Google Unveils Agent Development Kit (ADK) at Cloud NEXT 2025

he AI landscape just got a major upgrade. At Google Cloud NEXT 2025, Google announced the Agent Development Kit (ADK), a new open-source framework designed to simplify building intelligent, multi-agent systems.

This isn’t just another tool—it’s the same tech powering agents in Google’s own products like Agentspace and the Google Customer Engagement Suite (CES). Now, developers everywhere can tap into it to create production-ready AI applications with unmatched flexibility.

What’s ADK All About?

The Agent Development Kit is Google’s answer to the challenges of building complex, autonomous AI agents. Whether you’re crafting a single agent or a team of specialized ones, ADK streamlines the entire process—building, interacting, evaluating, and deploying.

It’s flexible enough to work with any model or deployment environment, making it a go-to for developers pushing the boundaries of AI.

Key highlights include:

Multi-Agent Systems: Create modular applications with agents that coordinate and delegate tasks seamlessly.

Model Freedom: Use Gemini, Vertex AI Model Garden, or third-party models via LiteLLM (think Anthropic, Meta, Mistral AI, and more).

Tool Powerhouse: Equip agents with pre-built tools (e.g., Search, Code Exec), integrate libraries like LangChain, or even use other agents as tools.

Real-Time Interaction: Enable human-like conversations with bidirectional audio and video streaming.

Developer-Friendly: Build and debug with a CLI, Web UI, and step-by-step execution insights.

Built-in Testing: Evaluate agent performance with tools to ensure reliability.

Easy Deployment: Containerize for any platform or deploy to Vertex AI for a managed, scalable runtime.

This is just the start. ADK shines when building collaborative systems—like a WeatherAgent that fetches forecasts but hands off greetings to a GreetingAgent. Clear agent descriptions and ADK’s smart routing make delegation a breeze.

From Prototype to Production

ADK doesn’t stop at building agents. It offers robust tools to test performance against datasets and deploy seamlessly. Want to go live? Containerize your agents or use Vertex AI Agent Engine for a fully managed setup. It’s a complete lifecycle solution.

ADK vs. Genkit: What’s the Difference?

Google’s AI toolkit includes Genkit, a versatile framework for broad AI experiences. So, how does ADK stack up?

ADK: Built for complex multi-agent systems, with streaming and LiteLLM support.

Genkit: Offers flexible building blocks for diverse AI projects.

If you’re diving into intricate agent workflows, ADK’s your pick. For general AI development, Genkit’s a solid choice.

Built for Google Cloud (But Works Anywhere)

While ADK plays nice with any environment, it’s optimized for Google Cloud. Expect tight integration with Gemini models, Vertex AI, and enterprise systems via 100+ connectors. Plus, you can tap into APIs managed by Apigee and deploy at scale without duplicating data.

Why It Matters

By open-sourcing ADK, Google is handing developers the keys to the future of AI. It’s a powerful, flexible framework that tackles the toughest parts of multi-agent development—control, integration, and reliability. Whether you’re a solo coder or part of a large team, ADK empowers you to build the next generation of AI applications.

Ready to get started? Check out the ADK documentation for examples and dive in. We’ll be watching to see what the community creates with this exciting new tool.