Google began rolling out early access to its Google Home Model Context Protocol server on Tuesday, extending the AI-agent race from browsers and office software into cameras, thermostats, lights and speakers.¹

The premise is unusually expansive. MCP-compatible agents—including Claude, Hermes, OpenClaw, ChatGPT and Google Antigravity—can be connected to a user’s Google Home ecosystem after the user creates a Google Cloud project, supplies configuration details and explicitly signs in to grant permission.¹ Google says the arrangement allows agents to securely access connected devices and their event history, not simply issue basic commands.

That opens the door to a more conversational smart home. Users could ask an agent to review camera summaries, track household activity, control devices or assemble a tailored dashboard. Google Home and Nest product manager Taylor Lehman said the integration lets compatible agents “securely work with all of the devices and event history in your Google Home ecosystem.”²

Google’s example use cases push beyond switching off lights. An agent could compare footage across cameras to answer what a child did after arriving home from school, examine device-state history to count laundry loads, or determine how long lights were left on.² It could also send an audio message through a Google Home speaker once a task is complete.

But the launch is deliberately narrow. Access is rolling out over coming weeks only to U.S. subscribers of Google Home Premium Advanced, the $20-a-month tier that includes longer video-event history, detailed alerts, search tools and daily summaries.¹ Google has not said whether Home MCP will reach cheaper plans or other markets.

That makes the release both a glimpse of a more capable home assistant and a premium trial of how much household data people are willing to place in an AI agent’s hands. Google is soliciting early-adopter feedback through its Smart Home for Developers Community as it tests that boundary.¹