Luz AIbox: plug-and-play AI agent hardware
A small, silent box that runs your AI agent around the clock — orchestration, memory and tool use on the device, large language models in the cloud.
What is Luz AIbox?
Luz AIbox is a plug-and-play AI agent box. The agent layer runs locally on the device — task planning, multi-agent orchestration, persistent memory and tool use — while inference calls go to the cloud LLM you already trust: OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, Qwen, or any OpenAI-compatible API.
There is nothing to build and no API key to configure before your first conversation. Power it on, connect it to your network, pair a chat channel, and you are talking to your agent in about five minutes.
An AI agent box = local agent runtime + cloud LLM connectivity + plug-and-play hardware, in a small silent box.
A brain on your desk, intelligence in the cloud
Local Agent Runtime
Task orchestration, multi-agent collaboration, memory and tool use all run on the device — lower latency and full control over your data.
Cloud Intelligence
Connect to GPT, Claude, Gemini, DeepSeek, Qwen — or any OpenAI-compatible API. Switch models per task without touching the hardware.
Plug and Play
Power on, connect to your network, pair a channel. No server to build, no Docker to babysit.
Everything a serious agent needs
Multi-Agent Orchestration
Run several agents side by side — each with its own role, tools and memory — and let them hand work to one another.
Tool Use
Built-in tool use and automation connect your agent to your apps, APIs and data sources.
Persistent Memory
Your agent remembers preferences, context and past conversations across sessions, stored on the device rather than a third-party server.
Privacy First
API keys and local memory stay yours. No account is required to run, and data does not leave the box without your permission.
Multi-Platform Messaging
Reach your agent from WeChat, WeCom, WhatsApp, Telegram and other popular messaging platforms.
Always-On, Low Power
Fanless and silent. Typical consumption is under 10 W — roughly $8–12 a year in electricity, less than a standard LED bulb.
Built for home and business
Home
A personal assistant that manages your calendar, controls smart home devices, researches topics and remembers your preferences — without sending your personal life to a third-party server.
Business
24/7 customer service agents, office automation, edge AI gateways and R&D prototyping. Handles customer data without routing it through someone else's cloud.
Three photos in. A video out.
A customer sent us three phone photos: a gold warrior figurine, a hand-painted horse, and the city wall they stand on.
- Upload — three photos, straight from the phone.
- Ask — one sentence: “make a video of them walking the wall.”
- Get it back — a 10-second clip, generated on the spot.
The clip was generated by the agent on the box calling a cloud video model.
Serious hardware, silent operation
Compact, fanless, built for 24/7 operation.
| Specification | Detail |
|---|---|
| Processor | Multi-core heterogeneous architecture with AI inference acceleration and millisecond-level local response. |
| Memory & storage | Ample memory and storage for smooth multitasking and instant access to chat history and your knowledge base. |
| Connectivity | Fast Wi-Fi plus Gigabit Ethernet, with automatic reconnection after an outage. |
| Ports | HDMI · USB-C power · USB 3.0 ×2 · 3.5 mm audio. |
| Power | 5V/3A USB-C input. Typical consumption under 10 W — it will run off a laptop power bank. |
| Operating system | AgentOS, a custom Linux-based OS. Ready out of the box, with lifetime free OTA updates. |
What's included
- Luz AIbox Pro hardware unit
- USB-C power adapter
- Quick-start guide
- Starter cloud LLM credits — ready to use, no cloud API key needed
- 1-year full hardware warranty (motherboard, power supply and enclosure)
- Lifetime free OTA updates for system, models and expert mode
Product questions
Does AIbox run large language models locally?
No. AIbox runs the agent layer on the device — task planning, memory and tool use. Large language model inference happens through the cloud API you connect it to, so you always get current model quality.
Which cloud LLMs can I connect?
OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, Qwen, and any OpenAI-compatible API. You can choose the best model for each task and change it at any time.
How long does setup take?
Usually about five minutes: plug in power, connect to Wi-Fi or Ethernet, and pair your chat channel. The device activates and pairs automatically — no API key configuration needed.
Does it work without an internet connection?
The agent runtime and its local memory keep working offline, but intelligent conversation and task processing need cloud LLM access. Without network access, AI inference is not available.
Which chat channels are supported?
WeChat, WeCom, WhatsApp, Telegram and other popular messaging platforms.
Tell us what you need
Leave your details and we will reply within one business day — bulk orders, enterprise deployment or technical support.
Ready to put an agent on your desk?
Luz AIbox Pro ships worldwide with starter credits included and a one-year hardware warranty.
View pricing