# OpenClaw Nodes Nodes are physical devices ; primarily iOS and Android phones/tablets, optionally macOS in node mode ; that connect to an [[OpenClaw Gateway]] and act as **sensing and rendering endpoints** for the agent. They are the bridge between an agent that lives on a server and the user's actual physical environment. Where channels carry text/voice in chat apps, nodes carry **device-native capabilities**: camera, microphone, GPS, screen capture, push notifications. ## Capabilities Exposed - **Camera capture** ; take a photo on demand and send to the agent (e.g., "what is this thing?") - **Audio** - Voice notes recorded on the device - Real-time transcription back to the agent - **Talk mode** ; always-on speech recognition for hands-free operation - **Voice wake** ; wake-word detection so the device listens without holding a button - **Location** ; share GPS to the agent when allowed - **Media understanding** ; image, video, and document analysis - **Screen recording** ; capture the device's own screen for agent context (e.g., "what's on my screen right now?") ## Pairing & Trust A node pairs with a gateway via a code/URL handshake. Once paired, the node is a trusted endpoint of that gateway. Multiple nodes per user are normal: phone + tablet + macOS as node = three sensing endpoints on the same gateway. ## Why It Matters The node model lets one server-side agent project itself onto every physical context the user is in: - on the phone in a kitchen ; "what can I cook with these ingredients?" via camera - in the car ; voice-only interaction via talk mode - at the desk ; macOS node mode for clipboard/screen access - offline / lock-screen ; widgets and notifications This is what separates OpenClaw's scope from purely-text harnesses like [[Claude Code]] or [[Codex CLI]] ; those handle code in a terminal; nodes handle the *body* of a personal assistant. ## Configuration & Operations ```bash openclaw devices list # see paired nodes openclaw devices add # initiate pairing ``` Most node-side configuration (wake word, ASR provider, transcription target) lives on the device app itself; the gateway exposes the resulting events through its WebSocket network. ## Since OpenClaw 2.0 [[2026-10-03 OpenClaw 2.0 - easier setup, shared sessions, same security defaults|OpenClaw 2.0]] (August 30, 2026) widened what paired devices do: - **Native apps as full clients**: iPhone, iPad and Android get one Chat surface for typing, dictation, voice notes, attachments, realtime Talk and model/session controls; macOS adds Quick Chat; Wear OS brings transcripts, replies and Talk to a paired watch - **Computer Use** on paired Macs and explicitly enabled Windows machines, subject to pairing, policy and OS permissions; view-only sessions reject input, and Linux control is still experimental - **Paired computers as execution targets**: a session's work can run on a paired device (connected with `openclaw connect`) while the Gateway keeps the transcript and credentials - Removing or re-pairing a device retires its old access ## Related Surface Areas - **Live Canvas** ; agent-driven visual workspace with A2UI support, increasingly relevant for nodes - **Web UIs** ; Control UI, Dashboard, TUI, WebChat ; complementary surfaces, but those are *clients*, not *nodes* ## References - Nodes troubleshooting: https://docs.openclaw.ai/nodes/troubleshooting - OpenClaw 2.0 native apps: https://docs.openclaw.ai/releases/2026.8.1/native-apps - OpenClaw 2.0 browser and Computer Use: https://docs.openclaw.ai/releases/2026.8.1/browser-and-computer-use ## Related - [[OpenClaw]] - [[OpenClaw Gateway]] - [[OpenClaw Channels]]