Google Agent Development Kit (ADK) Masterclass

Learn how the Google ADK provides a cohesive model agnostic and deployment agnostic framework to build, deploy, and orchestrate AI agents and workflows across the Google ecosystem.

Explore the ADK core concepts that empower developers to build, manage, evaluate, and deploy ai powered agents using pre-built tools, memory, session management, callbacks, and orchestration.

Explore the Google ADK masterclass capabilities: multi-agent system design, tool integration, flexible orchestration with sequential, parallel, and loop agents, memory management, and developer tooling for testing and deployment.

Learn the essential development tools for the Google Agent Development Kit Masterclass, including VS Code, Python with uv, and access to OpenAI, Anthropic, and Google Studio APIs.

Learn to build your first adk agent by setting up a python project, configuring a multi agent folder, selecting a local gemma 3 model, and testing via adk web.

Learn to configure an ADK agent to use OpenAI GPT five nano or Gemini 2.0 flash via light LLM, with dot env keys and vertex AI integration.

Define a tool for the agent using a detailed docstring, attach it as a tool, and enable the large language model to invoke it to fetch weather data.

Understand the ADK architecture. Explore base and LLM agents, workflow types such as sequential, parallel, loop, and custom logic, plus the A and B primary types, and final sequential composition.

master the ADK run command and sdk cli options to launch a fast api server, use the web ui, and run interactive agent sessions from the command line.

Add and test tools in the Google ADK Masterclass agent, such as a get current time tool for New York, then register it so the agent can call it.

Explore how a sequential agent processes input through sub agents in order—destination research agent, itinerary agent, and travel optimizer agent—to deliver output for a travel planning system.

In the Google ADK masterclass, build a sequential agent with destination research, itinerary building, and travel optimization, using Google search tools and output keys to generate a travel itinerary.

Explore parallel agents that run subagents in parallel and apply this to a blog content writer system with blog writer, SEO specialist, visual creator, and social media and email agents.

Construct a parallel agent system that orchestrates blog writer, SEO strategist, visual creator, social media, and email marketing agents using real-time trend monitoring and Google Search.

Explore how session and memory manage context in multi-turn conversations with a session object and service, history, events, and temporary state, while memory serves as a knowledge base across sessions.

Create and test a session object with an in-memory session service, showcasing async session creation, state storage, event handling, and session cleanup for the ADK workflow.

Create an in-memory session driven ecommerce agent using a custom profile service, Gemini integration, and a runner to personalize responses from customer data like name, favorite category, and recent orders.

Define a structured agent output with Pydantic, create a customer state output schema, and use an output key to store consistent, validated data in session state.

Learn to build a persistent chef recipe agent using a database session service to store recipes in a local sql database, enabling add, view, and delete across sessions.

Explore how callbacks hook into an agent's execution to observe, customize, and enforce guardrails in the workflow using before and after agent, model, and tool callbacks.

Explore before and after agent callbacks with the LM agent, logging inputs, measuring processing time, and using a callback context to inject debugging and observe agent interactions.

Explore before and after model callbacks in the Google Agent Development Kit (ADK) Masterclass, filtering inputs with LM request and LM response, blocking math content, and adding emojis to outputs.

Learn to run and test ADK agents via the api server and decay web interface, using commands such as web, run, eval, deploy, and create, with fast api docs.

Practice by building agents with the ADK, applying tool calls, callbacks, memory, orchestration, and debugging traces, while staying updated on GitHub and Google Docs.