A private AI appliance that lives in your office.
Kubrius Core runs its own AI model on your premises. Your team gets one place to use AI for all their work, confidential or not, and we set it up for them.
One small box. Everything your firm needs to use AI privately.
Kubrius Core runs on an ASUS Ascent GX10, built on NVIDIA's GB10 Grace Blackwell chip with 128 GB of memory. It's about 15 cm square, sits on a shelf or in your comms cupboard, and draws about as much power as a desktop PC. No rack, no data centre, no cooling project.
- A private AI model running on the appliance
- A workspace your team opens in their browser
- Knowledge bases built from your documents
- Assistants and skills set up for your work
- Policy routing and a record of every decision
- An API for connecting your other systems
What happens when someone asks a question.
Everyday work can go to an approved public model. Anything confidential is answered by Kubrius Core in your office and never leaves the building.
Your team asks in one place
Nobody has to decide which AI tool is safe for which job. They open Kubrius Core and ask.
Your policy decides where it goes
You set what counts as confidential, such as particular matters, document collections or kinds of work. Kubrius Core checks every request against it.
Confidential work is answered in your office
The private model on the appliance answers it, drawing on your own documents. Everyday work can go to an approved public model.
Every decision is recorded
What was asked, which rule applied and where it was answered. When a client or an insurer asks how you use AI, you have the answer.
Knowledge that stays with the firm, not the chat window.
Your documents and the work your team does with them build knowledge bases on the appliance. Answers come back with references to the source, so people can check them. Knowledge bases are access controlled, can be split by client or matter, and can be deleted in full when a client asks.
From delivery to useful, without an AI team.
The appliance is the easy part. What makes it useful is assistants built around the work your people actually do. We do that with you.
We learn how your firm works
A session with the people who will use it. We pick the work where AI will save the most time and agree what counts as confidential.
We build your assistants and skills
Assistants for that work, using your templates, precedents and house style, with the right documents behind each one.
We install it and show your team
We install the appliance in your office, connect it to your documents and show your people how to get good answers from it.
We keep it working
Support, updates and new skills as your needs change, included while you're with us.
One appliance for 20 to 200 people.
We tested how many people can wait on confidential answers at once, and published the method and the limits. For most firms of up to 200, one appliance is comfortable. Larger firms are sized individually.
Kubrius Core is ready to install, and we're taking on our first firms now. You'd be working directly with the team that built it. Our two founding partners get it at £395 a month with the first two months free.
Clear answers.
Does anything leave the building?
Everyday work can go to an approved public model. Anything confidential is answered by Kubrius Core in your office and never leaves the building. You decide what counts as confidential, and every decision is recorded.
Is our data used to train anyone else's model?
No. Your documents are stored and searched on the appliance, and the model on it isn't shared with anyone. Confidential work is never sent to a public AI service.
What does the appliance need from us?
A network connection, a power socket and somewhere sensible to put it. We handle installation, set-up and support.
What happens if the hardware fails?
If you rent, replacement hardware is included. If you buy, the appliance is yours and we'll get you running again under its warranty.
How many people can use it?
One appliance comfortably serves a firm of around 200, because most AI use is reading and checking rather than waiting. See how we measured it.
Can it work with the systems we already use?
Kubrius Core has an API, so your other systems can use it. We agree which connections make sense for your firm before installation.
What does it cost?
£9,995 to buy, including the first year of support, then £249 a month. Or £595 a month to rent. All prices exclude VAT. See pricing.
See Kubrius Core running.
Book 15 minutes with Chris. Bring a piece of work your firm can't put into a public AI tool, and we'll show you how Kubrius Core would handle it, on the real appliance.
- 15 minutes with Chris, the founder
- Kubrius Core running live, not slides
- A personal reply. No spam, no drip funnel
Thank you.
Your message has been received. Chris will reply personally to arrange a time.