Three pieces, one job: making sure the copy exists
An agent on every machine, a console for the IT team, a quiet tray icon for the user. All of it on your own equipment — nothing rented by the month.
The agent
Windows service
Installed on every workstation or server to protect. It evaluates the schedule, scans watched folders, packages, encrypts and uploads. It starts with the machine and restarts itself after a failure.
Automatic startup, controlled restart after an incident
Logging to the Windows event viewer
Centralized configuration; no user action required
Dry-run mode to validate a setup without sending anything
The console
Administration
The IT team’s screen: companies, machines, watched paths, frequencies, run history and restore. This is where “is that machine protected?” gets answered.
Installed on your server, reached from your own network
Fleet overview and last successful backup per machine
The only part the user sees: current backup status, and a way to pause it for a few hours without calling support.
Backup status in a tooltip, continuously refreshed
Pause and resume without administrator rights
Stop and start the service on managed machines
Under the hood
The mechanics that make the difference
Implementation details are what separate a reliable backup from a folder someone copied by hand.
Scheduling true to the calendar
Four frequency types — once, daily, weekly, monthly — evaluated continuously. Only one frequency fires per cycle: two schedules landing at the same time never collide.
Change detection against the server
The agent doesn’t trust a local state that could lie after a reinstall: it compares against the last snapshot recorded server-side. File names are normalized so accents never cause false positives.
Size-capped batch packaging
Changed files are split into compressed packages with a capped source size. A large folder goes out as several successive batches rather than one fragile transfer.
Cumulative manifest
Every batch is sent with the record of everything confirmed so far. If a batch fails, the server cannot believe an un-uploaded file is backed up — the state stays accurate even after a cut.
Fail-closed encryption
Encryption is applied as the package is written. If encryption is requested and no key is available, the agent sends nothing: never a plaintext upload “just to unblock things”.
History that sticks
Every run leaves a trace: files processed, volumes, duration, errors. That history serves diagnostics as much as proof, during an audit or an insurance claim.
Deployment
Installed with one command, updated the same way
Installation is scripted: agent copy, configuration, service creation, tray icon. Re-running the same script with new binaries updates the machine while keeping its configuration.
01
Install the service on your server
The MIRRO service, its database and the vault are placed on a server in your network. That is where every backup will live.
02
Declare the company and its paths
From the console: the company to protect, its agent account, the watched folders and their frequencies.
03
Deploy the agent on the machines
One installation script in an administrator console: the agent is copied, configured, registered as a service and started.
04
Check, then let it run
The first full run goes out and the console confirms the volumes received. Later runs are incremental, with an email report.
Requirements
A Windows server — physical or virtual — for the service and the vault
Disk space on that server, sized to your data volume and retention
Machines to protect running Windows 10/11 or Windows Server, 64-bit
No runtime to install: the agent ships self-contained
HTTPS access from the machines to the MIRRO server, on your network
A local administrator account for the install only
A temporary working folder for packaging
Already using SIIO backup?
MIRRO succeeds the original Java agent and keeps its data model: your companies, paths and history stay in place. Migration happens machine by machine with no loss of history — a new password is simply set for each agent account.
Is your data really backed up?
A 30-minute demo on your own fleet: scheduling, restore, logs. No strings attached.