How to control your OpenClaw from your Apple Watch

If you're using OpenClaw, you should try Apple Watch support. This guide walks through installing WristClaw, pairing it with your OpenClaw instance, and the four things you can do from your wrist — voice, visuals, theatre mute, and dialog. Free.

What you need

• An Apple Watch (Series 6 or later, watchOS 10+) and the iPhone it's paired to (iOS 17+).
• An OpenClaw instance somewhere — your laptop, a VPS, a home server. WristClaw is the client; OpenClaw is the agent.
• A TestFlight invite. Sign up on the homepage — the form returns the public TestFlight link as soon as you submit.

1. Install WristClaw

1. Open the TestFlight link on the iPhone you want to use. Tap Install.
2. WristClaw's watch counterpart installs alongside automatically — give the Apple Watch app a minute to appear on your watch face.
3. Open WristClaw on iPhone. Grant microphone permission when prompted (the watch will ask separately on first voice use).

2. Pair with your OpenClaw

The pair screen offers three paths. Pick whichever fits how your OpenClaw is set up — all three carry the same payload, just different transports.

Option A — Telegram bot (easiest)

1. On iPhone, tap Pair with your OpenClaw.
2. Tap Copy message.
3. Open your OpenClaw Telegram bot conversation, paste, send.
4. The bot installs the WristClaw plugin and registers this session. Wait ~10 seconds; the home screen flips to "OpenClaw connected".

Option B — Terminal on the OpenClaw host

If you're already SSH'd into the box where OpenClaw runs:

mkdir -p ~/.openclaw/plugins
cp ~/WristClaw/openclaw-plugin/catalog.json ~/.openclaw/plugins/catalog.json
openclaw plugins install ~/WristClaw/openclaw-plugin --link
openclaw plugins registry --refresh

openclaw channels add --channel wristclaw --url wss://relay.wristclaw.app/ws --token <sessionID>
openclaw channels login --channel wristclaw

The exact sessionID token is shown on the pair screen. Restart the gateway after pairing so the WristClaw channel connects.

Option C — Raw payload

For custom flows, the pair screen exposes the JSON payload and a wristclaw://pair?… URL form. Hand either to your OpenClaw host however you like; the contract is the same.

3. Talk to your agent from your wrist

The watch app has four tabs. Swipe horizontally.

• Talk. Hold the mic button to record, release to send. The status dot turns amber while OpenClaw thinks and back to green when the reply arrives — text on screen, audio in your ear if you allow it. On Apple Watch Series 9 / Ultra 2 and later, you can also double-pinch (index finger + thumb, twice) to start and stop hands-free.
• Visuals. If your agent sends images (charts, screenshots, scraped photos), they collect here as a thumbnail grid. Tap to view full-screen with Digital Crown zoom; long-press to enter edit mode and delete with the wiggling X.
• Theatre. A mute toggle. When on, audio replies are suppressed and text replies are chunked into 3–5 word "pages" you swipe through on the next tab. Built for trains, meetings, and quiet rooms.
• Dialog. Vertical-paged text dialog with the agent. Read with the crown, type a short reply from the wrist when theatre-mute is on.

4. Extensions (optional but the best part)

Your OpenClaw agent can push extensions — custom shortcut buttons — to your watch. Each one bundles a multi-step check into a single tap. The agent pushes a JSON descriptor; your watch grows a new tab.

Common ones people make: Am I late? (calendar + location + ETA), What happened here? (Wikipedia near current GPS), Daily memo (calendar + tasks). Ask your agent to add one — it learns what you actually use.

5. Local actions (optional)

From iPhone Settings → Watch Actions, you can give your agent permission to:

• Play / pause its own TTS audio reply on the watch.
• Start and stop HealthKit workouts and sleep sessions ("hey, start a 45-min run").
• Fire a local alert timer after a specified delay ("remind me in 20 min").

Each capability is off by default. Apple does not let third-party apps control system media (Apple Music, Spotify) on watchOS, so that one is intentionally limited to WristClaw's own audio reply.

6. Context signals (optional)

In Settings → Context, you can opt-in per signal — location, now-playing, battery, connectivity, heart rate, motion activity, altitude, workout, focus mode. Each signal has three states: off, on demand (sample when you take an action), interval (continuous). Everything is end-to-end encrypted before it leaves your device.

Useful when you want your agent to know "user is walking, near the lake, focus mode on" without you having to say so.

What's different vs. the workarounds

If you've seen guides recommending Telegram on Apple Watch or Siri Shortcuts: those work, but they route your voice through a third party (Telegram, Apple) and don't give you a wrist-native UI. WristClaw is a real watchOS app that talks to your OpenClaw directly through one end-to-end encrypted channel. Side-by-side comparison: Alternatives.

The OpenClaw-side source

Everything the OpenClaw agent needs to speak the WristClaw protocol — the channel plugin, the pairing handler, the catalog, the test vectors — lives in the same repo. The canonical entry point is the skill file:

• Skill file: github.com/salam/WristClaw/blob/main/skill.md — read this first. It describes how OpenClaw recognizes a pairing payload, what the WristClaw channel delivers (text + audio + images), and the invariants the agent should follow when replying to the wrist.
• Channel plugin sources: github.com/salam/WristClaw/tree/main/openclaw-plugin — the Node.js OpenClaw channel that joins the relay as the host peer. openclaw plugins install points at this directory.
• Relay (Go, self-hostable): github.com/salam/WristClaw/tree/main/relay — single binary, can't decrypt your traffic.

If you'd rather install the channel from ClawHub instead of cloning the repo, the skill is published as clawhub install wristclaw once the public release is live there.

Install WristClaw

Stuck on a step? Email hello@wristclaw.app — I read every message.