Course Overview: Build AI Agents with Agent Development Kit

Build Agents with the Agent Development Kit is an advanced-level, instructor-led course designed for software developers, engineers, platform and DevOps professionals, and technical or solutions architects who want to build enterprise-grade AI agent systems using Google’s Agent
Development Kit (ADK). The course provides a practical progression from building a simple ADK
agent to designing and deploying sophisticated multi-agent systems. Participants learn how ADK represents agents, tools, and runners, how tools are exposed to models through automatically generated schemas, and how session state enables coordination across agents. Learners then explore multi-agent orchestration using deterministic workflow agents, graph-based workflows, conditional routing, parallel
fan-out and joins, and task-based collaboration. The course also coversgrounding agents in enterprise data through ADK grounding tools,multi-source retrieval, and external Model Context Protocol (MCP)
integrations. Finally, participants learn how to deploy multi-agent systems to Agent Runtime and register and share agents through Gemini Enterprise, enabling organizational users to discover and access agent capabilities

After completing the Agent Development course, participants will be able to:

  • Describe the ADK agent model, including Agent configuration, tools, and the runner
  • Configure and run a functional ADK agent.
  • Use the ADK CLI to scaffold, run, and test agents locally.
  • Explain how ADK automatically generates tool schemas from Python functions.
  • Build custom tools using descriptive function names, docstrings, and parameter types
  • Use ToolContext to securely manage authentication, session state, and artifacts
  • Explain session state scopes and implement safe state modifications.
  • Design multi-agent workflows using shared invocation state and the output_key pattern.
  • Build multi-agent workflows using ADK’s deterministic template workflow agents.
  • Design graph-based workflows using conditional routing, parallel fan-out, and joins.
  • Configure coordinator agents and delegation modes for task-based collaboration.
  • Ground agents in enterprise data using structured and unstructured data retrieval tools.
  • Implement intent-based routing across multiple retrieval sources.
  • Explain the benefits of MCP for decoupling agent logic from tool execution.
  • Use McpToolset to integrate external MCP-based tools.
  • Describe the core capabilities of the Gemini Enterprise Agent Platform
  • Differentiate between Agent Sessions and Agent Memory Bank use cases.
  • Deploy agents using the ADK CLI and Agents CLI.
  • Deploy multi-agent systems to Agent Runtime.
  • Register and share agents through Gemini Enterprise.

Upcoming Batches

Loading Dates...

Key Features of AI Agent Development Training

  • ADK Agent Model and Core Components 

  • Building and Configuring Simple Agents

  • ADK CLI Development Workflow 

  • Custom Tools and Automatic Schema Generation 

  • ToolContext, State, Flow, and Artifacts 

  • Session State and Multi-Agent Coordination 

  • Multi-Agent Workflow Orchestration 

  • ADK 2.0 Graph-Based Workflows 

  • Task-Based Agent Collaboration 

  • Enterprise Data Grounding and MCP

  • Gemini Enterprise Agent Platform 

  • Multi-Agent Deployment and Sharing 

Who should Attend Build AI Agents with Agent Development Kit ?

  • Software Developers and Engineers
  • Platform Engineers
  • DevOps Engineers
  • Technical Architects
  • Solutions Architect
  • Cloud Architects
  • AI/ML Engineers
  • Application Developers
  • Professionals building AI agent applications
  • Professionals responsible for enterprise AI application deployment

Prerequisites for the Agent Development Kit Certification Course :

• Familiarity with Large Language Models (LLMs) and API integration. • Understanding of software architecture concepts. • Familiarity with Python programming and Python-based application development. • Understanding of basic AI/ML concepts is beneficial. • Experience working with APIs and cloud-based applications is recommended. • Familiarity with Google Cloud concepts is beneficial. • Basic understanding of agentic AI concepts is recommended.

Why choose CloudThat as your training partner?

  • Specialized Google Cloud and Generative AI Expertise CloudThat specializes in Google Cloud, Generative AI, agentic AI, and enterprise cloud technologies, delivering focused learning programs aligned with real-world enterprise use cases. 

  • Industry-Recognized Trainers 

    Our trainers are certified Google Cloud professionals with expertise in Google Cloud, Vertex AI, AI agents, Agent Development Kit, DevOps, and enterprise architecture. 

  • Hands-On Learning Approach 

    CloudThat emphasizes practical learning through guided exercises, agent development workflows, multi-agent orchestration, grounding scenarios, deployment demonstrations, and troubleshooting activities. 

  • Customized Learning Paths 

    Training programs are designed for developers, platform engineers, DevOps professionals, technical architects, and cloud professionals working with enterprise AI and agent systems. 

  • Interactive Learning Experience 

    Sessions include instructor demonstrations, guided development activities, architecture discussions, scenario-based exercises, troubleshooting workflows, and knowledge checks.

  • Continuous Learning and Updates 

    Course content is continuously updated to align with advancements in Agent Development Kit, Gemini Enterprise, agent platforms, enterprise AI, and Google Cloud technologies.

Learning Objective of Build AI Agents with Agent Development Course

  • This course enables learners to build enterprise-grade multi-agent AI systems using the Agent Development Kit. Learners will develop the ability to configure ADK agents, build custom tools, manage session state, implement multi-agent orchestration, construct graph-based workflows, collaborate through coordinator agents, ground agents in enterprise data, integrate external MCP tools, and deploy multi-agent systems to Agent Runtime.  Learners will also understand how to register and share agents through Gemini Enterprise, enabling organizations to make agent capabilities accessible to users across the enterprise.

Course Outline: Build AI Agents with Agent Development Kit Download Course Outline

  • Overview of the ADK framework and enterprise use cases
  • The four core ADK components.
  • Component Interaction and Execution Lifecycle

Learning Outcomes:

  • Describe the responsibilities of the four core ADK components.
  • Explain how the components coordinate to execute an agent’s reasoning loop.
  • Configure a functional agent by defining its essential parameters.
  • Identify the role of the agent name, model, and core instruction.

Activities:

  • Instructor-led walkthrough: ADK framework overview
  • Demonstration: Four core ADK components
  • Demonstration: Agent execution lifecycle
  • Guided discussion: Enterprise ADK use cases

  • ADK agent requirements and required parameters
  • Optional agent parameters

Learning Outcomes

  • Identify the required parameters for an ADK agent.
  • Configure a simple functional agent.
  • Apply optional agent parameters based on application requirements.
  • Understand how agent configuration influences agent behavior.

Activities

  • Demonstration: Creating a simple ADK agent
  • Guided configuration exercise
  • Instructor walkthrough: Optional parameters

  • Creating scaffolding for your project with the ADK CLI
  • Running the agent locally during development
  • Using adk web to test the agent in the browser

Learning Outcomes

  • Build an agent working directory using the ADK CLI.
  • Run agents locally during development.
  • Test an agent using the adk web command.
  • Explain the ADK development loop and how it supports agent development and testing.

Activities

  • Demo: ADK CLI project scaffolding
  • Hands-on: Create an ADK agent project
  • Hands-on: Run the agent locally
  • Demo: Test the agent using adk web

  • Automatic schema generation from Python functions
  • Tool components
  • Inter-tool dependencies

Learning Outcomes

  • Explain how ADK generates tool schemas from Python functions.
  • Describe how function names, docstrings, and parameter types inform the model’s reasoning.
  • Create tools with descriptive function names and explicit docstrings.
  • Understand how tools can interact and depend on one another.

Activities

  • Demonstration: Automatic tool schema generation
  • Guided exercise: Build a custom Python tool
  • Demonstration: Tool dependencies
  • Tool design discussion

  • ToolContext and its role within an ADK tool
  • State, flow, and artifacts
  • The four capabilities ToolContext provides

Learning Outcomes

  • Explain the role of ToolContext within an ADK tool.
  • Implement ToolContext in a custom Python tool.
  • Securely handle authentication through ToolContext.
  • Read and write scoped session state.
  • Manage large data payloads as artifacts.

Activities

  • Demo: ToolContext capabilities
  • Hands-on: Implement ToolContext
  • Hands-on: Read and write session state
  • Demo: Working with artifacts

  • The four session state scopes and their prefixes and lifespans
  • Writing state safely using managed contexts
  • State as a coordination mechanism across agents

Learning Outcomes:

  • Categorize the four session state scopes.
  • Determine the appropriate prefix and lifespan for different types of agent data.
  • Implement safe state modifications using managed contexts.
  • Use the output_key parameter to persist agent outputs.
  • Design a multi-agent workflow using shared invocation state.

Activities:

  • Demonstration: Session state scopes
  • Guided exercise: Safe state modification
  • Demo: Using output_key
  • Design activity: State-based multi-agent coordination

  • Multi-Agent Orchestration
  • Workflow Agents Collaboration

Learning Outcomes:

  • Explain when a single agent may not be appropriate for complex tasks.
  • Identify use cases involving context window limits, task specialization, and parallelism.
  • Use ADK’s deterministic template workflow agents.
  • Select an appropriate orchestration pattern based on task dependencies.

Activities:

  • Demonstration: Multi-agent orchestration
  • Demo: Sequential workflow
  • Demo: Parallel workflow
  • Scenario activity: Select the appropriate orchestration pattern

  • Graph-based workflow fundamentals: the ADK 2.0 Workflow class, nodes, and edges
  • Conditional routing within a graph workflow
  • Parallel fan-out architectures

Learning Outcomes:

  • Explain how the ADK 2.0 Workflow class represents complex execution paths.
  • Describe how nodes and edges model agent workflow execution.
  • Construct conditional routing within a graph workflow.
  • Design parallel fan-out architectures.
  • Use a JoinNode to coordinate independent agents running concurrently.

Activities:

  • Demonstration: Graph workflow fundamentals
  • Guided exercise: Conditional routing
  • Demonstration: Parallel fan-out
  • Demonstration: JoinNode orchestration

  • Coordinator agents and defined sub-agents
  • The three delegation modes and when to use each
  • Collaboration constraints in multi-agent system design

Learning Outcomes :

  • Differentiate between the three delegation modes.
  • Select appropriate delegation modes for different collaboration scenarios.
  • Configure a coordinator agent with defined sub-agents.
  • Create explicit instructions for accurate request routing.
  • Apply collaboration constraints when designing multi-agent systems.

Activities:

  • Demo: Coordinator agent
  • Demonstration: Delegation modes
  • Guided exercise: Configure sub-agents
  • Scenario activity: Route user requests using a coordinator

  • Core grounding concepts and ADK grounding tools
  • Multi-source routing and intent-based retrieval selection
  • Model Context Protocol (MCP) and decoupling agent logic from tool execution

Learning Outcomes:

  • Differentiate between structured and unstructured data grounding tools.
  • Identify risks such as schema hallucination and semantic drift.
  • Implement intent-based routing across multiple retrieval tools.
  • Design an agent that dynamically evaluates user intent and selects an appropriate retrieval source.
  • Explain how MCP decouples agent logic from tool execution.
  • Utilize McpToolset for external MCP integrations.

Activities:

  • Demonstration: Grounding tools
  • Demo: Structured and unstructured data retrieval
  • Guided exercise: Multi-source routing
  • Demonstration: MCP integration using McpToolset

  • The four core pillars of the Gemini Enterprise Agent Platform
  • Scaling features for agent applications
  • Alternative runtime options for agent applications

Learning Outcomes:

  • Describe the four core pillars of the Gemini Enterprise Agent Platform.
  • Identify the tools and services supporting the agent lifecycle.
  • Understand the platform capabilities for scaling agent applications.
  • Differentiate between Agent Sessions and Agent Memory Bank.
  • Select the appropriate mechanism for immediate conversational context versus long-term user preferences.
  • Understand alternative runtime options for agent applications.

Activities :

  • Demonstration: Gemini Enterprise Agent Platform
  • Platform capability walkthrough
  • Scenario discussion: Agent Sessions vs. Agent Memory Bank
  • Demonstration: Scaling agent applications

  • ADK CLI vs. Agents CLI and selecting the appropriate deployment path
  • SDK configuration and deployment sources
  • The four stages of the complete agent lifecycle
  • Sharing agents across the organization

Learning Outcomes:

  • Contrast the ADK CLI and Agents CLI.
  • Determine the appropriate CLI for different deployment requirements.
  • Configure SDKs and deployment sources.
  • Explain the four stages of the complete agent lifecycle.
  • Understand the importance of an accurate agent description during registration.
  • Explain how agent descriptions support AI-driven routing.
  • Deploy multi-agent systems to Agent Runtime.
  • Register and share agents through Gemini Enterprise

Activities:

  • Demonstration: ADK CLI vs. Agents CLI
  • Guided deployment walkthrough
  • Demo: Deploying to Agent Runtime
  • Demonstration: Agent registration
  • Demo: Sharing an agent through Gemini Enterprise

Select Course date

Loading Dates...
Add to Wishlist

Course ID: 31178

Course Price at

Loading price info...
Enroll Now

FAQs for Build AI Agents with Agent Development

This course is designed for software developers and engineers, platform and DevOps engineers, and technical or solutions architects who want to build and deploy multi-agent AI systems using the Agent Development Kit.

The Agent Development Kit (ADK) is a framework used to build AI agents and multi-agent systems, providing capabilities for agent configuration, tools, state management, orchestration, grounding, and deployment.

Yes. The course covers deterministic workflow agents, graph-based workflows, coordinator agents, delegation modes, parallel execution, conditional routing, and task-based collaboration.

Yes. Module 08 introduces ADK 2.0 graph-based workflows, including nodes, edges, conditional routing, parallel fan-out, and JoinNode-based workflow coordination.

Yes. Module 10 covers grounding tools, multi-source retrieval, intent-based routing, and external MCP integrations.

Talk to Us