ADDing agent-context annotations for existing API specs (github.com)

A new open standard called ADD (AI Device Description) has been introduced to facilitate seamless communication between IoT devices and AI systems. By enabling any HTTP-capable IoT device to publish a self-describing JSON document at a specified endpoint, these devices can now articulate their functionality, permissions, and operational conditions without requiring custom integration or cloud services. This minimal implementation means developers can make devices AI-ready with just a single HTTP endpoint, significantly lowering barriers to entry for integrating AI into everyday hardware. The significance of ADD lies in its emphasis on safety and ethical considerations within AI interactions with physical devices. Unlike traditional methods that treat safety as an afterthought, ADD embeds safety protocols directly into the device’s self-description and decision logic. This ensures that AI models possess a nuanced understanding of the device's operational context and limitations, including explicit rules for actions and ethical frameworks that govern behavior. By prioritizing risk management, including features like tool fingerprinting and confirmation flows for ambiguous instructions, ADD exemplifies a pioneering approach to responsible AI-hardware integration—setting a new standard for the industry.