Google Home Opens Smart Home to AI Agents via MCP

Simon Lüthje
Simon Lüthje · 4 min. read

Google is opening up its smart home ecosystem to third-party AI agents. Using the Model Context Protocol (MCP), third-party assistants such as Claude, Hermes, and Open Claw will be able to access connected Google Home devices, evaluate their status, and perform actions. The rollout will begin in the coming weeks—but for now, only in the U.S. and behind a paywall.

What is the Google Home MCP?

The Model Context Protocol is an open standard introduced by Anthropic in 2024 that has since established itself as the de facto interface between AI agents and external services. Simply put, MCP works like a USB-C for AI: a standardized connector that links different systems together—without requiring custom code to be written for each integration.

Google is now using this standard for Google Home. The Google Home MCP server does not replace Google’s in-house Gemini for Home, but rather serves as an additional interface for external AI agents. For example, anyone using Claude, Hermes, or Open Claw as assistants will be able to connect them directly to their Google Home ecosystem in the future.

What exactly can external AI agents do?

The possibilities sound impressive on paper. According to Google, external agents can use the MCP server to:

  • Analyze device data and usage history—for example, how often the washing machine ran or how long the lights were on during the day
  • Perform cross-device camera analyses
  • Build custom dashboards
  • Play voice messages through connected Nest speakers, such as after completing an automated task

However, there are clear security limits: external agents are not allowed to unlock door locks. Google takes a hard line here when it comes to physical access to buildings.

Home Assistant has been able to do this for a while—with more flexibility

The basic idea isn’t new. Anyone who runs smart home setups with open platforms is already familiar with the MCP principle from the Home Assistant ecosystem. There, it’s been possible for quite some time to set up an MCP server that grants Claude, ChatGPT, or other agents access to all connected devices and automations—including the ability to write YAML automations using natural language. The Home Assistant setup is significantly more open for hobbyists and developers than Google’s new solution.

Google is taking a more deliberate, secure approach: MCP access is more restricted, but it’s officially supported and embedded in an existing ecosystem.

How much does Google Home MCP cost—and who can use it?

This is the biggest catch. At launch, the following restrictions apply:

  • Availability: Exclusively in the U.S. —Germany, Austria, and Switzerland are not included for the time being
  • Subscription required: An active Google Home Premium Advanced subscription is required: $20 per month or $200 per year
  • Technical hurdle: The setup is clearly aimed at developers and tech-savvy users—you must create and configure your own Google Cloud project
  • Rollout: The launch is scheduled for the coming weeks

Google Home Premium replaces the former Nest Aware. Advanced is the higher of the two subscription tiers and includes, in addition to MCP access, AI event descriptions, a searchable 60-day video history, and 10 days of continuous 24/7 recording.

What does this have to do with Gemini for Home?

The Google Home MCP is not a replacement for Gemini for Home. The in-house AI assistant remains the primary interface for voice control and everyday smart home commands. The MCP is aimed at advanced users who want to deeply integrate their preferred external AI agent—such as Claude instead of Gemini—into their connected home.

Google continues to expand its Gemini ecosystem in parallel: Just recently, Google brought the Gemini app to the Windows desktop via a keyboard shortcut. The MCP feature complements this strategy by opening up the smart home ecosystem to non-Gemini agents as well.

Conclusion: Interesting for developers, but still a long way off for everyday use

Google’s move to open Google Home to external AI agents via the Model Context Protocol is conceptually exciting. The MCP principle has proven itself as a standard for connecting AI with external services—and the fact that Google is now adopting this standard in the smart home underscores the growing importance of agent-based AI control.

For the general public, however, this feature is still a long way off: it’s not available in Germany, there’s a $20 paywall, and setup requires developer knowledge. Anyone who wants to integrate AI into their smart home right now will find Home Assistant and a corresponding MCP setup to be a much more flexible option. Google’s official solution could change that in the long run—if it becomes more widely available internationally and more beginner-friendly.