Security
Privacy is a visible processing boundary
SafeShell is local-first, but does not pretend every configuration is equally private. Users should always know what data is sent, where it goes, who operates the destination, and what retention applies.
Company-approved processing routes
Explicit consent before crossing a new boundary
Human approval for consequential writes
No prompts or company knowledge in the commercial backend
Security
You choose where intelligence runs
SafeShell makes the processing destination visible. Use models on your device, infrastructure controlled by your company, or an external provider selected by you.
On this device
On-device by default where supported
Company-controlled infrastructure
Company-approved processing routes
Your selected provider
Explicit consent before crossing a new boundary
