AI/TLDR

Google · 2026-07-01 · major

Gemini Interactions API GA — Google's unified endpoint for models and agents

Google's Interactions API leaves public beta as the default endpoint for both Gemini model inference and managed agents, with server-side state, background execution, and Deep Research on one interface.

Google Interactions API general availability announcement graphic

One endpoint for Gemini model calls and agent runs, now stable and the default across Google's AI stack.

Quick facts

MakerGoogle
StatusGenerally available
Public beta sinceDecember 2025
EndpointSingle API for models + agents
Managed AgentsDeep Research, custom agents in Linux sandboxes
StateServer-side via previous_interaction_id
SDKsPython, JavaScript

What is it?

The Interactions API is Google's new primary interface for building on Gemini, replacing generateContent as the default across AI Studio, docs, and partner integrations. A single endpoint accepts either a model ID for inference or an agent ID for Deep Research and other managed agents, with a stable schema after seven months in public beta.

How does it work?

Each call creates an Interaction resource that Google stores server-side, so a follow-up turn only needs previous_interaction_id instead of the full chat history. Managed Agents run in remote Linux sandboxes and can execute in the background, with observable step traces for debugging. Data is retained 55 days on paid, 1 day on free, and callers can opt out with store=false.

Why does it matter?

The Interactions API is where new Gemini capabilities land first — future frontier models, tools, and agent features will ship here, not on generateContent. Multi-turn calls also get cheaper because cached state means fewer input tokens per turn.

Who is it for?

Gemini API developers, agent builders, Deep Research users

Frequently asked questions

Does the old generateContent API still work?
The generateContent API remains fully supported, but Google says new Gemini models, tools, and agent features will land first on the Interactions API. Teams that stay on generateContent will still get security updates, but frontier capabilities will move to the new endpoint by default.
What are the Interactions API's managed agents?
Managed Agents are Google-hosted agents that run inside remote Linux sandboxes over background executions. Deep Research is the first shipped one; developers can also register custom agents and drive them by agent_id through the same API call they use for models.
How does Interactions API cut token cost?
Each Interaction is stored server-side, so follow-up calls only send previous_interaction_id instead of resubmitting the whole chat history. Google says that reduces input token counts and improves caching across turns, which is meaningful for multi-turn agents and long conversations.
How does Google handle data on the Interactions API?
By default Interactions are stored 55 days on paid tiers and 1 day on the free tier, under Google's standard API terms. Callers can pass store=false to opt out of server-side retention entirely, though they lose previous_interaction_id chaining when they do.

Try it

Python/JS SDKs at ai.google.dev/gemini-api/docs/interactions

Sources · 2 outlets

Tags

  • google
  • gemini
  • api
  • agents
  • deep-research
  • developer-tools

← All releases · Learn AI