Google has begun opening its smart-home ecosystem to third-party AI agents, giving them a route into connected devices, camera histories and household routines through the Model Context Protocol, or MCP. The company’s bet is that an agent can move beyond answering questions to helping manage a home—but, for now, only paying early adopters can test it.

On Tuesday, Google started rolling out early access to its Google Home MCP server. Users can connect compatible agents—including Claude, Hermes, OpenClaw, ChatGPT and Google Antigravity—to Google Home after creating a Google Cloud project, configuring Home MCP and granting the chosen agent permission. The integration covers Nest products as well as Works with Google Home and Matter devices, such as lights and thermostats.¹

The practical appeal is straightforward: natural-language control and a wider view of what happened at home. Google says an agent could review camera summaries, monitor activity, operate devices or build a personalized dashboard. Its product team frames the change as securely connecting agents to “all of the devices and event history in your Google Home ecosystem.”²

That could mean asking what a child did after getting home from school, using cross-camera analysis; checking how many laundry loads were run in a week; or calculating how long lights stayed on. Agents could also announce completed tasks through a Google Home speaker, rather than merely returning text on a screen.²

But the rollout has clear limits. Access is initially reserved for U.S. subscribers to Google Home Premium Advanced, the $20-a-month tier that already includes expanded video history, detailed alerts, video search and daily summaries. Google has not said whether MCP access will reach cheaper plans or other countries, and it is soliciting feedback from early users through its developer community.¹