Privacy / Local Execution
SmallClaw Privacy and Local Execution
SmallClaw is designed to keep task structure, permissions, local automation, execution records, and user confirmation inside clear Mac boundaries. Model requests go through the model services configured by the user.
What Local-First Means
- Work structure and execution history are organized around the local app.
- File access, automation, and sensitive actions are constrained by macOS permissions and in-app confirmation.
- Users can connect local model services or supported cloud model services.
Limitations
Whether SmallClaw is fully offline depends on the user's configured models, channels, and external services. When connecting cloud models or third-party channels, the privacy terms and network behavior of those services still apply.
FAQ
Can SmallClaw work with local LLMs?
Yes. If the user connects a locally hosted compatible model service, SmallClaw can send model requests through that local setup.
Does SmallClaw bypass macOS permissions?
No. Local files, automation, and sensitive actions remain subject to macOS and app-level permission boundaries.