Google Home Puts AI Agents Behind a $20-a-Month Paywall

MOUNTAIN VIEW, Calif. — Google plans to bring Anthropic’s open-source Model Context Protocol to its smart home ecosystem in an early access build arriving in the coming weeks. The feature will be limited to subscribers of the Google Home Premium Advanced plan, priced at $20 per month or $200 annually.

The protocol will allow Google Home to connect various third-party and internal AI models with smart devices and event logs. Google Antigravity, Anthropic’s Claude, Hermes, and Open Claw are among the supported agents.

MCP functions as a standardized software layer through which AI systems can make external API calls, eliminating the need for customized software bridges for each connected device or service. Within Google Home, agents will be able to parse historic device data and conduct multi-step automation routines through natural language inputs.

Planned uses include creating cross-room security camera summaries, tracking household activity patterns, controlling connected hardware, and assembling dynamic smart home control panels tailored to individual users. The integration places advanced automation capabilities behind Google Home Premium Advanced, part of a broader technology-sector effort to monetize agentic AI tools through high-tier subscription paywalls.

Anthropic introduced MCP in late 2024 as an open standard for unifying the way large language models connect with external data repositories and application software. The protocol has gained rapid adoption among developers, although cybersecurity analysts have repeatedly raised concerns about its lack of built-in, enforced authentication rules.

Early versions of MCP do not mandatorily require strict identity verification or granular access permissions. Implementation flaws have consequently been linked to real-world data security incidents and unauthorized access vulnerabilities across connected platforms over the past year.

Connecting AI agents to home surveillance networks and physical access points brings those concerns directly into the smart home. External models could receive access to historical sensor feeds, camera logs, and device control endpoints, creating potential attack vectors such as prompt injection exploits and data leakage when endpoint permissions are not strictly contained.

MCP has nonetheless gained support as a common interface for agentic workflows. OpenAI, Microsoft, SpaceXAI, and Google have adopted the protocol across desktop, enterprise, and developer environments. Google’s move to incorporate the standard into Google Home will make multi-agent interoperability available in consumer hardware to users who pay the higher subscription cost.