What can I help with?
Ask anything, run commands, explore files, or manage your scheduled tasks.
Message Hermes…
Plug.
Play.
Local AI.
*Currently sold out
[1/4]
WHAT IS IT
Plug & Play Local AI
Local AI. Ready to use the day you get it.
The Hardware
Bought once. No subscription.
The Integration
The model, the agent, the interface — tuned to this hardware, with privacy, access control, and updates handled for you.
Your computer
Your phone
Your API
What it's like to use The Box
This is the actual interface you'll use — the same one that ships on your box. Pick one of the tasks and watch it run. These are live session replays (not live sessions, so you can't send new messages)
Chat
What's inside
Hardware
- ▸AMD Ryzen AI Max+ 395 APU (64 GB / 128 GB)
- ▸Radeon 8060S integrated GPU
- ▸M.2 NVMe PCIe 4.0 storage
- ▸5 GbE Ethernet + Wi-Fi 7
- ▸~4.5 L Mini-ITX chassis
Software
- ▸Qwen 3.6 local model (27B / 35B MoE)
- ▸Hermes Agent (Nous Research)
- ▸Hermes WebUI browser interface
- ▸Our install, update & tuning layer
The hard parts — kept honest, handled for you.
THE BOX
Security
We handle it. Every update we ship — model, agent, harness — is vetted before it reaches your box.
Model tuning
KV cache. GGUFs. Quantization. Speculative decoding. Context windows.
^ if you have no interest in learning these terms, that's what we specialize in.
Privacy
Your data is yours. Everything runs on-prem — nothing leaves the box, and we never see it.
Pricing
One all-in price. Free US shipping, tax at checkout, and ~2-week built-to-order delivery.
Baseline
Best for most$3,200
- AMD Ryzen AI Max+ 395
- 64 GB unified memory
- 500 GB SSD storage
- ▸Runs models up to 40B parameters
- ▸2 concurrent agents/users @ 260k context
- Free US shipping
- Tax at checkout
- ~2-week built-to-order delivery
*Currently sold out
Enhanced
$4,500
- AMD Ryzen AI Max+ 395
- 128 GB unified memory
- 1 TB SSD storage
- ▸Future-proof — runs larger models as they're released
- ▸4 concurrent agents/users @ 360k context
- Free US shipping
- Tax at checkout
- ~2-week built-to-order delivery
*Currently sold out
Enterprise
Let's talk
Custom builds, more-than-standard memory, business/multi-box, tailored support.
Questions, answered.
Is it really plug-and-play, or will I have to set it up?
It arrives ready to use. We do the hard part — choosing the hardware, installing and tuning the model, agent, and interface to it, securing it, and keeping it updated — so you never have to learn any of that. Turn it on, connect from your computer or phone, and start using it the day it lands. Turn-key by design.
How is this different from just buying a Mac Mini or Mac Studio?
You could buy comparable hardware yourself — but the box isn't really the hardware, it's everything we do to it. We pick parts with the memory and bandwidth to run capable models well, optimize open-weight models to that hardware, keep them current as better ones release, and secure the whole thing. A Mac Mini leaves all of that to you; the box hands it to you finished.
How is it different from the cloud AI I already use (Claude, ChatGPT)?
It's yours, and it's genuinely unmetered — no token counter, no usage caps, no subscription, and your data never leaves the device. We're honest that a self-contained appliance makes different tradeoffs than a giant frontier cloud model; it's excellent at everyday text work, and you can watch exactly what it does in the live demo above and judge it for yourself.
Is my box safe when it's connected to the internet? Can anyone else access it?
It's your device on your network. Nothing is exposed to the outside by default, and no one — including us — can reach it unless you choose to open access, over a secure private link (Tailscale). You decide who and what it talks to, and you can turn that off at any time. Private by default, connected by choice.
Does my data actually stay on the box?
Yes. Your prompts, files, and conversations live on your device and never pass through us — we never see them, store them, or use them for anything. The AI runs entirely on the box, so it works fully offline whenever you want it to.
What's actually inside it — models, storage, memory?
An AMD Ryzen AI Max+ 395 with 64 GB or 128 GB of unified memory (Baseline vs Enhanced) — enough to run a capable Qwen 3.6 open model right on your desk — plus fast NVMe storage, with the Hermes agent and WebUI on top. See the full list in “What's inside” above.
How do I reach it — remotely, from my phone?
At home it's on your network. To reach it from anywhere, you connect over a secure private link (Tailscale) from your computer or phone — same box, same privacy, wherever you are. It also exposes a standard open API, so you can point your own apps and tools straight at it.
How does it stay updated as new models come out?
We ship vetted model, software, and security updates. Both the initial setup and every update afterward are reviewed before they reach your box, so you keep improving as better open models release — without having to track releases or judge what's safe to run yourself.
Have more questions? Email us. or