Executive Summary
Google has announced the general availability (GA) of its Interactions API, establishing it as the primary and recommended interface for developers building with Gemini models and agents. This unified endpoint simplifies development by handling both simple model inference and complex, stateful agentic tasks through a single API structure. The GA release introduces significant new capabilities, including managed sandboxed agents, asynchronous background execution, and improved tool combination, signaling a strategic shift towards supporting more advanced, agent-based AI applications.
Key Takeaways
* Unified Endpoint: Provides a single, streamlined API to interact with both standard Gemini models for inference and autonomous agents for complex tasks.
* Managed Agents: A single API call can provision a remote Linux sandbox, allowing an agent to reason, execute code, browse the web, and manage files.
* Background Execution: Long-running tasks can be executed asynchronously on the server by setting a `background=True` parameter in the API call.
* Enhanced Tooling: Developers can mix built-in tools like Google Search and Google Maps with their own custom functions in a single request, with tools now able to return images.
* New Pricing & Developer Tiers: Introduces "Flex" and "Priority" tiers, allowing optimization for cost (Flex offers a 50% cost reduction) or latency. The paid tier includes 55-day retention for past interactions.
* Official Standard: The Interactions API is now the default for Google AI Studio and all official documentation. While the legacy `generateContent` API remains supported, future frontier capabilities for agents will be exclusive to the new API.
Strategic Importance
This launch solidifies Google's strategic focus on agentic AI by providing developers with a purpose-built, stateful API designed for complex, multi-step workflows, moving beyond the limitations of traditional stateless chat completion APIs.