Google Home MCP lets third-party AI agents control smart

Google Home MCP connecting AI agents to smart devices inside a modern house

Google Home now lets any AI agent that supports the Model Context Protocol (MCP) read your smart home data and act on it through Google Home’s device layer. The integration, called Home MCP, is designed to give tools such as Claude and Open Claw a direct line to the lights, locks, thermostats, and event history already tied to a Google Home setup, so they can analyze what is happening in the house and respond on the user’s behalf.

At launch, availability is limited to Google Home Premium Advanced subscribers in the US, at $20 per month or $200 per year, with access rolling out in the coming weeks. Setup requires creating a Google Cloud project and configuring it to use the Home MCP API.

What Home MCP actually does

According to Taylor Lehman, group product manager at Google Home and Nest, Home MCP “allows any AI agents that support MCP, including Google Antigravity, Claude, Hermes or Open Claw, to securely work with all of the devices and event history in your Google Home ecosystem.” The point is to move beyond simple voice commands and let an agent reason across everything Google Home has recorded.

Practical tasks highlighted for the integration include:

Cross-camera analysis, so a user can ask the agent what happened at home after the kids arrived from school.
Pulling device state history to answer questions like how many loads of laundry ran last week, or how long the lights were left on.
Voice replies through a Google Home speaker when an agent finishes a task.
Building a custom dashboard inside Google Home that the agent can use to control devices.

Home MCP is an additional layer on top of Google Home, not a replacement. Gemini for Home remains the front-end assistant inside the Google Home app and on Google Nest smart speakers, and it stays in place after Google Assistant was retired. Third-party agents connected through MCP can send voice messages through those speakers, but Gemini for Home still drives the on-device assistant experience.

Security and safety limits built into the rollout

Handing an AI agent the keys to smart locks, thermostats, HVAC systems, and connected appliances raises immediate questions about privacy and physical safety. Google says Home MCP enforces rate limits and safety protections, including a rule that blocks any agent from unlocking doors.

Lehman also warned that, depending on the agent in use, “connecting it to Home MCP can result in unexpected or even undesired behavior,” and pointed users to Google’s developer policies and terms of service before turning the integration on.

Home Assistant, an open-source smart home platform, already ships a comparable MCP integration that shows what is possible when an agent is wired into a smart home system end to end, including troubleshooting devices, designing dashboards, and writing automations in natural language.

Why Google is pushing this now

The bigger play behind Home MCP is Google positioning itself as the infrastructure layer for the smart home, the same way AWS became the backbone for a generation of internet and software businesses. Google opened API access to its smart home in 2024 and, more recently, turned Gemini for Home into a full-stack AI offering. Home MCP extends that strategy by letting outside companies build consumer-facing products and services on top of Google’s device graph.

Google has also opened a path for developers to use Antigravity, its agentic coding tool, to build custom AI agents for the smart home that tap directly into Google Home device data and control. Even if developers pick a different underlying agent, as long as the home runs on Google Home, Google still captures the infrastructure layer.

The risk for Google is its track record. The company has launched and killed several smart home platforms over the years, including Android @ Home, Weave, Project Brillo, Works with Nest, and Google Assistant. Developers now have to decide whether Google Home is the platform the company is finally committed to keeping around.

What changes for the smart home itself

The headline shift is not that Claude can now flip a light switch through Google Home. It is that a third-party agent now has access to the underlying data and control layer, so it can read device history, reason across it, and act. A system that can troubleshoot a misbehaving device, suggest improvements to automations, and track patterns across the home is more useful than one that just responds with “I’m sorry, I can’t do that.”

That move from command-and-control to something more contextual and proactive is the real bet behind Home MCP, and the reason Google is inviting any MCP-compatible agent in rather than locking the door to its own assistants.

FAQ

What is Google Home MCP?

Home MCP is a new integration in Google Home that lets any third-party AI agent supporting the Model Context Protocol, including Claude, Hermes, Google Antigravity, and Open Claw, read device data and control smart home devices connected to Google Home.

Who can use Google Home MCP right now?

Home MCP is limited to Google Home Premium Advanced subscribers in the US at launch, priced at $20 per month or $200 per year, with broader access rolling out in the coming weeks. Setup requires creating a Google Cloud project and configuring it for Home MCP.

Can a third-party AI agent unlock my doors through Google Home?

No. Google says Home MCP blocks agents from unlocking doors and applies rate limits and other safety protections. Google still warns that, depending on the agent, behavior may be unexpected, so users should review the developer policies and terms of service before enabling it.

Related coverage

This article summarizes reporting from theverge.com. See our editorial disclaimer for how our articles are produced.

Google Home opens its smart home data to third-party AI agents through MCP