Nothing leaves
Every question and answer stays on your own hardware. Always.
Private AI · Local LLM
Want AI but can't risk your data leaving?
You don't have to send sensitive information to the cloud to use AI. We set up powerful AI that runs entirely on hardware you own — in your office, server room, or data centre. Nothing leaves your building.
Air-gap capable
$ uclab deploy --air-gapped
✓ Llama 4 Scout · 70B
✓ Air-gap verified
✓ Outbound traffic: 0 B
ready on rack-01.local
Why local?
Because some data is nobody else's business.
When you use ChatGPT or similar tools, your questions and data go to someone else's servers. With a local setup, everything stays with you.
Compliance sorted
GDPR, HIPAA, financial regulations — green out of the box.
Internet optional
Works without internet. No cloud dependency, no downtime.
Sub-second
No waiting for cloud servers. Answers come back instantly.
Regulations? Covered.
If your industry has strict data rules, a local AI setup helps you meet them without extra effort.
6 frameworks · 0 effort
GDPR
Data stays in your jurisdiction
HIPAA
Patient data never hits cloud APIs
Financial
Transaction data stays internal
SOC 2
Audit-ready local infrastructure
ISO 27001
Aligns with InfoSec standards
NIS2
Reduces your attack surface
The AI models
Powerful, free, and they're yours.
These are open-source AI models — completely free, no subscriptions, no per-question fees. We pick the right one for your needs.
Meta · 01Top pick
Llama 4 Scout
All-rounderFree · commercial
Handles massive documents, answers questions, writes content. The go-to choice for most businesses.
DeepSeek · 02
DeepSeek R1
Complex reasoningFree · MIT
Excels at multi-step problems, coding, and analysis. Great for technical teams.
DeepMind · 03
Gemma 4
CompactFree · Apache 2.0
Near top-tier performance in a compact package. Perfect for edge and office hardware.
Moonshot AI · 04
Kimi K2
Tool useFree · MIT
Can use tools, browse the web, and chain actions together. Built for AI that does things.
Alibaba · 05
Qwen 3.5
MultilingualFree · Apache 2.0
Supports 200+ languages and handles huge documents. Fastest in its class.
Mistral AI · 06
Mistral Small 4
SpeedFree · Apache 2.0
Lightweight and fast. Understands images and text. Great for customer-facing apps.
Hardware
What hardware do you need?
It depends on team size and what you want the AI to do. Here are the options — we'll recommend the right one.
Entry01 / 04
~$1,500
Mac Mini M4 Pro
A small, silent box that sits on a desk. Handles everyday AI tasks like answering questions, summarising documents, and internal search.
Ideal for:Small teams · 1–10
Professional02 / 04
~$3,000
Mac Studio M4 Max
More power for bigger models and more users at once. Runs the most capable open-source AI models comfortably.
Ideal for:Medium teams · 10–50
Performance03 / 04
~$2,500 (GPU)
RTX 5090 Workstation
Dedicated GPU for maximum speed. Best when you need fast responses for lots of users at the same time.
Ideal for:High-volume teams
Enterprise04 / 04
Enterprise
H100 / B200 Cluster
Full-scale server infrastructure for the biggest deployments. Handles the most demanding AI workloads.
Ideal for:Large orgs · 50+
Day to day
How your team actually uses it.
A simple chat interface that looks and works like ChatGPT — your team types questions and gets answers. No technical knowledge needed.
01
We install the AI on your hardware
We handle all the technical setup. You don't need to touch anything.
02
Your team gets a simple chat interface
Looks like ChatGPT. Type a question, get an answer. Works in any web browser.
03
Everything stays on your network
No data goes to the internet. Like having your own private ChatGPT only your team can access.
Curious if this would work for you?
Tell us what you're trying to do and we'll give you an honest answer — including what it'd cost. No sales pitch, just a straight conversation.
Book a free call