Google Launches Managed Agents in Gemini API for Simplified AI Development
Google introduces Managed Agents in the Gemini API, allowing developers to spin up AI agents that reason and execute code in secure cloud sandboxes. The new Antigravity agent removes the need for complex infrastructure management.
Google introduces Managed Agents in the Gemini API, enabling developers to spin up AI agents with a single API call. Powered by the new Antigravity agent built on Gemini 3.5 Flash, this feature allows AI to reason, use tools, and execute code within an isolated, ephemeral Linux environment. Developers access this capability through the Interactions API and Google AI Studio.
This update removes the burden of building complex infrastructure for production-grade AI applications. The Antigravity agent provisions a remote Linux sandbox where the AI executes code, manages files, and browses the web for live data. Developers maintain session continuity across multiple calls, ensuring all files and state remain intact throughout the interaction.
Developers extend the Antigravity agent by creating custom instructions and skills using simple markdown files like AGENTS.md and SKILL.md instead of writing complex orchestration code. Following the earlier launch of Deep Research as a managed agent, Google now opens its internal agent harness and infrastructure so developers can define and register their own versionable custom agents.