On September 16, 2026, Google began rolling out early access to the Model Context Protocol server, or MCP, for the Google Home ecosystem. The move allows AI agents compatible with the protocol to perform tasks within the smart home. According to Google, agents can control connected devices, review security camera summaries, and access event and home activity history using natural-language commands.
Supported agents include Claude, Hermes, OpenClaw, ChatGPT, and Google Antigravity, provided that the agent being used supports MCP. The range of devices is not limited to Google Nest products; it also includes Nest doorbells and thermostats, as well as devices compatible with Google Home or the Matter standard, such as smart lights.
What does the new connection enable?
Users can direct requests to their agent instead of interacting separately with each device or control screen. Examples of use cases cited by Google include reviewing camera summaries, monitoring smart-home activity, controlling connected devices, and building customized dashboards for the home.
However, access does not happen with a single click. Users must first create a project in Google Cloud and configure it to use Home MCP, then provide their chosen agent with MCP configuration data. The agent will then ask the user to sign in and grant the necessary permissions. Google said its developer center will also provide a guide for completing the setup.
Current availability and limitations
The rollout begins in the United States and will continue over the coming weeks for Google Home Premium Advanced subscribers, the highest-priced subscription tier, which costs $20 per month. According to the source material, this plan includes benefits such as longer retention of event-related videos, descriptive notifications and detailed alerts, tools for searching video history, and daily summaries.
Google has not clarified whether Home MCP will later become available to subscribers of other plans or in markets outside the United States. The company also requested feedback from early users through the Smart Home for Developers community during the early-access period.
Why does this news matter?
This development moves AI agents from answering questions to performing actions in a physical environment, while unifying control of multiple devices through a single language interface. MCP is important here because it provides a connection layer that different agents can use with the Google Home ecosystem, rather than limiting interaction to a single app or voice assistant.
At the same time, the setup process shows that this type of integration requires granting the agent permissions to access home devices and activity data, including camera-related content. Managing permissions and defining what each agent can read or execute therefore remains a key practical consideration for users. The source does not yet specify broader details about permission levels or the date of general availability, leaving a full assessment of the experience and security limitations until the program expands.