Resilient backups for businesses that can't afford to lose anything.
A small agent backs up whole drives, file by file, into per-device encrypted cloud storage the machine has no permission to delete from. Scheduled, versioned, verified.
Windows servers and laptops · Agent-based · Encrypted on the machine with its own key per device · Deduplicated and compressed before anything is uploaded
The machine cannot delete its own backups.
Ransomware does not just encrypt files. It looks for the backup target first. So the storage credential we issue to an agent can write new data and read that device’s own history — and carries no permission to delete backup data at all. There is no command, no flag and no administrator action on that machine that removes a restore point.
Said precisely, because the distinction matters to anyone who has read a storage spec: this is credential scoping, not object-lock or WORM storage. We do not use S3 Object Lock, and we do not claim to. What we claim is narrower and checkable — the key that lives on your machine has no delete permission on it. The whole argument, on the security page →
Twelve phrases, and none of them is a promise about the future.
Everything below is running today on Windows servers and Windows laptops. There is no macOS agent and no Linux agent, and there is no disk imaging or bare-metal recovery anywhere in the product.
Write and read. Never delete.
The key on the endpoint has two verbs and not a third. An attacker who owns the machine completely still owns a key that cannot remove a single restore point.
Every run is a new point
Nothing overwrites. A run adds a restore point beside the ones before it, so there is always a version from before the encryption event — not a mirror of it.
Verified after it is written
When a run finishes, the snapshot is walked against the repository and every object it references is confirmed present. A check that cannot finish is recorded as unverified, not as a success.
Runs with nobody logged in
A Windows service that starts with the machine. The tray application only shows status and lets someone start a run by hand.
Only what changed leaves
After the first run only new and changed content is uploaded. A server that changed by 2 GB moves roughly 2 GB, not the whole drive again.
Silence is the real failure
Agents check in about once a minute, and every missed schedule slot is detected. The failure mode of most backup setups is not a loud error; it is months of quiet.
The two things most businesses are already doing.
Both hold a copy of something. Neither tells you when it stopped.
| Cloud file sync | An external drive | Comsky Backup | |
|---|---|---|---|
| Get back a file deleted last Tuesday | If version history goes back far enough | If someone plugged the drive in on Tuesday | Any retained restore point. |
| Recover from ransomware | The encrypted file syncs too | Encrypted as well, if it was connected | The machine’s own key cannot erase stored backup data. |
| Get a dead machine’s data onto a working one | Only what was in the sync folder | If you can find the drive, and it still reads | Pick the dead device’s restore point, pick any online machine, restore. |
| Back up a machine nobody logs into | Needs a signed-in user | Needs a person | Runs as a Windows service. |
| Trust that last night’s snapshot is readable | No check | No check | Verified after it is written; short runs and skipped files are flagged. |
| Find out it stopped happening | Per-user, no fleet view | Ask around | Every missed slot is detected and alerted — and it tells a machine that was off from an agent that did not fire. |
| Rebuild a machine from bare metal | No | Only if it was a disk image | No. We back up files, not disks. If you need to rebuild from bare metal, buy something that does that. |
The last row is the honest one, and it is the same answer we give on the phone. Backup does not sell or support disk imaging. A dead-server day looks like this: you install Windows and your applications from your own media, and Comsky Backup puts the data back — or you restore that machine’s data onto a different machine in the meantime and keep working.
Six pages, each answering one question properly.
How it works →
Three moving parts, and you only ever install one. The agent, the encrypted vault, and one console for the fleet.
What we back up →
Whole drives file by file, open files read through a shadow copy, schedules and retention set on a policy.
Restore →
Look inside a restore point before you commit, then land it in one of four places — including a browser download.
The console →
Every machine, every job, every failure on one screen, streaming live as jobs actually run.
Security →
Why the machine cannot erase what it has already stored — and the exact limit of that claim.
Pricing →
Priced on the storage you hold after dedup and compression. Nothing else is gated. No egress charge.
The ones people actually ask.
The evaluation is about ten minutes of your own time. Install the agent on one machine, let it run overnight, and restore something back in the morning. Restoring is the only part worth testing, so test that.
What exactly runs on my machines?
One Windows service and a small tray application. The service does the work and starts with the machine, so backups happen whether or not anyone is logged in. The tray app is only there to show status and let someone start a run by hand. Nothing on the machine accepts a connection from the network — the agent reaches out, so there is no inbound port to open and no firewall hole for an attacker to aim at.
Do you have a macOS or Linux agent?
Not today — the agent ships for Windows only, and we will not put a date on another platform.
Can you read my data?
Data is encrypted on your machine before it is uploaded, and each device’s repository has its own key. What we operate is the orchestration — schedules, job state, device health — not a readable copy of your files. The repository key itself is issued and held by the service, encrypted at rest, so restores can be driven from the console and a forgotten password does not mean lost backups. If you need a key that only you hold, we do not do that today.
Can it rebuild a whole machine, bare metal and all?
No. Comsky Backup does not sell or support disk imaging or bare-metal recovery. It backs up files — including every file on a drive, if that is what you point it at — and it restores files. It will not rebuild a machine’s operating system, its installed software or its licences. If a one-click whole-machine rebuild is what you are shopping for, this is not the product for that job, and we would rather you knew now than in an incident.
What happens when a laptop is asleep or off the network?
The run is missed and the console says so — a sweeper walks every schedule slot using the same arithmetic the dashboard predicts with, so a missed slot raises an alert and can tell a machine that was switched off from an agent that failed to fire. The agent picks the schedule back up when the machine wakes, and it holds the machine awake for the duration of a run it started. A missed slot is not caught up retrospectively: the next scheduled run is the next one.
Does it slow the machine down?
The first run is the expensive one, because everything is new — and it does not have to finish in one sitting. A run cut short by a dropped connection, a machine going to sleep or the agent updating itself is recorded as interrupted rather than failed, and it picks up from its last checkpoint instead of starting the upload again. After the first run only new and changed content is hashed and uploaded, and repeated content inside that machine’s own history is stored once. Schedules exist so the heavy runs land at 02:00 rather than at 14:00.
Where is the data stored?
In S3-compatible object storage — a dedicated bucket per workspace, and one repository per device inside it. Pointing a workspace at your own bucket or an on-premise store is on the roadmap, not shipped.
What if we leave?
Restores and egress are free, so you can pull everything back down before you go. There is no exit fee, and no charge that only appears when you try to read your own data.
Every device name, job and figure shown in the console mockups across these pages is demo data. The real console is at backup.comsky.ai.
Start with Comsky Backup.
Grow into the ecosystem.
One account opens every Comsky product — and one invoice covers them, however many you run.