Google Releases Interactions API with Enhanced Gemini Deep Research Agent

April 21, 2026

Details

  • Google AI Developers announced a major update to the Gemini Deep Research Agent, now accessible via the new Interactions API in public beta through Google AI Studio.
  • The API provides a unified RESTful endpoint for interacting with Gemini models like Gemini 3 Pro and agents, simplifying agentic app development.
  • Key features include server-side state management to offload history handling, interpretable data models for debugging complex interactions, background execution for long tasks, and remote MCP tool support.
  • Gemini Deep Research (Preview), labeled deep-research-pro-preview-12-2025, handles complex, long-horizon research workflows, synthesizes findings into reports, and supports arbitrary MCP with rich visualizations and planning before execution.
  • Developers can start building immediately using Gemini API keys; future expansions include custom agents, Agent Development Kit support, and Vertex AI integration.
  • The update contrasts with prior generateContent by adding agent-specific capabilities like interleaved messages, tool calls, and reduced client-side complexity.

Impact

Google's Interactions API positions Gemini as a stronger contender in agentic AI development, matching OpenAI's recent agent toolkit advancements and Anthropic's tool-use expansions by offering server-side state and built-in research agents in a single endpoint. This lowers barriers for developers building multi-step workflows, potentially accelerating adoption in research and enterprise apps while narrowing the gap with rivals through async execution and MCP integration. Coming soon to Vertex AI, it aligns with cloud trends, pressuring competitors to enhance API robustness amid rising demand for production-ready agents.

Read more