Know what’s running.
Know what it covers.

1. Check your computer

Current desktop downloads are for Windows 10/11 with an Intel or AMD x64 processor, and Apple Silicon Macs (M1 or newer). Windows ARM, Intel Macs and phone releases are not included in these downloads.

On Windows, open Settings → System → About and check System type. On Mac, choose Apple menu → About This Mac and check Chip. If you cannot confirm the processor, ask support before buying.

The Windows installer is unsigned. The Mac download is signed and Apple-notarized. If a security warning prevents installation, contact support; keep security protections on.

2. Install and choose what to observe

Choose the matching download, install it and open AI Protection Services (the puppy icon). Choose folders in Settings and explicitly start monitoring. Local observations and rules operate independently of subscription registration.

Check that monitoring reports running, then create a harmless new text file in a selected test folder and look for its activity entry. An entry confirms that change was observed; it does not identify which AI agent caused it or prove a threat was blocked.

3. Connect an agent for permission decisions

This needs a developer integration through the local Node SDK. Start the local permission service, enroll the cooperating agent, keep its token private, and ensure the agent requests and consumes authorization before acting. Adding a process name or installing the desktop app does not connect an agent.

Use a disposable test folder to verify that a denied request is honored by that integration. Applications that do not cooperate can bypass this workflow. The source-distributed Ollama text-file adapter is a developer preview, not a bundled one-click integration.

4. Understand Pro cloud checks

Pro checks up to 10 domains you save roughly hourly, using warning patterns and an operator-maintained indicator list. It also offers optional scam-pattern review of submitted text. Review check history and timestamps in your dashboard.

These checks do not inspect your inbox, calls, other apps or website content. They cannot detect every threat. Your device being offline does not stop server-side domain checks, but you need a connection to submit or view results. iPhone push alerts and AI Agent Defense are not included today.

5. Read status accurately

• App stopped or computer asleep/off: do not assume local activity is being observed.
• Monitoring running: selected-folder observations may be available; this is not device-wide prevention.
• Permission service running: coverage still depends on an enrolled, cooperating integration.
• Internet unavailable: local monitoring can operate while the app is running; account, billing and new cloud requests need connectivity.
• Paid registration active: your account has registration access; this alone is not monitoring or agent-connection evidence.

Need a hand before choosing?

Email teamsafe@aipsafe365.com with your OS version, processor type and intended workflow. Do not send passwords, agent tokens, private files or sensitive prompts. We can help clarify compatibility; an email is not a protection assessment.