L0 Capability Modelv0.1

The L0 Capability Model is the foundation layer of the Open AIoT specification: it defines how a device describes itself to AI — so that any compliant AI can understand and control your devices without per-device customization.

The problem it solves

Today's devices each ship their own app and their own private protocol. For an AI to control a light bulb, it first has to learn that vendor's interface — device count × vendor count makes integration cost a multiplication. L0 turns it into an addition: a device states what it can do in one standard capability descriptor, and the AI only needs to read that one descriptor.

The three capability elements

• Property — a readable (or writable) piece of device state, such as a light's on/off switch or a thermostat's current temperature.
• Action — an operation the device can perform, such as “turn on” or “set temperature to 24°C”; dangerous actions require human-in-the-loop confirmation per L2.
• Event — a message the device reports proactively, such as “overheat protection triggered”.

Highlights of v0.1

• The basic structure of the capability descriptor: identity, type, declarations of the three elements, plus units and value ranges.
• A first standard capability vocabulary: example capabilities for common device types such as lights, thermostats and plugs.
• Capability discovery: how an AI enumerates the devices under a workbench and their capability descriptors.

Want to see the capability model in a real interaction? Head to the demo page and chat with the virtual devices (light / thermostat / plug).