Particle.news

Google Opens Google Home to Third‑Party AI Agents via MCP

Approved external agents will be able to read device data to run permitted actions as Google frames Home as an open infrastructure layer for developers.

Overview

  • Google announced Wednesday that Google Home now supports the open‑source Model Context Protocol, letting approved third‑party AI agents access device data and trigger allowed actions in the Home ecosystem.
  • Access will enter early access for Google Home Premium Advanced subscribers in the United States and requires users to create and configure a Google Cloud project to enable the MCP connector.
  • Supported agents such as Claude, Antigravity, OpenClaw and Hermes can monitor devices, analyze event history across cameras and sensors, send voice messages through Nest speakers, and build custom smart‑home dashboards.
  • Google will keep Gemini as the native Home assistant, while imposing safety protections and rate limits on MCP agents and explicitly blocking high‑risk actions such as unlocking doors, but it warns that outside agents may still cause unexpected behavior.
  • The move extends Google’s push to make Home an infrastructure layer for smart‑home developers by adding an open, agent‑friendly control layer on top of earlier API and Gemini integrations, a change that could speed new automations while raising fresh privacy and security tradeoffs for users.