- New platform simplifies no-code AI agent creation and sharing.
- Advanced networking and AI protection secure sensitive data and language.
- Agent Registry fosters collaboration between developers and business users.
Google Cloud has introduced the Gemini Enterprise Agent Platform, a comprehensive solution designed to empower both developers and business users to build, secure, and share AI agents with unprecedented ease and control.
The new platform leverages years of foundational work in Google's robust networking infrastructure, ensuring that agent communications are standardized and secure from the ground up. Ines Envid, Senior Director in Product Management, highlighted the crucial role of the underlying 'one network' in distributing orchestration controls for policies and identity protocols. This deep integration means that capabilities like the Agent Gateway act as a management construct, orchestrating identity without requiring separate infrastructure stacks.
A key focus of the platform is AI protection, especially concerning human language data. The system is designed to intelligently differentiate between benign information and sensitive data like credit card numbers, preventing exfiltration at the application layer. Furthermore, the Model Context Protocol (MCP) offers sophisticated remote management, integrating seamlessly with Google's extensive API management capabilities, which extend beyond Google Cloud to services like Android and Maps. This allows for integrated authentication, authorization, and quota management without the need for users to deploy their own MCP servers.
The platform also addresses the complexity of building effective AI tools and managing agent 'skills.' Envid likened AI agents to 'gifted teenagers' – highly capable but requiring careful governance. The platform provides a catalog of pre-built tools and emphasizes treating skills as critical software artifacts, complete with versioning, vulnerability scanning, and robust management. The Agent Registry then facilitates sharing and collaboration, enabling users to discover, control access to, and even form 'teams of agents' to tackle complex tasks, bridging gaps in human communication and fostering experimentation. The ultimate vision includes increasingly autonomous and proactive agents, capable of diagnosing and remediating issues like cost anomalies or security vulnerabilities before they are even reported.
“You had binaries before, now you have skills. As dangerous or more.”
- Ines Envid, Senior Director in Product Management
