Google Updates Gemini API Managed Agents with 3.6 Flash Support and Hooks

Google announced significant updates to Managed Agents in the Gemini API on July 28, introducing support for the Gemini 3.6 Flash model and a new Hooks system. The announcement was made by Google engineers Philipp Schmid and Mariano Cocirio.

Managed Agents now support the Gemini 3.6 Flash model, Google's latest high-efficiency offering that balances inference speed with response quality. This makes it particularly suited for latency-sensitive agent scenarios where quick responses are critical.

The new Hooks mechanism allows developers to inject custom logic at key points in an agent's execution flow. This enables finer-grained control over agent behavior, such as data validation, logging, or permission checks before and after external API calls.

Managed Agents is Google's enterprise-grade hosted service in the Gemini API that handles agent lifecycle management including state maintenance, task scheduling, and error recovery. The service already supported multiple models and tool integrations before this update.

Google emphasized that these new capabilities are designed to improve agent reliability and customizability without increasing operational overhead for developers. The Hooks feature is particularly significant, as it allows agents to adapt to complex enterprise workflows rather than being limited to simple Q&A scenarios.

For the developer community, the addition of Gemini 3.6 Flash expands model options. In cost-sensitive and high-concurrency scenarios, developers can deploy 3.6 Flash within Managed Agents to optimize inference costs while maintaining reliable task execution.

This update further intensifies competition among Google, OpenAI, and Anthropic in the AI agent platform space. As major vendors roll out managed agent services, model diversity, customizability, and deployment flexibility are becoming key differentiators.