Using your own computers (devices)

Your agents normally work on your Lowkey box — the always-on machine that runs your projects. Devices let them reach past that box and act on your actual computers: your laptop, your desktop, a home server. Once you pair a computer, an agent can run commands on it, read and write its files, drive its browser, and even control its screen — all from the same chat.

It’s the difference between “the agent can work in its own workspace” and “the agent can do something on the machine in front of you.”

Pairing a computer

Open Settings → Devices. Choose Generate pairing code, and Lowkey gives you a command to run on the computer you want to connect. Run it there once; the computer then shows up under My devices, marked online or offline depending on whether it’s reachable right now.

You can pair as many computers as you like — each appears in the list with its own name (rename it anytime so “that mini PC” becomes “Home server”).

Choosing which computer

When you have more than one device paired, you tell Lowkey which one to act on by setting an active target — the device a request goes to by default. Pick a device and choose Set active (it has to be online first). You can also just name the computer in your message, and the agent will target that one.

What an agent can do on a device

Once a computer is paired and online, an agent can:

• Run commands on it — on a Mac or Linux machine that’s a normal shell; on Windows it’s PowerShell, and your agents know the difference.
• Read, write, and list files in their real locations on that machine.
• Drive its browser — open pages, click, type, read what’s on screen.
• Control its desktop — click, type, and take screenshots of the actual screen.

So you can say things like “on my laptop, find every screenshot from last week and move them into a Receipts folder,” or “open my home server and check why the backup job failed,” and the work happens on that machine.

Staying in control

Devices hand real power to your agents, so you stay in charge of it:

• Browser and desktop control are off by default. Driving your screen or browser needs you to turn it on, per conversation — an agent can’t reach for it uninvited.
• Pause or unpair anytime. A device you pause goes dark to every agent until you bring it back; unpairing removes it entirely.
• Everything is logged. Each device keeps an audit trail of what ran on it — the commands, their results, and any screenshots — so you can always see exactly what happened.

What to read next

• To understand who’s doing the work → /agents/
• To have a device checked on a schedule → /automations/