OpenClaw’s mobile apps are companion nodes, not standalone assistants, which means they rely on a separate Gateway running on your own computer. This design solves the problem of sending personal data to cloud services, but it introduces a setup hurdle for users who want AI that can use phone hardware like camera, location, or voice without compromising privacy. The main friction points are: installing the Gateway, pairing the phone over WebSocket, and granting explicit approval for privacy‑sensitive actions such as camera snap or screen record.
A practical solution is to follow the minimal walkthrough: first install the Gateway on macOS, Linux, or Windows (WSL2) with a single command, then launch the iOS or Android app and either scan the QR code or enter the host and port manually. Approve the pairing request from the Gateway CLI, after which the phone appears as a node. To enable hardware features, edit the Gateway configuration to allow specific commands—this keep‑list approach ensures that nothing runs unless you explicitly permit it, addressing the common concern of uncontrolled access.
For remote access, pair the Gateway via Tailscale and use a wss:// endpoint, keeping the connection encrypted and avoiding the need to open ports on your router. Remember that camera and screen capture only work when the app is in the foreground, so plan your workflows accordingly. By keeping the AI core on your own machine and treating the phone as a peripheral, you retain full control over keys, configs, and data while still gaining mobile‑specific capabilities.
#AI #Product #OpenSource #Privacy #SelfHosted #MobileAI