Remotely power cycle a frozen PC: the recovery rung below everything else
A machine that has truly frozen — hung kernel, crashed driver, a box a power cut left dark — cannot be woken by any packet and cannot be reached by any software, because nothing on it is running to listen. The fix is the one you would perform standing next to it: cut the power, put it back, let it boot. A certified power switch — a smart plug or an inline relay module — lets you do exactly that from wherever you are, through your Rooster Wake device on the same network.
This guide is the setup, in the order that matters: the BIOS setting that decides whether the machine comes back at all, the network arrangements that keep the switch findable, what the ladder does before it ever cuts power, how to prove the recovery path once on purpose rather than during a real failure, and what the watchdog will — and deliberately will not — do.
First, the BIOS setting: Restore on AC Power Loss
When mains power returns, the motherboard decides what to do — and out of the box, nearly every board decides to stay off. The setting that changes this is called Restore on AC Power Loss on many boards, or AC Power Recovery, After Power Loss, AC BACK, Power On After Power Failure, depending on the vendor — and it usually ships set to Power Off. Find it in the BIOS or UEFI power section and set it before you rely on a power switch for anything.
It has two useful values, and the difference matters. Last State returns the machine to whatever it was doing when power was lost: a machine that was on comes back on, a machine that was off stays off. That is enough for rescuing a hung machine, which was by definition on when you cut it. Power On boots the machine whenever mains appears, whatever it was doing before — which is what you want if the switch should also start the machine cold, from off at the wall.
One side effect to weigh before choosing Power On: any blip in the mains — a breaker reset, a brief outage overnight — boots the machine, wanted or not. If that would annoy you, choose Last State and accept that cold starts stay a job for the wake path rather than the switch.
The ladder: what happens before power is cut
A hard cut is never the first move and never one click. Ask for a power cycle on a machine that has the free agent installed and reachable, and the sequence — enforced on our side, not left to habit — is: ask the agent for a graceful shutdown first; wait for the machine to go down, up to a timeout; only then cut the mains, hold it off for a few seconds, and restore it. A machine that was merely misbehaving gets a clean shutdown; the hard cut is reserved for the machine that cannot answer at all.
And when it truly cannot answer, be clear about what the cut is: a dirty cut. Whatever was unsaved is lost, exactly as if you had pulled the plug by hand — because that is what it is. Modern filesystems take it in their stride and hardware tolerates it far better than the folklore says, but it is a recovery action, not a convenience, and the confirmation step says so to whoever presses the button, every time, on purpose.
Give the switch a fixed spot on the network
The switch is identified by its hardware (MAC) address rather than its IP address, and the drivers re-find it when your router hands out a new lease. Even so: give it a DHCP reservation in your router, the same fixed-address treatment a printer gets. It costs a minute in the router's admin page and removes a whole class of slow moments at exactly the wrong time — and it is a recommendation rather than a requirement: nothing breaks without it, recovery is simply faster with it.
You need a Rooster Wake device on the same network
The switch is not driven from our servers and not driven from your phone: it is driven over local HTTP by a Rooster Wake device on the same network as the switch. The dongle is the guaranteed way — always on is its entire design, so the path to the switch survives everything short of the whole site losing power. The free agent on another machine works too, with the caveat that it only helps while that machine is awake. An agent on the machine you want to cycle is no use here, by definition: when the machine hangs, everything running on it hangs with it — the same physics as the sleeping host it cannot wake.
That local arrangement is the security story, so it deserves saying plainly: the switch needs no cloud account, no vendor app, and no internet access at all. Firewall it off from the internet and everything still works, because the only path to it is your Rooster Wake device on your network. You leave the vendor's cloud switched off during setup; there is no vendor login to phish and no vendor outage in your recovery path.
Prove the recovery once, on purpose
The BIOS setting above has an unkind failure mode: you believe it is set, and you find out otherwise during a real emergency. So the dashboard offers a recovery verification: at a moment you choose — explicitly, behind a consent step that says exactly what is about to happen — it cycles the machine's power once and records whether the machine came back. Only that recorded result lets a machine's page say "hard recovery verified": the claim comes from a stored fact about your actual machine, never from an assumption about your BIOS.
It never runs by itself. It does not run during setup, it is not a background check, and nothing ever cycles your machine to keep a status light green. Pick a quiet moment — not while somebody is mid-keystroke on the machine — run it once, and the verified state stands until the hardware changes.
The watchdog is alert-only: it tells you, it never acts
Every certified model meters power, and that number is a diagnosis: a machine drawing its normal idle wattage whose agent has gone silent looks hung, while a linked machine drawing zero watts is off at the wall — two different problems, and the dashboard uses two different sentences for them. On the paid plans a watchdog can watch for the hung pattern and send one alert — email, or a push if the dashboard is installed as an app — rather than nagging.
What it will never do is act on its own. The watchdog is alert-only: it never cuts power on its own, and there is deliberately no setting to change that — automatic recovery is absent from the product, not hidden behind a toggle, because a machine that reboots itself on a heuristic is a machine that reboots itself during exactly the wrong write. You get the alert; you decide; the ladder runs when you say so.
What to buy, and what it costs to run
The hardware is a certified power switch: a plug-in adapter for a machine at a wall socket, or an inline relay module for a fixed installation. The certified list says what each model is for and what it is rated to switch. The short version: the plug-in adapters are entirely self-serve, and an inline module is an electrician's job — in the UK that is Part P territory, and neither this guide nor any other page here will instruct you to wire mains yourself.
Hard power cycling sits with sleep, restart and shutdown on the paid plans — the power ladder — while waking stays free on every plan. Starting from nothing, the order is: a plan, a Rooster Wake device — dongle, or the free agent on an always-on box — and then the switch.