Google Home adds AI support for Claude and OpenClaw

Google has enabled Google Home to work with third-party AI agents including Anthropic’s Claude and the open-source OpenClaw.

The rollout relies on Model Context Protocol technology, the same framework Google previously used to let AI systems plug into Google Workspace, and lets any MCP-compatible agent reach every device and event history stored inside the Google Home ecosystem.

Anthropic’s Claude, the open-source agents OpenClaw and Hermes, and the Google Antigravity coding platform all appear among the systems Google names as compatible, opening the door for voice-driven thermostat control, custom automations and camera-based security summaries beyond what Gemini for Home currently offers.

Those personalised dashboards mark a notable shift for users, since compatible agents can now build their own home-control interfaces instead of relying solely on the standard Google Home app layout.

Availability and access requirements

The feature began rolling out on Wednesday 16 September, though Google has restricted early access to Google Home Premium Advanced subscribers in the United States.

Google has flagged the release as early access, meaning bugs and missing functionality remain likely, and the company is directing developers toward its smart home feedback channel for reports of issues encountered during testing.

That milestone puts Google Home ahead of Apple Home and Amazon Alexa, making it the first of the three dominant smart home platforms to open its ecosystem to AI agents built outside its own walls.

A head start from third-party platforms

Third-party AI integration is not new to smart homes, since the open-source Home Assistant platform has supported Claude and OpenClaw for several years through an extensive library of community setup guides.

Hardware makers have followed that lead too, with universal hubs from Homey and a dedicated OpenClaw-compatible hub from SwitchBot giving privacy-focused households more routes into third-party AI control than Google Home has offered until now.

That expanded access carries fresh risk, since prompt injection attacks, sometimes called smart home promptware, can hide malicious instructions inside ordinary emails or calendar entries to seize control of connected devices without a user noticing.

Google encountered similar vulnerabilities in early versions of Gemini for Home and patched them quickly, though it has not confirmed a timeline for expanding third-party AI access beyond Premium Advanced subscribers in the United States.