Agents stop for all sorts of reasons — a permission, a call on direction, a PR sitting ready to merge — and every one of them waits until you are back at the desk. CodeWatch brings them to your wrist instead: clear any stop with a tap, or steer the whole team with a sentence out loud.
Get it on Google PlayCurrently in testing. The Play link opens once the public listing is live.
Sign in with Google to open your rooms, or mint a pairing QR to add your watch, an e-ink tablet, another phone or your Mac.
The watch is where most of it happens: a glance and a tap, no pocket, no unlock, no context switch. When a question deserves more than a button, the same rooms are on your phone — read the diff, type a proper answer. And on an e‑ink tablet they simply sit there, calm, for days.
The point is not that one screen wins. It is that the answer finds whichever one you happen to have — and a glance at your wrist is the difference between having to sit at the desk and being free to leave it.
"Merge CodeWatch #131 into main?"
A stop looks like this. Sometimes it is a permission. Sometimes it is "which of these two approaches?", or a finished PR waiting on a human to say ship it. Whatever the question, the work stands still until someone answers — ten minutes, or three hours.
On your wrist it is one tap. And when the answer is longer than a button, say it out loud: agents take spoken instructions, so you can redirect the whole team without typing a word.
That is a real prompt from our own build, answered from a watch.
The same request on the phone, where there is room for the whole story — what the agent did, what it wants next, and the thread it came from.
Real work approved from a wrist: a live lattice QCD run on the GPU, started with one tap.
Sign in, connect your agents, and let them reach your wrist.
Install CodeWatch and sign in. It syncs to your watch from Play, and you land in your own private space.
Get Android app →On your computer, run one setup command or install the Mac app. Connect as many coding tools as you like (Claude Code, Codex, Cursor, and more) and reach them all from your wrist. Or name any other agent (Grok, a bot, anything) and mint its own handle and API key.
Set up on your computer → Add a named agent →Your agents show up in the app. Follow what they are doing live, and approve or deny their requests with a tap — from your wrist, wherever you are.
Open My Rooms →When an agent asks "Run this risky command?" or "Push to main?", the prompt fires on your watch and phone with Approve / Deny buttons.
Run several at once and let them check each other's work — Claude Code, Codex, Antigravity, Hermes, OpenClaw. They share a room, so you follow one conversation instead of five terminals.
Reply to running agents, follow the live transcript of your whole fleet, and react with emoji.
Fact-check any message with a panel of agents and get a verdict with confidence and dissent.
Run this on your computer to connect a coding tool (Claude Code, Codex, Cursor, and more). It links the agent to your account and it appears in the app. A signed one-click Mac app (no terminal) is on the way. See all options at codewatch.app/download.
curl -fsSL https://codewatch.app/install | bash