Try it free

Nowhere to put a server? We will rent you one.

The appliance. Configured, ready for first login.

Same product. Same install. Same boundaries. We provision the hardware and hand you the keys.

Not a cloud version. Not a shared instance. Not a cut-down tier. One customer, one box.

Twelve steps from a cleared payment to a login screen — not one of them is a person

Support included, first year We are managing every appliance ourselves while there are few enough of us to do it properly. That is included for a year and we will tell you the renewal figure before you reach it.
01 They pay tier and region 03 A box exists in their region 06 It is named and certified 08 Hardened then installed 10 Our key dies and we prove it 12 They log in and set their own Twelve steps. Not one of them is a person.

From cleared payment to a sign-in page, with nobody in between.

What you are renting, and what you are not

We operate the machine

The operating system, the disk, the certificate, the backups, the releases, and the box being up at three in the morning. That is ours, and for the first year it is included.

You operate the platform

Your clients, your tickets, your estate, your decisions, and everything Genie does inside it. That is yours and we cannot see it.

And the line is provable rather than promised The product enforces its own boundaries whether it runs on your rack or on hardware we rent you. We can restart the machine. We cannot read your estate, and nothing about the appliance changes that.

What it costs

The licence does not change. $6 USD per managed asset, every module, AI never metered. The appliance is a separate line, and if you move to your own hardware later the licence price is the same.

Tier
Card
Assets
Per month, USD
Tier 1
NVIDIA RTX 4000 Ada ×1, 20 GB
5,000measured
$650
Tier 2
NVIDIA RTX 4000 Ada ×2, 40 GB
10,000derived
$1,400
Tier 3
NVIDIA RTX 4000 Ada ×4, 80 GB
20,000derived
$3,200
High availability
a second box, warm
any tier
from $1,300

Tier 1 measured at 2,905 inferences an hour · tiers 2 and 3 derived from it · prices in USD · the method

On a 5,000-asset practice $30,000 of licence and $650 of appliance. The hardware is a little over two per cent of what you pay, and it is the two per cent you would otherwise be buying, racking and replacing yourself.
The prices are firm. Tier 1 is measured. Every price above is the hardware we rent plus what it costs us to run it, and it does not move. The asset column carries its basis: Tier 1 is a measured load run — 2,905 inferences an hour on the box we run ourselves — and Tiers 2 and 3 are derived from it, labelled derived in the table rather than in a footnote. So pick the tier with us rather than from the table, and if we size you wrong we move you and you pay the difference from that month, not from the start.

On-premises is still cheaper, and we will say so

If you have somewhere to put a server and somebody to look after it, buy the hardware and run it yourself. It is a one-time cost against a monthly one, and over three years it wins.

The appliance is for the practice that has neither — no rack, no spare hands, and no appetite for a GPU purchase before they have decided the software is worth it.

Cancelling

The appliance is billed monthly and charged for the days it ran. Cancel and we take a backup, hold it for a stated period, destroy the instance, and you pay to the day it was destroyed and nothing after.

The onboarding fee is not refundable, because it pays for work that has already happened — the provisioning, the first estate sweep and the handover.

Where the box can be built

Frankfurt answers the data residency question.

Every region below was driven, not looked up — one instance created, booted to running, destroyed, and the destruction confirmed before the next was started.

Europe

Frankfurt

Tier 1 built in 90 seconds

North America

Seattle · Chicago

Tier 1 provisioned

Asia Pacific

Singapore · Mumbai · Osaka

Tier 1 provisioned

Not every tier is available in every one of those. Only Tier 1 was driven, and only Frankfurt and Seattle carry the full range of plans — Mumbai carries four, Osaka three.

Unknown does not sell. A published table is not a promise that a box can be built today. The provider is asked live, per tier per region, at the moment you buy — and if it cannot confirm, the sale stops rather than proceeding on an assumption.

Either way, it is your deployment

Every client you manage sits inside it. The model runs on that box, so AI usage is never metered and your clients’ telemetry never leaves your control. We receive ten fields and a signature, or nothing at all if you never turn it on.

The question nobody asks until it is too late

What happens to your practice when your vendor gets bought?

You have seen it. The roadmap stops. The price goes up at renewal because the new owner models it differently. Support becomes a queue. And every client, every site and every device you manage is sitting in their cloud, which is the reason you cannot simply leave.

The software is on your hardware

One deployment, in your rack or on a box we rent you. Nobody can switch it off, reprice what is already running, or hold your data while you decide.

And leaving is a supported path

Your estate exports in open formats, on demand, without asking us. A product you can leave is the only kind worth trusting with four hundred clients.

We are a small company and we are not going to promise you we will never be acquired. We built it so that it would not matter.