Xiaomi’s Open-Source HachimoDock Turns a $20 Desk Gadget into an Agent Hub

Introduction

A small display, several physical buttons and a joystick are enough to make Xiaomi’s HachimoDock look like a miniature retro game console. Add a West Highland white terrier that changes expression, and the device could easily be mistaken for an electronic pet. Its more interesting role, however, is as a physical interface connecting users with computer agents and smart-home devices. The hardware can reportedly be built for about 150 yuan, making it unusually approachable for an AI project.

Key points

• From voice input to an active companion: Users can issue requests directly to the device. A computer-side agent can then handle tasks such as organizing files, gathering information or writing code, while HachimoDock displays progress through animations, text and sounds.
• Multiple interaction modes: Voice, buttons, a joystick and a touchscreen support both productivity workflows and interactive creations such as games, dashboards and daily utilities.
• Customizable hardware and characters: Users can replace the on-screen pet with personal photos or original characters, and can also design new keycaps, shells and software components.
• Open materials for replication: The project publishes its software source code as well as a bill of materials, structural models and assembly instructions. Community-made components can be shared, reused and modified.
• A route into the smart home: Together with Xiaomi’s open-source MiLoCo project, HachimoDock can act as a personalized controller for connected household devices.

Why it matters

HachimoDock is not important because any single feature is entirely unprecedented. Its significance comes from combining several AI interactions in a low-cost, visible and modifiable object. Agents usually live inside a browser window, terminal or chat interface. Here, a physical character can show whether a task is running, waiting for confirmation or finished, giving the agent a more tangible presence on the desk.

The project also suggests a practical form of “hardware Vibe Coding.” Instead of beginning with a complete embedded-systems curriculum, users can start by making a timer, weather card, stock display or simple game. These small experiments expose the relationships between input, display and device control without requiring a large upfront investment.

There are still practical dependencies. The experience will rely on agent integrations, speech recognition, the quality of community components and the availability of hardware parts. Reports of shortages or higher prices for some components also show that open hardware projects can create supply-chain pressure when they suddenly become popular. Even so, HachimoDock offers a clear direction for AI products: moving beyond software-only assistants toward devices that users can observe, operate and reshape themselves.

Source: QbitAI