Google is launching early access to an MCP server for Google Home, opening its home automation ecosystem to external AI agents. According to TechCrunch, tools such as Claude and ChatGPT will be able to access Google Home devices and automations, subject to permissions granted by the user.
The server uses the Model Context Protocol (MCP), designed to connect AI models and agents to external services. A person using Claude or ChatGPT could therefore ask their usual assistant to interact with their Google Home setup, rather than relying solely on Google's and Gemini's tools.
Permissions that need to be framed
This opening is not limited to viewing information: an agent can potentially trigger an action on physical equipment or launch an automation configured by the user. The central question is therefore permissions: which devices can an agent view or control, which automations can it use, and under what circumstances?
Privacy is also at stake. Data on presence, routines or device status can reveal living habits; users will need to understand precisely what they are authorizing and for which agent.
Details still unknown
The rollout and available capabilities may still evolve, as the service is in early access. In the information reported, TechCrunch does not detail the exact list of devices, automations or actions accessible to each agent.
Local availability, particularly in France and Europe, as well as how consent will be presented, also remain to be clarified. Interoperability can simplify the use of home automation, provided that the control granted to agents remains proportionate.
Comments· No comments yet
Be the first to react.