Google will now let any AI agent run your smart home

Jennifer Pattison Tuohy at The Verge frames the same Google Home MCP launch as Google inviting third-party agents—including Claude and Open Claw—into the home graph. Agents can analyze home data, control devices, and build custom dashboards.

This is Google betting that the winning smart-home layer is the device and history fabric, not a single branded chatbot. For households already deep in Nest and Matter, agent choice becomes a product feature.

It also raises the privacy bar: home data plus control is the combination scammers and buggy agents both love.

We’re watching Google turn Home into an agent-agnostic control plane.

The Verge emphasizes analysis of home data alongside control and dashboards. That’s broader than “turn on the lights.” Summaries of camera activity and custom views mean agents will hold patterns about when you’re home, who’s at the door, and how the house behaves. Useful for caregiving and security. Dangerous if a compromised agent or sloppy OAuth path leaks the diary of a house.

We’re advocates for useful AI here: natural-language home control that isn’t locked to one assistant is overdue. Matter already tried to break silos at the device layer; MCP tries at the agent layer. Stacked together, they look like the first credible path to a truly multi-agent home.

The downside if it scales: consent fatigue, shadow permissions, and a household that can’t answer “which bot can unlock what.” Google’s early-access framing with explicit sign-in helps. The test is what happens at GA.

I’m watching Verge and developer docs for pricing and release-date specifics Tuohy’s framing teases—and for whether Google publishes an agent action audit consumers can actually read.

Google Home MCP early access landed September 16, 2026, overlapping TechCrunch’s coverage. Claude and Open Claw are among the third-party agents called out in The Verge’s report.