Google Interactions API reached general availability and became the main entry point to the Gemini models and agents. State (the conversation's memory on the server), background tasks and tools stop being a side feature.
- Google announced general availability of Interactions API: the main entry point to the Gemini models and agents (June 22).
- One endpoint with server-side state: the app doesn't resend everything on every request.
- Inside: managed agents in a Linux sandbox, background tasks, built-in tools and media generation.
The app stops repeating everything from scratch on every request. Google announced that Interactions API reaches general availability and becomes the main entry point to the Gemini models and agents - one unified endpoint (the API's entry point) with state that now lives on the server.
Short and clear - for developers this is a signal. Server-side state, background execution, built-in tools and multimodal generation are no longer a side feature. This is the main way Google wants you to build on Gemini.
Less code here. But more of your logic now lives on Google's server, not yours. Convenience that comes tied to dependency - you pay for it in control, not in money.