Google has introduced the Interactions API, a unified foundation for models and agents like Gemini 3 Pro and Gemini Deep Research. This API provides a native interface for handling complex context management in agentic applications.
The Interactions API supports Gemini’s suite of models and its first built-in agent, Gemini Deep Research (Preview). It enables developers to build and bring their own custom agents using one API. The API offers a single RESTful endpoint for interacting with models and specialized agents.
Key features include:
* Optional server-side state for offloading history management
* Interpretable and composable data model for complex agentic histories
* Background execution for long-running inference loops
* Remote MCP tool support
The Interactions API is currently in public beta, but it will provide a more robust developer experience compared to the existing generateContent API. As the primary path for standard production workloads, generateContent remains supported.
Developers can start building with the Interactions API using their Gemini API key from Google AI Studio. The company encourages feedback and testing to ensure the API solves real friction points and enables seamless experiences for agents in production.
Source: https://blog.google/technology/developers/interactions-api