Google will now let any AI agent run your smart home

Google is allowing third-party AI agents to access and control devices in Google Home through a new Model Context Protocol (MCP) integration. The company says agents that support MCP — including examples like Claude, Open Claw, Google’s Antigravity and Hermes — will be able to interact with devices and the event history stored in a user’s Google Home system. According to Google, granting an agent MCP access connects it to real-world events and device state, enabling features such as cross-camera analysis, queries about device histories (for example, how many loads of laundry ran last week or how long lights were left on), and the ability to send audio messages over Google Home speakers when tasks are completed. Agents can also build custom dashboards and otherwise use their own interfaces to control Google Home devices. Google stresses that this integration does not replace its own Gemini for Home assistant, which remains the company’s primary interface for interacting with Google Home through the Home app and Nest speakers after Google Assistant was retired. MCP adds a separate layer that lets third-party agents operate directly with devices and historical data rather than routing everything through Google’s interface. The initial rollout is limited: access will be available to Google Home Premium Advanced subscribers in the U.S. (priced at $20 per month or $200 per year) and will require users to create a Google Cloud project and configure it to use Home MCP. Google also highlights built-in protections such as rate limits and safety checks; for example, it will not permit agents to unlock doors. Google advises developers and users to review its developer policies and terms of service because connecting an external agent can produce unexpected or undesired behavior. Observers say the practical significance of MCP is that it gives agents access to the underlying data and control layer of a home, letting them reason across device histories and automate or troubleshoot in ways traditional interfaces often cannot. Google has been positioning itself as the smart-home infrastructure provider — it opened API access to its smart-home platform in 2024 and has made Gemini for Home a full-stack AI offering — and MCP creates another path for developers to build consumer-facing services on top of Google’s layer. Still, some developers may hesitate given Google’s history of launching and retiring multiple smart-home platforms in past years.