Claude and ChatGPT Can Now Control Your Google Home Devices — Google's Home MCP Is Live

by Geth Author

On September 16th, Google opened early access to Home MCP, a Model Context Protocol server that lets third-party AI agents monitor, read event history, and control devices across the Google Home ecosystem. The practical effect: if you use Claude, ChatGPT, or another MCP-compatible assistant, you can now ask it to check your Nest camera, adjust your thermostat, or toggle connected lights — and it will actually do it.

What MCP Makes Possible Here

The Model Context Protocol has become a standard interface for letting AI agents connect to external tools and data sources, and it turns out to be a natural fit for smart home control. Google's Home MCP server exposes the full device graph — from Nest doorbells and thermostats to anything that supports Works with Google Home or the Matter standard — as a set of callable tools. An agent can query which devices are on, pull recent camera events, issue control commands, and chain those into broader tasks.

Google named Google Antigravity, Claude, ChatGPT, Hermes, and Open Claw as the AI agents that can connect to it. Notably, this isn't a proprietary integration — any agent that speaks MCP and can be granted permissions can participate.

Availability and Setup

Access is currently gated to Google Home Premium Advanced subscribers in the US, which runs $20 per month. Setup requires standing up a Google Cloud project and explicitly granting MCP permissions — there's no one-click flow yet. The rollout is English-only to start.

The early access label is doing real work here: Google is clearly watching how people use autonomous agents with home control before widening the gate. Control commands over cameras and door locks carry different risk profiles than adjusting a playlist, and the permission model reflects that caution.

Why This Matters Beyond Convenience

Smart home control has been promised by AI assistants for years, but the actual experience has been fragmented — different apps, different voice commands, inconsistent state. What the MCP approach changes is the model: instead of a bespoke integration for each assistant, there's a standardized interface that any MCP-capable agent can consume. Your preferred AI becomes the consistent layer above all your home devices, regardless of brand.

It also sets a precedent. If Home MCP succeeds, expect similar moves from other platform holders. The smart home wars are being refought on AI agent turf, and Google just drew first blood.