Private AI on your terms
A ready-to-use AI appliance for your office: chat assistant, document work and automation — fully offline. Plug it into your network and your team gets a private AI within minutes.
Overview
Ready to go
Models and interface are pre-installed and start automatically.
Maximum privacy
Everything you send and receive stays inside your network.
Familiar interface
A browser-based chat your team already knows how to use.
Features
Work with internal documents
Ask questions about your own files and knowledge base (RAG) — without uploading them anywhere.
Your tone of voice
The assistant uses your terminology and style; the interface can carry your brand.
OpenAI-compatible API
Connect existing tools, agents and scripts to a local endpoint.
Team access
Accounts and access control for your whole team.
Modern open models
Current open-weight models, from fast assistants to large reasoning models.
Updates and support
We keep models and software up to date and help your team get the most out of it.
Configurations & pricing
Choose the option that suits you: build it yourself, have us set it up, or get a turnkey solution.
EdgeBox Basic
A compact appliance for a team of up to 20 people.
- Models up to ~120B parameters (MoE)
- Chat interface and API
- Installation and onboarding
- 12 months of support
EdgeBox Pro
A workstation for demanding teams, tuned to your documents and tasks.
- Large models up to ~235B (MoE)
- Search over your documents (RAG)
- Customisation for your tasks
- Priority support for 12 months
Enterprise
Server infrastructure, training on your data and on-site support.
- Multi-GPU servers
- Fine-tuning on your data
- Integration with your systems
- SLA and on-site support
Prices in USD. Shipping and taxes are calculated individually. Questions: mail@cakescats.com
Technical specifications
| Deployment | On your premises, fully offline; no cloud connection required |
|---|---|
| Models | Current open-weight LLMs (e.g. Qwen, Llama, Mistral families), updated by us |
| Interfaces | Web chat, OpenAI-compatible API, document search (RAG) |
| Access | Accounts, roles, HTTPS on your local network |
| Basic / Pro hardware | 128 GB unified-memory appliance / workstation with a 96 GB professional GPU |
| Setup time | About 10 minutes after connecting to your network |
Frequently asked questions
Does it work without the internet?
What do I need to get started?
How does it compare to cloud AI?
Can it be adapted to our company?
Why are prices higher than last year?
The CakesCats family
Every site and product in one place.
See private AI in action
Book a demo — we'll show EdgeBox on your kind of documents.