Google's Interactions API Goes Live as the Primary Way to Build Gemini Agents

Google just made the Interactions API generally available, and with it comes a clear signal: this is now the primary way to build with Gemini. It is no longer a side experiment. The API has reached general availability and is now Google's primary API for interacting with Gemini models and agents. With this GA release, the API now has a stable schema, and Google also added major new capabilities that developers asked for, including Managed Agents, background execution, and Gemini Omni support coming soon.

The problem it was built to solve

The old generateContent API was designed for single-turn inference. You send a prompt, you get a response. That model breaks down fast when you try to build anything stateful , a multi-turn chat, a long-running research task, or an agent that needs to use tools across several steps. Every call required you to resend the entire conversation history, manage state yourself, and stitch together tool calls manually.

The Interactions API offers server-side history management via previous_interaction_id, observable execution steps that make it easy to debug complex flows and render UI for intermediate events, and native support for multi-step tool use and orchestration through typed execution steps. The server holds the conversation state for you. You just pass the ID of the last interaction to continue.

What actually changed at GA

The beta launched in December 2025. Since then, Google has shipped a substantial set of new features based on developer feedback. Here is what is new at GA:

• Managed Agents: A single API call provisions a remote Linux sandbox where an agent can reason, execute code, browse the web, and manage files. Managed agents let you extend the Antigravity agent with your own instructions, skills, and data. You can customize the agent inline at interaction time, or save the configuration as a managed agent you invoke by ID.