Google allows ChatGPT and others to control smart home devices.

Google announced that any third‑party AI agent can now control and monitor a user’s smart‑home devices through its new Model Context Protocol (MCP). The move, revealed in a developer blog on Tuesday, means tools such as Claude, OpenAI’s ChatGPT, and Anthropic’s Claude 2 can read sensors, adjust thermostats, lock doors, and more, provided users grant permission. The MCP standardizes how devices and agents exchange data, replacing Google’s older, proprietary Home Graph API.

The shift comes as generative‑AI assistants race to dominate the home‑automation market. Last year Amazon’s Alexa and Apple’s HomeKit added “AI‑skills” that let third‑party services run routines, but the integration was limited to voice commands. Google’s new framework gives AI agents deeper access, potentially turning a simple voice command into a full‑blown scenario that watches a room’s temperature, predicts a user’s mood, and adjusts lighting and music. The change follows a wave of regulatory pressure on tech firms to open their ecosystems. In the EU, the Digital Services Act requires platforms to provide APIs that let competitors build on their data. Google’s MCP is a direct response to that mandate and to consumer demand for more flexible, AI‑powered home control.

The protocol works by having the Google Home hub publish a “context bundle” that includes device status, sensor readings, and user preferences. Third‑party agents can subscribe to these bundles, run local or cloud‑based logic, and send back commands to the hub. The system includes built‑in privacy controls: users can see which agents are connected, revoke access, and set limits on data that can be shared. Google has also partnered with the Open Connectivity Foundation to certify devices that support MCP, aiming for a broad rollout across brands. Early beta tests show that an AI agent can reduce a household’s energy use by up to 15 % by learning when occupants are likely to be away, or it can automatically dim lights at sunset and play soft music for bedtime routines.

Next steps involve regulatory reviews and consumer education. Google will release a public dashboard where users can audit agent activity and adjust permissions. The company plans to roll out the feature in the U.S. and the U.K. next month, with a global launch slated for later in the year. The move could reshape the smart‑home landscape, making AI assistants the central hub for daily routines while raising new questions about data security, privacy, and the balance of power between big tech and independent developers.