Take your AI agent from "good demo" to "real system" by giving it custom tools and specialist teammates.

In Part 2 of our ADK series, we move beyond built-in tools and show how to connect agents to external APIs and other agents. You'll build practical patterns you can reuse in production-style projects.

If you missed part 1, we have the recording and slides available here.

🎯 WHAT YOU'LL BUILD

In this hands-on workshop, you'll build two upgraded agent systems:

• Weather-Aware Planner that calls a live weather API before making trip suggestions

• Trip Data Concierge that orchestrates multiple specialist agents using Agent-as-a-Tool

You will implement these key Part 2 patterns:

• FunctionTool pattern (Python function as a tool)

• AgentTool pattern (one agent calling another agent)

• Tool context state sharing (pass data across tool calls in one flow)

💡 WHAT YOU'LL LEARN

By the end of this session, you will be able to:

• Write clean tool functions with strong docstrings so ADK can use them effectively

• Connect an ADK agent to a real-time public API (weather.gov)

• Design multi-agent delegation flows with clear responsibilities

• Use orchestrator instructions to enforce step-by-step tool usage

• Build more reliable, composable agent behavior

🛠️ WHAT YOU'LL NEED

Required (Please Bring):

• Laptop with a web browser

• Google account (for Colab access)

• Access to your Gemini API key from Part 1 setup

Recommended:

• Completed Part 1 (Agent, Session, Runner basics)

• Basic Python familiarity

No local installation required. We will run the notebook in Google Colab.

📚 WORKSHOP FORMAT

Duration: 90 minutes

Agenda:

• 0-10 min: Quick recap from Part 1 + Part 2 architecture overview

• 10-35 min: Build a FunctionTool using a live weather API

• 35-70 min: Build Agent-as-a-Tool orchestration with specialist agents

• 70-85 min: Debugging patterns, prompt/tool design tips, and Q&A

• 85-90 min: Next steps for Part 3 (memory + adaptive planning)

🎓 WHO SHOULD ATTEND

This workshop is for:

• Developers who attended Part 1 and want to go deeper

• Engineers building practical agent workflows with tool calls

• Anyone curious about multi-agent orchestration patterns in ADK

📖 BASED ON

This session continues from the official Google codelab: ADK Crash Course - From Beginner To Expert by Qingyue (Annie) Wang.

Notebook segment covered: Part 2: Supercharging Agents with Custom Tools

💬 JOIN OUR COMMUNITY

Discord: https://discord.gg/8MAHjHj

• Ask setup or debugging questions

• Share your Part 2 agent implementations

• Stay updated on future ADK sessions

📧 Questions? Email us at gdgchapelhill@gmail.com or just place a comment here on this event page, and we'll get back to you!

⚠️ DISCLAIMER

GDG Chapel Hill is an independent group; our activities and the opinions expressed here should in no way be linked to Google, the corporation. To learn more about the GDG program, visit https://developers.google.com/community/gdg/.

✅ READY TO CONTINUE?

Bring your laptop and let's supercharge your agents with real tools.