Google has launched the preview of its Gemini API Managed Agents, offering developers access to a cloud-based Linux sandbox environment for AI development. Announced at Google I/O 2026, this new service allows developers to create isolated, temporary environments with a single API call, eliminating the need for complex server maintenance and setup. The Managed Agents, powered by the Gemini 3.5 Flash model, enable multi-step reasoning, task planning, and tool usage within a secure sandbox. Developers can define agent roles and skills using versioned Markdown files, simplifying the process of creating custom managed agents. This approach mirrors Anthropic's Claude Managed Agents, as both companies aim to provide containerized cloud platforms for secure agent management. The Gemini API Managed Agents are currently available in preview, with enterprise users having access to the Gemini Enterprise Agent Platform for private preview.