Particle.news

Google Opens Google Home to Third‑Party AI Agents

The early-access rollout gives external agents access to device states and event history, signaling Google's move to sell Google Home as a developer-focused smart‑home platform.

Overview

  • Google announced Home MCP on Wednesday and began early access for Google Home Premium Advanced subscribers in the U.S., letting compatible AI agents connect to Google Home devices and event logs.
  • Home MCP uses the open Model Context Protocol to let agents such as Claude, Hermes, OpenClaw, ChatGPT and Google Antigravity read device state, query event history, and trigger allowed actions across cameras, lights and thermostats.
  • Setup requires creating a Google Cloud project, configuring Home MCP, sharing MCP details with the chosen agent, and granting permissions when the agent signs in, so the release is aimed at developers and advanced users.
  • Google enforces rate limits and hardcoded safety protections that block high-risk actions like unlocking doors, but it warns third-party agent behavior can be unexpected and recommends reviewing developer policies.
  • The move builds on MCP's open-source roots and Google's prior smart‑home APIs and Gemini work, and it could shift the market by making Google Home an infrastructure layer that other companies use to build AI-driven home services.