Gemini API Managed Agents: 3.6 Flash, hooks, and more
Google Blog 2026-07-28 16:00:00
Context: Google has announced updates to its Gemini API Managed Agents, which enable developers to create autonomous AI workflows. The updates include new features such as environment hooks, model selection, and free tier access, allowing developers to build more sophisticated and cost-controlled AI-powered applications. These updates aim to enhance the capabilities of Managed Agents in handling complex tasks.
Key Facts
- Managed Agents in Gemini API now default to Gemini 3.6 Flash, with no code changes required for existing agents, such as the antigravity-preview-05-2026 agent.
- Environment hooks have been introduced, allowing developers to run custom scripts before or after tool calls made by the agent inside its sandbox, using a hooks.json configuration file.
- The Gemini API supports multiple models, including Gemini 3.5 Flash-Lite for lower-cost applications, and developers can explicitly select models by passing agent_config.model when creating an interaction or managed agent.
- Budget controls have been added, enabling developers to cap total token consumption using the max_total_tokens parameter in agent_config, preventing runaway tasks and allowing for paused and resumed execution.
- Managed Agents are now available on free tier projects, allowing developers to experiment with agentic workflows using an API key from a project without active billing.