Google opens Home ecosystem to third party AI agents through MCP

Google is opening Google Home to third party AI agents through a new Model Context Protocol integration, allowing tools including Claude, Hermes and OpenClaw to access connected devices and smart home data.

Google Home MCP gives supported AI agents the ability to monitor devices, analyze event history and act on behalf of users across the Google Home ecosystem.

The integration could allow users to ask an AI agent to analyze footage across multiple cameras, determine how long lights were left on or calculate how many loads of laundry were completed during a given period. Agents can also send voice messages through Google Home speakers and create custom smart home dashboards.

The new system does not replace Gemini for Home, which remains Google’s primary assistant across the Home app and Nest speakers. Instead, MCP creates an additional control layer that allows outside AI agents to interact directly with devices through their own interfaces.

Access will initially be limited to Google Home Premium Advanced subscribers in the US, which costs $20 per month or $200 per year. The feature is expected to roll out over the coming weeks and requires users to create and configure a Google Cloud project.

Google said the integration includes safety protections and rate limits. AI agents will not be allowed to perform certain actions such as unlocking doors, although the company warned that connecting outside agents could still result in unexpected behavior.

The launch builds on Google’s broader push to become an infrastructure layer for smart home developers. Google previously opened API access to its smart home platform and is now giving developers another way to build AI powered products using Google Home device data and controls.