (nothing uploads by default; a person presses a button that says where data goes) applies here without exception.
- Never sent: your code, your files, your local scan results, your posture answers, your policies. Cloud has no way to receive them.
- Sent when you sign in: the Cloud page sends the client type (
web) and app name (Keryn web app) toapi.keryn.appto start browser pairing. Sign in opens your Keryn account there (email and password, an emailed link, a passkey, GitHub, Google or your company’s single sign-on). The Cloud page polls for approval using the pairing code. After you approve the browser, it keeps a device token for your organization that you or an admin can revoke. - Sent when you press Scan: the address you typed, the kind of scan, and your session token, to
api.keryn.app. The Cloud page says so above the form: your code stays on your machine, and the address is all that goes. - Sent when you add a domain: the domain name.
- Kept for your organization: its members, its domains with their verification token, and for each scan what you asked for, when, a summary and the report, for 365 days unless an admin sets another period (1 to 730 days, Account > Data).
The data and limits page has the retention, the rate limits and how to delete things, and the privacy page lists every call.
What it is made of
Section titled “What it is made of”Scans run in functions on Vercel behind api.keryn.app, using the same Rust and WebAssembly engines the local tools use. Accounts, verified domains, scan history and results are in a Postgres database at Supabase, in a private schema that only the Cloud API reaches, with row-level security per organization. The browser talks only to api.keryn.app, and only from the Cloud page.
What it does not do
Section titled “What it does not do”- It does not scan your code, and it cannot be pointed at a private address. Targets are public host names; IP addresses, private ranges and cloud metadata addresses are refused when the connection is made.
- It does not call a tool of an MCP server. It reads the handshake and the lists.
- It does not run active probes on a host you have not proved you control: see verified domains.
- It is not a compliance service. Findings and scores describe what the scan observed at that moment.
- Run the three scans.
- Verify a domain to allow port checks.
- Data, retention and limits.
- Single sign-on and directory sync for organization admins.