Google DeepMind has officially set the Interactions API as the default interface for Gemini models and agents, marking a significant shift in how developers build with the company’s flagship AI platform. The API, which entered beta in December 2025, is now generally available and replaces the longstanding generateContent interface in Google AI Studio and all official documentation. While the older API continues to function, all new agent-centric features will ship exclusively through the Interactions API going forward.
Interactions API Replaces generateContent as the Default Interface for Gemini
The transition from generateContent to Interactions represents more than a simple version bump. Google’s developer relations lead Logan Kilpatrick characterized the move as foundational for the next phase of AI development, stating that “Interactions sets the stage for the new era of Agents.” The company has published a migration guide to assist developers in transitioning their existing workflows to the new interface.
New Capabilities: Managed Agents, Background Execution, and Multi-Tool Chaining
Recent additions to the Interactions API include Managed Agents, which operate within their own Linux sandbox environment, providing isolated execution for complex tasks. Background execution support enables long-running processes without blocking the main thread, while tool chaining integrates Google Search and Maps directly into agent workflows. Media generation capabilities have also been expanded to include images, music, and speech, broadening the range of applications developers can build with the platform.
Schema Simplification: Typed Steps Replace Role-Based Labels
The API schema has been streamlined since its beta release. The older role structure, which used labels such as “user” and “model,” has been replaced by typed steps. In the new model, every action — from user input to function calls — is represented as its own defined step. This change is designed to improve clarity and flexibility when constructing complex multi-turn interactions and agentic workflows.
Flex and Priority Modes: Balancing Cost and Performance
Developers now have access to two distinct inference modes within the Interactions API. Flex mode reduces costs by 50 percent, making it suitable for high-volume or budget-sensitive applications where latency is less critical. Priority mode, by contrast, optimizes for speed, delivering faster response times for latency-sensitive use cases. This tiered approach allows developers to choose the appropriate tradeoff between cost and performance for each specific workload.
What Is the Interactions API and Why Does It Matter?
The Interactions API is Google DeepMind’s new default interface for interacting with Gemini models and building AI agents. It replaces the generateContent API and introduces support for managed execution environments, background processing, tool chaining with Google services, and multi-modal media generation. For developers, the API represents a unified foundation for building production-grade agent systems, with the promise that all future agent capabilities will be delivered through this single interface.
Who Should Migrate to the Interactions API
Developers currently using the generateContent API for Gemini should plan their migration to the Interactions API, as all new agent features will be exclusive to the new interface. The migration guide provided by Google covers the key changes, including the transition from role-based labels to typed steps. For new projects, starting directly with the Interactions API is the recommended approach. The old API remains functional for existing deployments, but adopting the Interactions API now positions teams to take advantage of future capabilities as they become available.
