We built a desk display for your AI agents

One agent was easy to watch. A hundred isn’t.

A year ago, running an AI agent meant one chat window and your full attention. Now a single developer might have OpenClaw shipping a refactor, Claude Code fixing tests, Codex drafting a migration, and a couple of cron agents running overnight, all at the same time. Delegation finally works. The problem is what comes after you delegate: is it done? is it stuck? what did it just cost me? is one of them waiting on my approval to keep going?

The honest answer today is that you go looking. You alt-tab through terminals, you refresh a dashboard, you scan logs. That is fine for one agent. It does not scale to ten, and it actively breaks at fifty, because the cost of checking is higher than the cost of just letting an agent sit blocked for an hour.

What you actually want is ambient, the way a wall clock is ambient. You do not “open” a clock. You glance, you know, you move on.

What the ClawMetry Device does

It is a small screen that sits on your desk and stays on. At a glance it shows you:

• Who is working right now, across every agent runtime you use.
• What it has cost you today, so a runaway loop is obvious before the bill is.
• What is stuck, the agent that errored or has been idle when it should be moving.
• Who is waiting on you, the agent that paused for an approval and cannot continue until you decide.

And when an agent is blocked on you, the device shows the exact request and you tap Approve or Deny right on the glass. The agent unblocks. That is the only thing the device can change; it stays read-only otherwise, and the decision runs through the same approvals path as the ClawMetry web app.

It is not a pet for one assistant

Most of the “AI buddy” gadgets appearing right now only speak to one model. The ClawMetry Device is deliberately runtime-neutral. It shows every agent tool ClawMetry supports: OpenClaw, NVIDIA NemoClaw, Claude Code, Codex, Cursor, Aider, Goose, opencode, Qwen Code, Hermes, PicoClaw and NanoClaw. One integration covers them all, so the device never goes blind the day you add a new tool. See the full list of runtimes.

Private by construction

Observability usually means shipping your activity to someone else’s servers in the clear. Not here. Everything is encrypted on your machine before it leaves. The device holds your key and decrypts on the screen itself; our cloud only ever stores an unreadable blob. The thing on your desk can show you everything precisely because the company behind it can see nothing.

A finished thing, not a science project

This is not a kit and not a breadboard. It arrives cased, configured and ready: set it on your desk, power it over USB-C, and pair it in about two minutes. We do the hard hardware work so you never have to think about it, and improvements ship to the device over the air, so the unit you buy keeps getting better without you touching anything.

Set it up in two minutes

The device shows your ClawMetry account’s data, so you install the free software first, then pair the screen to it:

pip install clawmetry

Then the device shows a code and a QR. Scan it with your phone or type the code at clawmetry.com/activate, approve it as a trusted device, and it is linked. You can revoke it any time from your account. The free plan covers OpenClaw and NVIDIA NemoClaw; the other agent tools are part of the paid plans.

Founders pre-order

The Founders Edition is $99 plus shipping, fully refundable before it ships. It has an onboard microphone and speaker, and voice (“Hey Claw, what’s stuck?”) is coming as a free over-the-air update for Founders units, so you do not buy new hardware to get it. The glance and tap-to-approve work from day one.