Google

Google Unveils Interactions API to Unify Model and Agent Development


Executive Summary

Google DeepMind has launched the Interactions API, a new unified interface available in public beta through the Gemini API and Google AI Studio. This API is specifically designed for developers building complex agentic applications, providing a single endpoint to interact with both models like Gemini 3 Pro and new built-in agents, starting with the "Gemini Deep Research" agent. It aims to simplify the development of sophisticated AI agents by introducing features like server-side state management and background execution, addressing the limitations of the older `generateContent` API for modern, stateful AI workloads.

Key Takeaways

* Product Name: The Interactions API.

* Primary Function: A unified interface to manage complex, stateful interactions with AI models and agents, simplifying the creation of agentic applications.

* Key Capabilities:

* Unified Endpoint: A single RESTful endpoint (`/interactions`) for communicating with both models and specialized agents.

* Server-Side State: Optionally offloads conversation history and context management to the server, simplifying client-side code.

* Background Execution: Allows long-running tasks to be executed on the server without maintaining a persistent client connection.

* New Agent Access: Provides access to Google's first built-in agent, Gemini Deep Research (Preview), for complex research tasks.

* Target Audience: Developers building advanced AI applications and agents.

* Availability: Available in public beta now via the Gemini API in Google AI Studio. It is also slated to be integrated into Vertex AI.

* Stated Goal: To provide a native, robust foundation for the next generation of agentic systems, moving beyond the stateless request-response model of the `generateContent` API.

Strategic Importance

The Interactions API marks a strategic move by Google to provide foundational tooling for the growing field of AI agent development. By unifying model and agent access and handling complex state management, Google is positioning its ecosystem as a more comprehensive and developer-friendly platform for building sophisticated, autonomous AI systems.

Original article