Documentation

AI audits

Weekly and on demand, evidence attached, within your plan's allowance.

Cadence and allowance

An audit runs automatically once a week on every monitored server, and whenever you start one from the server page. Both count against the monthly allowance of your plan, shown on the pricing page.

What it reads and sends

The agent's telemetry and configuration evidence: exposed services, the effective SSH configuration, packages and CVEs, accounts and sudo grants, cron and persistence locations, posture. This is sent to OpenAI in the United States. No file contents are sent except the 200-byte excerpt of a file the web-shell scanner flagged.

A finding

Fields
severity, category, title, what was observed, evidence, recommendation
Proposable actions
block an address, install a package update, apply a hardening fix; each waits for approval
Limits
A finding can be wrong; the evidence is shown so you can check. An audit is not a penetration test and not a compliance certification.

Reviewed against the implementation on 2026-09-20. Something wrong? Tell us.

Trust Center