You Can Now Ask Your AI Assistant to Run Your Google Home
Google opened early access to a new Google Home connection for AI assistants on September 16, 2026. TechCrunch It lets outside AI assistants control smart home devices and check device history.
The connection uses the Model Context Protocol, or MCP, which works like a universal remote any compatible assistant can use. At launch it works with Claude, ChatGPT, Hermes, OpenClaw, and Google Antigravity. They do not need a separate add-on for each assistant. Access is limited. Control is handed to the assistant you choose.
Google says you can use normal language. You can ask for camera summaries, check home activity, control connected devices, and make custom home dashboards. In short, you can ask about what happened before, or tell a device what to do now, all in conversation.
It works with devices that already work with Google Home. That includes Google Nest doorbells and thermostats, and Works with Google Home or Matter devices such as light bulbs. There is no extra approval for MCP. If it is already in Home, it can be controlled.
Setup takes a few steps. You must create a Google Cloud project, set it for Home MCP, give those setup details to your chosen assistant, then sign in and allow access. The Cloud project holds the rules. You give permission when you link accounts.
Early access started September 16, 2026 and will roll out over the coming weeks to Google Home Premium Advanced subscribers in the U.S. That plan costs $20 per month. Google declined to comment on if or when it will come to other plans or countries. Google is asking early users for feedback in its Smart Home for Developers Community.
Separately, Google Home Developers keeps an MCP Servers page at developers.home.google.com. It describes a separate MCP server for developers that helps AI assistants give correct technical answers about Google Home developer programs and technology. A July 29 note says this developer server works with several developer tools, with or without sign-in.
The broader context here is where the thinking happens. Old routines keep it inside one company's app. This change lets any compatible assistant use device information and controls, and put the steps together itself.
In my view, the questions to watch are permissions and stored history. Camera and activity summaries are helpful because they shorten long records, but they raise questions a light switch does not. Which devices, which dates, and which family members an assistant can see will matter more in shared homes than how many devices are covered. The Cloud project and sign-in steps suggest Google wants those limits clear from the start.
Looking at what this enables, custom dashboards may matter as much as control by voice. If an assistant can see what is happening now and what happened before, it can build a simple screen on the spot. Open connections like this tend to lead to automation that lasts longer than closed routines, as long as people trust the setup.
