Build AI Agents with ADK and Gemini
A workshop introducing the Agent Development Kit (ADK): Google's open-source framework for building production-ready AI agents with Gemini: covering agentic AI concepts, ADK architecture, multi-agent design, and deployment options on Google Cloud.
Events
| Name | Organizer | Date | Location | Attendees | Links |
|---|---|---|---|---|---|
| Google I/O Extended Bacolod 2025 | GDG Bacolod | 2025-08-02 | Negros Women for Tomorrow Foundation (NWTF) Head Office, Bacolod, Philippines | 80 | 📊 Slide Deck |
QR Code for this session
Abstract
AI agents represent the next major shift in how software is built: moving from static, deterministic systems to dynamic applications that reason, plan, and act. But building agents that are reliable, extensible, and production-ready is a different challenge altogether.
This workshop unpacks what AI agents actually are, how they've evolved from simple LLM prompts to complex multi-agent systems, and when (and when not) to reach for an agentic architecture. It then introduces the Agent Development Kit (ADK): Google's open-source framework designed to give developers a powerful, flexible foundation for building complex agents with Gemini: with minimal boilerplate, a built-in dev UI for debugging, and native support for multimodal streaming. The session covers the full agent-building toolkit available on Google Cloud: from no-code Conversational Agents to full-code ADK with Vertex AI Agent Engine: and close with a practical look at how to deploy ADK agents to Cloud Run or Vertex AI Agent Engine, with codelabs attendees can follow along with live.
Agenda
- What are AI agents? Evolution from prompts to multi-agent systems
- When (and when not) to reach for an agentic architecture
- Google's agent spectrum: Conversational Agents → ADK → Agent Engine
- ADK architecture: models, tools, sessions, memory, and callbacks
- Codelab: Building your first agent with ADK and Gemini
- Debugging with ADK's built-in dev UI
- Deploying your agent to Cloud Run or Vertex AI Agent Engine
Key Takeaways
- Agents reason and act dynamically - they are fundamentally different from prompt-response chatbots
- ADK minimizes boilerplate while giving you full programmatic control over agent behavior
- Google Cloud's agent spectrum lets you choose the right abstraction for your complexity and scale
- ADK's built-in dev UI makes it practical to debug and observe multi-agent flows
- Production deployment - to Cloud Run or Agent Engine - is a first-class feature, not an afterthought