Google DeepMind is moving Gemini development onto a new default path. The Interactions API is now generally available and has become the main interface for Gemini models and agents in Google AI Studio and all documentation.
The change matters because it defines where Google will put new agent capabilities next. The old generateContent interface remains usable, but the forward-looking feature set is now tied to Interactions.
What Changed For Gemini Developers
The Interactions API had been in beta since December 2025. With general availability, Google DeepMind is replacing the older generateContent interface as the default way developers are guided to work with Gemini models and agents.
That does not mean existing generateContent usage immediately stops working. The source makes clear that the older API still works. The important shift is strategic: new agent features will only ship through the Interactions API going forward.
For teams building on Gemini, this creates a clear decision point. Existing projects can continue using the old interface, but anyone planning around future agent functions now has a strong reason to evaluate the migration path.
Google has also published a migration guide for the switch. That is a practical signal that this is not just a naming change in the documentation, but a platform transition developers are expected to plan for.
Why Interactions Is Centered On Agents
The framing from Google is direct. Logan Kilpatrick, Google's developer relations lead, writes:
"Interactions sets the stage for the new era of Agents,"
The API is being positioned around multi-step activity rather than a single request-and-response pattern. That aligns with the way agent systems are described in the source: they may involve user input, tool use, background work, and media generation.
Recent additions listed with the Interactions API include Managed Agents with their own Linux sandbox. The source also points to background execution for long-running tasks, which suggests the interface is meant to support workflows that do not finish instantly.
Tool chaining is another major addition. Google Search and Maps are named as supported tools in that context, giving developers a way to connect agent steps with external information or location-based capabilities through the newer interface.
The New Step-Based Schema
Google also changed the structure of the API since the beta. The older role model, with labels such as "user" and "model," has been replaced by typed steps.
In this approach, each action is treated as its own defined step. The source identifies user input and function calls as examples of actions represented in this structure.
That change is significant because it gives the interaction a more explicit shape. Instead of organizing an exchange mainly around who produced a message, the schema organizes the work around what happened.
For agent workflows, that distinction is useful. A system that moves through inputs, tool calls, generated media, and longer-running tasks needs a record of activity that can describe each part of the process clearly.
Cost, Speed, And Media Generation
The Interactions API also gives developers a choice between Flex and Priority mode. According to the source, Flex cuts costs by 50 percent, while Priority optimizes for speed.
That split gives developers a simple tradeoff to consider. Some workloads may favor lower cost, while others may need faster responses. The source does not provide further technical detail, so the key takeaway is the availability of the two modes and their stated priorities.
Media generation is also part of the recent additions. The source lists image, music, and speech generation as supported areas, placing the Interactions API beyond text-only model calls.
Together, the additions show why Google is making this the default interface for Gemini agents. The API is designed to handle agent execution, tool use, sandboxed managed agents, long-running work, and generated media within a newer schema.
What To Watch Next
The practical message for developers is straightforward. The old generateContent interface continues to work, but the center of gravity has moved.
Future Gemini agent features will arrive through the Interactions API. That makes the migration guide the relevant starting point for teams that want to keep pace with Google's agent roadmap.
For now, the clearest facts are the status change, the replacement of generateContent as the default in Google AI Studio and documentation, the step-based schema, and the new agent-focused features. Those details define the direction of Gemini development after the beta period that began in December 2025.