Google opens Home MCP so AI agents can control smart devices

Google is opening a new way for AI agents to interact with smart homes, rolling out early access to a Model Context Protocol server for Google Home that lets compatible assistants control devices, review activity and work with event history. The move matters because it could turn AI agents from chat tools into practical home controllers for millions of Google Home users.

Starting on Tuesday, Google began releasing the Home MCP feature to subscribers on its $20-per-month Google Home Premium Advanced plan in the U.S., with access expanding over the coming weeks. The company is positioning the update as a bridge between agentic AI systems and everyday household automation, while also using the rollout to collect feedback from early testers.

What Google is launching

Google’s new Home MCP server gives AI agents a standardized way to connect to the Google Home ecosystem. In practical terms, that means an agent that supports the Model Context Protocol can request permission to interact with supported smart home hardware and related data, instead of needing a one-off custom integration for each service.

The company says the feature is designed for consumer use cases, especially people experimenting with agents that can handle real-world tasks beyond text generation. That includes asking an assistant to check what a camera saw overnight, summarize home activity, or trigger a connected device with a natural-language request.

How does Home MCP work?

Home MCP works by letting a user connect an AI agent to their Google Home environment through a configured MCP server. Once the agent has the right configuration, the user signs in and authorizes access before the system can act on the home setup.

Google says users will first need to create a Google Cloud project and enable Home MCP. From there, they can hand the configuration details to an agent that supports the protocol, and the assistant will walk them through the sign-in and permissions steps.

Google also plans to provide a setup guide through its Google Home Developer Center, which should make the process easier for developers and power users who want to build custom automations or dashboards.

Which AI agents can use it?

Any AI agent that supports MCP should be able to connect, according to Google. That list includes well-known systems such as Claude and ChatGPT, as well as other MCP-capable agents like Hermes, OpenClaw and Google Antigravity.

The broader significance of that compatibility is that Google is not limiting the Home integration to a single chatbot or a single vendor’s assistant. Instead, it is embracing an open protocol that could allow multiple agent ecosystems to work with the same home devices, so long as they follow the required authorization and connection steps.

Google is betting that a common protocol will make AI agents more useful in the home, letting them handle tasks like checking video history, monitoring activity and controlling devices without a custom app for every brand.

Why Google is doing this now

Google has already been adopting MCP across parts of its broader product stack, including cloud services, data platforms, developer tooling and Google Workspace. Bringing that approach to Google Home extends the company’s strategy into the consumer smart home market, where agentic AI is increasingly being framed as the next major interface layer.

The timing also reflects a larger industry push to move AI assistants from isolated chat windows into systems that can safely take action. Smart homes are a natural testing ground because they combine frequent, everyday tasks with devices that already respond to software triggers, routines and voice commands.

For Google, the appeal is twofold. It gives the company a new way to make Google Home more useful for paid subscribers, and it creates a controlled environment where it can observe how people use agents when those agents are allowed to act on physical devices.

What can users do with Home MCP?

Users can ask AI agents to perform a range of home-management tasks through natural language. Google says the supported experiences include reviewing camera summaries, monitoring home activity, controlling connected devices and building customized smart home dashboards.

Those capabilities point to a future in which an AI assistant can serve as a home operations layer rather than just a voice remote. Instead of opening separate apps for cameras, thermostats or lights, a user may be able to ask a compatible agent for a quick summary or tell it to make a home-wide adjustment.

Examples of possible uses

• Ask for a summary of camera activity from the previous night.
• Check whether doors, lights or thermostats are in the expected state.
• Request a custom dashboard for frequently used home controls.
• Use an agent to monitor event history and surface relevant alerts.
• Control compatible devices using plain-language instructions.

Google says the system works with devices across its ecosystem, including Nest doorbells and thermostats, as well as “Works with Google Home” products and Matter-compatible devices such as smart bulbs.

Who gets access first?

Access is starting with subscribers to Google Home Premium Advanced in the United States. That subscription costs $20 per month and already includes features such as longer event-based video history, descriptive notifications, detailed alerts, video-history search and daily summaries.

Google has not said whether Home MCP will eventually be offered to other subscription tiers or expanded into additional countries. For now, the company is keeping the release relatively narrow, which is typical of an early access launch designed to test reliability and user demand before a wider rollout.

How the rollout will unfold

Google says the release starts now and will continue over the next several weeks. That staged approach suggests the company is watching for technical problems, permission issues and user feedback before deciding whether to expand access more broadly.

The company is also inviting early adopters to comment through its Smart Home for Developers Community. That feedback loop could be important because the real challenge of bringing AI agents into the home is not just connecting the software; it is making sure users trust the agent to act correctly and only within approved limits.

Why early access matters

Early access gives Google room to learn how people actually use agent-powered smart home tools. In theory, the protocol may unlock richer automation. In practice, it will need to prove that it is simple enough to set up, useful enough to keep and safe enough for users to let an AI interact with cameras, thermostats and other connected hardware.

That safety question is especially important because smart homes are inherently personal. A home assistant may know when someone is present, what rooms they use and what the camera sees. Google’s permission-based setup suggests the company understands that trust will be as important as functionality.

How does this fit into Google’s broader AI strategy?

Google Home MCP is part of a wider effort to make AI agents more interoperable across the company’s products. Google has already supported MCP in cloud infrastructure, data systems, developer tools and Workspace applications, which indicates a broader push toward standardized agent connectivity.

For Google, the smart home is a particularly visible consumer extension of that strategy. Unlike a backend cloud integration, a home agent is something users can see, interact with and judge directly in daily life. If it works well, it could deepen the value of Google Home and strengthen the case for paid subscriptions.

It also places Google in the middle of a fast-moving industry conversation about how open agent protocols should function. As more AI systems become capable of taking actions on behalf of users, interoperability standards like MCP could become central to how companies connect assistants to the digital and physical world.

Why the smart home is becoming an AI agent battleground

The smart home has long been a proving ground for voice assistants, routines and automation platforms. AI agents raise the stakes by promising a more conversational and flexible experience, where users describe goals instead of tapping through menus or prebuilt scenes.

That promise is attractive for consumers, but it also raises practical concerns. Agents must understand context, respect permissions and avoid making risky assumptions. A system that can control lights might also need to know when not to change settings, how to handle ambiguity and when to ask follow-up questions.

Google’s move suggests the company believes the market is ready to test those ideas at scale, at least among premium users. It also reflects the company’s effort to stay competitive as other AI platforms look for ways to become more action-oriented and more embedded in everyday life.

What this could mean for Google Home users

For Google Home customers, the biggest change may be convenience. Instead of switching between an app, a voice assistant and separate automation tools, users could increasingly rely on a single agent interface to check, control and summarize their home.

That could be especially useful for people who already use smart cameras, doorbells, lights and thermostats, or who want to create dashboards tailored to their own routines. It may also appeal to developers and advanced users who want to experiment with custom agent workflows built around home data.

Still, this is an early access launch, not a finished consumer product. Availability is limited, Google has not committed to a wider rollout, and the feature depends on both technical setup and the growing ecosystem of MCP-capable agents.

Timeline of the rollout

What happens next?

The immediate next step is limited but important: Google will see how users and third-party agents handle the new connection layer in real homes. If the rollout goes smoothly, the company may eventually broaden access beyond the highest-tier U.S. subscription or extend the feature into more markets.

For now, the launch is a signal that Google is serious about bringing AI agents into the smart home on a standardized footing. If the company can pair convenience with trust, Home MCP could become one of the clearest examples yet of AI moving from conversation to control.