What this Mac exposes.
Encryption, firewall state, sharing, listening ports, sleep, and screen lock.
Security for agentic Macs
Your AI has access. See how much.
Templeton Protect scans the Mac your assistants run on—their logs, standing permissions, exposed services, and the code they can reach—then turns it into one clear list of findings.
Read-only No account Runs locally
One scan surface
Three places risk hides.
A key in a chat log, a firewall that is off, and an .env file in a repository are one problem—this Mac—not three separate reports.
What leaked. What they may do.
Saved conversations, MCP servers, hooks, approval policies, and standing permissions.
The mistakes that become breaches.
Committed keys, unignored environment files, private keys, and secrets still alive in git history.
Signal, not noise
A finding you can act on.
Every result names the thing, explains why it matters in plain language, and points to the next action. Secrets are always redacted.
Names the exact issue.
No vague risk score. Protect tells you which setting, file, service, or permission needs attention.
Explains why it matters.
Each finding connects the technical fact to the way an AI assistant actually works on your Mac.
Gives you the next move.
Clear, direct steps—without printing the credential or changing your system behind your back.
The coverage
Broad enough to matter. Local by design.
The scanner combines precise system checks with secret patterns and assistant-aware configuration audits—without uploading your files for somebody else to inspect.
The only network request is a small, optional update check.
Byte patterns, file modes, and direct system answers. Never a model's guess.
Plus JWTs, PEM keys, and database URLs.
Including Claude, Codex, Cursor, and more.
Nothing changes until you explicitly ask.
The trust contract
Security software should explain itself.
Protect treats access to your machine as a responsibility, not a growth metric.
Nothing leaves this Mac.
No analytics, telemetry, or account. The optional update check carries only the version you are running.
It never prints the secret it found.
Every finding, export, and notification redacts the credential first. A report carrying the key has only copied it somewhere new.
Read-only until you ask.
Nothing is changed without a click. Protect removes a leaked key, not the conversation around it.
Every finding is anchored to a fact.
A byte pattern, file mode, git answer, or system state—not a model's guess.
The engine is open source.
The code that finds and explains problems is MIT-licensed and readable on GitHub.
Simple on purpose
Free to scan. Subscribe to keep monitoring.
The complete scanner stays free. Protect+ keeps checking and tells you when something changes.
Protect
Free, and the whole scanner.
$0/forever
No account, no card, no trial to run out.
- All three scans and every rule
- Fixes and PDF, CSV, and Markdown reports
- Command-line tool for CI and pre-commit
- Open source under MIT
Protect+
The part that runs on your behalf.
$5/month
or $50 a year — two months free. Free for 14 days first, no card.
- Lives in the menu bar and scans on a schedule
- Explains what changed since last week
- Flags a key when it enters a conversation log
- Cancel any time; 14-day refund, no questions
Know what your AI can reach.
Run the scan locally. Get a ranked list. Fix what matters before the next agent session does.
Download Templeton Protect