Your data, your storage

Look after the machines, not just their backups.

Four things one agent does for Mac, Windows, Linux, FreeBSD, Haiku and pfSense firewalls. Backup — encrypted and deduplicated before it leaves the machine, written to storage you own, and provable: browse inside a restore point and pull back a single file. Remote access — desktop, shell and files, in the same browser tab, off until you switch it on. Monitoring — every disk, every machine, and an alert when one stops reporting in. Quick support — one file for somebody who was never enrolled, and no account for them to make.

One command per machine. No agent to babysit.

What you get

Built around the parts that usually go wrong

Most backup failures are not exotic. They are a job that quietly stopped, a restore nobody tested, and a bucket the ransomware could reach.

Look inside a backup

Walk the folders of any restore point and pull back one file. No restoring a whole tree to find out whether the thing you need is even in there.

Read more

Immutable by default

Object Lock keeps backups undeletable for a retention window you set — including by anyone holding your credentials. Keys are wrapped, and every unwrap is audited.

Read more

Storage you own

Point a policy at your own S3 bucket or keep it on the device. No per-gigabyte markup, no hostage situation, and your data stays in the region you chose.

Read more

An honest activity log

Every backup, restore, failure and sign-in, with the agent's own error message. A job that fails on a remote machine shows up here, not just in that machine's logs.

Read more

Sensible defaults per platform

Starter policies that already know where the files live and which caches to skip — so a new machine protects the right things without you writing exclusion rules.

Read more

Agents that update themselves

Upload a build, promote it, and the fleet moves — checksum-verified and swapped atomically. Rolling back is promoting the previous one.

Read more

Reach the machine itself

A remote desktop, a shell, a file browser and the agent's log following live — in a browser tab, on the machine you were already looking at. The same console that tells you a backup failed lets you go and see why.

Read more

Off until you turn it on

Shell, forwarding, desktop, file transfer and logs are five separate switches, all off by default, at the tenant and at the machine — and the agent enforces its own copy locally, so the console cannot talk its way past a machine that says no.

Read more
Getting started

Protected in about a minute

One command per machine. It installs the agent as a system service, so backups run whether or not anyone is signed in.

Install

Run the one-line installer from your console. It enrols the machine and sets up the background service.

Pick a policy

Start from a platform template, choose how often it runs, and point it at your storage. New machines can inherit it automatically.

Verify

Watch restore points appear, browse inside one, and pull a file back — the only way to know a backup works.

Coverage

Runs where you run

Five operating systems and six processor architectures, built and published together in every release — plus pfSense, which is FreeBSD underneath but arrives as its own firewall package: a pkg, an rc.d service and a settings page in the firewall's own web interface. Supporting an awkward platform is not a slogan here — it is a line in the build, a toolchain somebody maintains, and in one case a machine somebody had to go and find.

macOS Windows Linux FreeBSDexperimental Haikuexperimental pfSenseexperimental
5 operating systems — macOS, Windows, Linux, FreeBSD and Haiku, each with its own service, its own installer and its own screen backend.
6 processor architectures — 64-bit x86 and ARM, 32-bit ARM and x86, RISC-V and POWER.
14 agent builds in every release, published in one go. A platform quietly dropped from that list does not fail — it leaves those machines on the old build until somebody notices.

The second tier is deliberate

A 32-bit Raspberry Pi, an ARM NAS, a decade-old x86 box. Ordinary things to find in a small office, and all three are exactly the machine nobody has a backup of. The 32-bit ARM build is pinned to ARMv6 rather than left to the compiler's newer default, because the reason to publish it at all is the hardware too old for the 64-bit one.

Haiku is not a stunt

It is not a platform the Go compiler ships, so the agent is built with a patched toolchain, five of its dependencies are carried in a workspace beside it, and its tray is a native Deskbar replicant compiled on the machine itself. Backup, restore, enrolment, self-update and remote desktop were each run on real hardware before it was called supported.

And where one cannot, it says so

Haiku takes no filesystem snapshot, because BFS has none. Remote desktop carries sound from Linux and FreeBSD and not yet from macOS or Windows. FreeBSD's remote desktop wants an X session and will not capture Wayland. Each of those is written on the screen it applies to, in the machine's own words, rather than left for you to discover.

What "experimental" means on those two chips.

Exactly what the console says beside them: verified on real hardware, but not yet proven across a fleet. Backup, restore, self-update, remote shell and remote desktop have each been run end to end on a FreeBSD machine and on a Haiku machine — on one of each, for about a day. That is a different claim from the years behind macOS, Windows and Linux, and somebody choosing a platform for a customer's backups should be told which is which.

Remote access

Backup is where you find the problem. This is where you fix it.

Every machine running the agent can be reached from the same console — no second tool, no second agent, nothing for the person at the far end to install, and no inbound port opened on their network. Both ends dial out to a relay on 443 and are spliced together there. Nothing dials in.

Remote desktop, in a browser tab

The agent captures the screen and drives the keyboard and mouse itself, so a machine needs no RDP, no Screen Sharing and no VNC server for this to work. That matters most on the machines that have none: Windows Home, a headless server, a Mac nobody enabled Screen Sharing on. Where a native screen server is running, you can point the session at that instead.

A shell that asks who you are

A real terminal on a real PTY, resized to your window. The tunnel gets you as far as a login prompt and no further: the machine's own username and password are what open the shell, through the same login program somebody sitting at it would meet. Reaching a machine and being somebody on it stay two different things.

Files, and the log as it happens

Browse the machine's filesystem, pull a file back, push one across. Confined to one directory unless somebody deliberately widens it, and there is no delete, no rename and no chmod — a mistyped path should not be able to become an incident. Beside it, the agent's log and its helpers' logs, followed live while you watch.

Sound, where the machine can send it

Macs, Linux and FreeBSD machines send what they are playing, with a speaker and a volume control in the viewer — enough to answer "can you hear that noise?" and "does the video play?". Windows machines do not, yet. Rather than a control that quietly does nothing, the viewer shows a struck-through speaker carrying the machine's own reason for it.

Pick a screen

A machine with more than one monitor offers all of them and you choose which you are looking at, mid-session, without reconnecting. A machine with one offers no control at all — a dropdown with a single entry is a question with one answer.

Port forwarding

Point a local port at a service on the managed machine — a database, an admin page — and use your own tools against it over the same tunnel. That one runs from a small command-line helper the console hands you rather than from the browser, because a web page cannot open a listening socket.

This is the most dangerous thing in the product, and it is built that way.

Shell, forwarding, desktop, file transfer and logs are five independent switches, all off by default, set per tenant and narrowed per machine, with the agent enforcing its own copy rather than trusting what it is told. Permission to use them is a separate grant and never a rank — an owner who has not been given it is refused like anybody else, so "who could have reached that machine?" is a short list of names. A forward reaches only the machine's own loopback unless somebody deliberately widens it, because an agent that will forward anywhere is a way into the network behind it.

And a session that ends does not leave a door open.

Sessions are time-boxed — thirty minutes by default, four hours at the outside — and the control plane cuts one at its expiry rather than when the conversation finishes. On macOS, Windows and Linux a desktop session asks for the machine's own username and password before the first frame is drawn; your Outback Backup password is not it. A desktop session that showed something then locks the console on the way out, so a support call never leaves somebody's desk sitting signed in. Every session is in the same activity log as the backups: who opened it, against which machine, when, and how long it lasted.

Quick support

For the machine that was never enrolled.

Somebody rings who is not on your fleet — a customer, a contractor, a machine you have never seen. They download one file, run it, and read out the number it shows. You type that number in and you are on their screen. Nothing is installed, no account is made, and it stops existing when they close the window.

Enter the support code your technician gave you. It decides whose helpdesk the machine asks for, so the download is stamped with it.

Enter a code to enable the downloads.

Our pricing plans

Per device, per month — never per gigabyte

Storage is yours and billed by your provider, so what you pay us does not grow with your data.

$2 / device / mo

Starter

Small fleets that need the basics done properly.

Backup

  • File and folder backup, versioned
  • Encrypted and deduplicated on the device
  • Your own S3, SFTP or WebDAV storage, or keep it local
  • Browse a backup and restore single files
  • macOS, Windows, Linux, FreeBSD and Haiku

Monitoring

  • Health, disk and metrics for every machine
  • Every volume on a machine, not just the ones backed up
  • Alerting when a machine stops reporting in
  • Backup failures raised as they happen

Volume pricing from the first device.

$3 / device / mo

Standard

Backup, plus hands on the machines when something needs fixing. Everything in Starter, plus:

Backup

  • Everything in Starter

Remote access

  • Remote desktop, shell and file access, off by default
  • 5 simultaneous sessions
  • Unattended access with per-machine capability switches

Quick support

  • Attended support for machines that were never enrolled
  • One file the customer downloads — no install, no account
  • Desktop and file transfer, ending when they close it

Monitoring

  • Managed agent rollout and rollback

Where remote support starts.

$4 / device / mo

Pro

When retention and immutability have to hold up to scrutiny. Everything in Standard, plus:

Backup

  • Grandfather-father-son retention
  • Object Lock immutability against ransomware
  • Pinned restore points retention cannot remove
  • App-aware backup — PostgreSQL, MySQL/MariaDB, SQL Server
  • Virtual machines — Hyper-V, KVM, Proxmox VE, bhyve
  • Point-in-time recovery from WAL and binlogs (in development)

Remote access

  • 25 simultaneous sessions
  • Session recording and an audited history

Quick support

  • Your own branding on the customer's download
  • Every session recorded against the technician

Monitoring

  • Backup traffic relayed for sites with no route to storage

The plan most fleets land on.

Talk to us

Enterprise

Where the security review is the hard part. Everything in Pro, plus:

Backup

  • Multi-tenant isolation with per-tenant storage
  • mTLS device identity, audited key access
  • Bring your own KMS key (in development)

Remote access

  • Unlimited simultaneous sessions
  • SSO and data-residency controls (in development)

Quick support

  • Unlimited attended support sessions

Monitoring

  • Priority support

Quoted on tenancy, residency and support terms.

Simultaneous sessions are counted across remote access and quick support together — both are a technician on a live connection. Prices are per device per month, billed monthly, in Australian dollars and include GST. Enterprise is quoted because it depends on tenancy, residency and support terms. Items marked in development are on the roadmap and not yet shipping — we would rather tell you that here than after you have signed. Storage costs go to whoever hosts your bucket; we never resell it.

FAQ

Questions worth asking

The ones we would ask before trusting anyone with a restore.

Where does my data actually live?

In storage you own — your S3-compatible bucket, or a repository on the device itself. You choose the provider and the region, and the bill for it comes from them, not us.

That also means leaving is not a hostage situation: the backups are already in your account. Stop paying us and the data is still yours, in place.

Can you read my backups?

No. Data is encrypted and deduplicated on the machine before anything leaves it, and the control plane holds metadata and keys — never the file contents.

The repository key itself is wrapped rather than stored in the clear, and every unwrap is written to the audit log. Even browsing a backup from the console is a round trip to the device that holds the keys, because we cannot open it ourselves.

What stops ransomware deleting the backups too?

Object Lock. Backups can be written but not deleted or overwritten until the retention window you set expires — in compliance mode, not even by someone holding your credentials.

You can also pin individual restore points so that no retention schedule can age them out, which is the case that usually bites: the backup you needed was quietly pruned months ago.

How do I know a backup would actually restore?

By restoring from it, which is the only real answer. You can walk the folders inside any restore point and pull back a single file in a few clicks, without restoring the whole tree to find out.

Every job also reports back with the agent's own result — "backed up /Users — 412 files, 8.1 MB", or the error that stopped it — so a failure on a remote machine appears in the console rather than only in that machine's logs.

Which systems does it run on?

macOS, Windows, Linux, FreeBSD, Haiku and pfSense. On 64-bit Intel and ARM everywhere, plus 32-bit x86 for Windows and Linux, and 32-bit ARM, RISC-V and POWER for Linux — fourteen agent builds published together in each release. The agent installs as a system service, so backups run whether or not anyone is signed in.

On macOS the desktop app is signed and notarised by Apple, and the agent can take a consistent APFS snapshot before reading. FreeBSD and Haiku are marked experimental in the console: both work and both were verified on real hardware, but neither has been run across a fleet yet, and we would rather say that than let you find out.

pfSense is newer still, and marked experimental for a plainer reason: it has been installed on one firewall. It ships as a FreeBSD package rather than a bare agent, because a firewall is configuration rather than data — the default policy protects /conf, which is config.xml and with it every rule, VPN, user and certificate. Two things are worth knowing before you rely on it. Restoring config.xml onto a rebuilt firewall does not bring the package back, so reinstalling the agent belongs in your rebuild steps or the backups stop silently. And the firewall is the one machine that has to reach your storage without anything else in front of it.

Can your people get onto our machines?

Not unless you switch it on. Remote shell, port forwarding, remote desktop, file transfer and log following are five separate switches, all off by default, set for your organisation and narrowed per machine — and the agent keeps its own copy of that answer, so a machine that says no cannot be talked round.

Being senior does not grant it: remote access is an explicit per-person grant that an owner does not get by being an owner. Beyond that, a desktop session asks for the machine's own credentials, a shell opens on a login prompt, sessions expire on a clock, and every one of them lands in the activity log with a name against it.

What do we have to install or open up for remote access?

Nothing, and nothing. The remote desktop, the shell, the file browser and the live log all run in a browser tab against the agent that is already there — no viewer to install at either end, no plugin, no VPN, and no inbound firewall rule, because the machine dials out rather than being dialled into.

The one exception is port forwarding, which needs a small command-line helper on your own machine — a web page cannot open a listening socket, and we would rather say so than pretend.

What happens to laptops that are off, or off the network?

Nothing is lost. Work queued for a device waits until it checks in, and the schedule resumes when it comes back. The console shows plainly which machines are online and when each was last seen and last backed up.

How do agents get updated?

You upload a build and promote it; the fleet moves to it on request, verified by checksum and swapped atomically. Rolling back is promoting the previous build. Nothing updates itself behind your back.

Are you SOC 2 certified?

Not yet, and we would rather say so. The architecture was built for it — tenant isolation, audited key access, mTLS device identity, an append-only activity log — and formal certification is on the roadmap rather than done.

If a compliance programme is a condition of buying, we would love to have a chat about it — where it sits, and what timing you need.

Talk to us

Tell us how many machines, and where the data should live.

We will come back with a number. No trial to cancel, no call to book before anyone will quote you.