Your Mac stays yours

We built Whissle so you get a voice-first AI companion without giving up control. No backdoor, no remote takeover, no silent execution of unvetted code — and we're clear about what we do and don't do.

What we don't do

×We do not have remote control of your Mac. We cannot take over your machine or run actions without you.
×We do not install backdoors or hidden access. Nothing runs unless you trigger it (e.g. by holding fn and speaking).
×We do not execute arbitrary code from the cloud. Scripts and key presses are optional and only run if you explicitly allow "elevated" actions.
×We do not send your raw voice or sensitive data to our servers unless you choose a cloud AI provider — and then only what's needed for that provider.

In-call widget — no bots in the call; you stay in control

What we do

Encrypted storage: Sensitive data (e.g. API keys, preferences) is stored in the macOS Keychain, not in plain text.
Certificate pinning: When we talk to our cloud backend, we use SSL pinning so only the intended server is used (e.g. Cloud Run).
On-demand and automatic deletion: You can delete your data anytime. We also auto-expire certain data after 90 days.
Optional local-only: You can run voice recognition and even the AI entirely on your Mac via Docker — no data need leave your machine. See Technology and Docker.

You initiate every request — hold fn to speak

Permissions we need — and why

Whissle requests only the permissions required for the features you use:

Microphone: So we can hear you when you hold fn and speak. You can revoke this in System Settings at any time.
◇Accessibility: Required for features like typing text into fields, pressing keys, or controlling the UI when you ask (e.g. "click OK"). We only act when you trigger an action.
◇Screen recording (optional): Only if you use features that need to see your screen (e.g. live assist with context). Not required for basic voice commands or writing.