Google ADK Visual Agent Builder: Build Your First AI Agent

Google Agent Development Kit(ADK) Visual Agent Builder converts plain English descriptions into working AI agents through a browser interface. The platform eliminates Python coding, YAML configuration, and manual debugging by generating agent architectures in seconds.

This tutorial covers building a workout planner agent that creates customized fitness routines based on user inputs: fitness level, available time, equipment, and training goals.

What is Google ADK Visual Agent Builder?

Google ADK Visual Agent Builder is a browser-based IDE for agent development, released with ADK 1.18.0 in November 2025. The platform turns natural language specifications into agent implementations without code.

Core components:

• Visual canvas: Displays agent architectures as graph structures with nodes representing agents and edges showing data flow
• AI Assistant: Accepts English descriptions and generates agent configurations, including model selection, instructions, and tool assignments
• Testing interface: Provides real-time chat for agent validation with execution monitoring

Agent types:

• LLM agents: Connect to language models (Gemini 2.0 Flash, Gemini 2.5 Pro) for natural language processing and responses
• Sequential agents: Execute sub-agents in fixed order, passing state between steps
• Parallel agents: Run multiple sub-agents concurrently for independent tasks
• Loop agents: Iterate sub-agent execution until conditions meet or limits trigger

Output format:

Generated configurations export as YAML files containing model endpoints, instructions, and orchestration logic. Files integrate with git workflows and deploy via CI/CD pipelines.

How to install and setup Google ADK

The first step is installing the ADK library through Python’s package manager.

Step 1: Install ADK library and verify version

Open a terminal and run the following commands to install ADK and check the version:

# Install or upgrade to the latest ADK version 
pip install --upgrade google-adk 

# Verify the installation 
adk --version 

Copy to clipboard
Copy to clipboard

Step 2: Configure Gemini API credentials

The Visual Agent Builder needs an API key to access Gemini models. Google AI Studio provides free API keys with usage quotas.

Navigate to Google AI Studio in a browser and click “Get API Key” in the bottom left corner. Select “Create API Key” and choose an existing Google Cloud Project or create a new one. Copy the API key immediately after generation.

Set the API key as an environment variable:

# Windows PowerShell 
setx GOOGLE_API_KEY "your_actual_api_key_here" 

# macOS/Linux 
export GOOGLE_API_KEY=your_actual_api_key_here 

Copy to clipboard
Copy to clipboard

Restart the terminal after setting the environment variable. Verify the configuration:

# Windows 
echo %GOOGLE_API_KEY% 

# macOS/Linux 
echo $GOOGLE_API_KEY 

Copy to clipboard
Copy to clipboard

Step 3: Launch development server

Start the ADK web server to access the Visual Agent Builder interface:

adk web 

Copy to clipboard
Copy to clipboard

The server starts on port 8000. Open http://127.0.0.1:8000/dev-ui/ in a browser to access the interface. The landing page displays existing agent projects and controls for creating new ones.

Building the workout planner with Google ADK Visual Agent Builder

This section covers building a workout planner that generates exercise routines based on fitness level, time, equipment, and goals. The agent is created by describing requirements to the AI Assistant, which generates the complete configuration in seconds.

Step 1: Create a new agent project

Click the “+” icon in the interface toolbar to open the project creation dialog. Enter workout_planner as the agent name.

Project names become directory names where ADK stores configuration files. Use lowercase letters and hyphens instead of spaces to avoid path issues.

Step 2: Write the agent specification

The assistant panel opens after project creation. Enter the agent requirements in plain English:

Build an agent that creates personalized workout routines. The agent should collect four parameters: fitness level (beginner, intermediate, or advanced), workout duration in minutes, available equipment (bodyweight only, basic home equipment, or full gym), and primary goal (strength gain, cardiovascular fitness, or flexibility). Based on these inputs, generate a complete workout session including warm-up activities, main exercise sequence with set and rep counts, rest period guidance, and cool-down protocol. Exercise selection must match the stated fitness level. Beginners require fundamental movement patterns with technique emphasis, while advanced individuals need progressive overload strategies. Equipment availability dictates exercise substitutions. 

Copy to clipboard
Copy to clipboard

Press Enter to submit.

The AI Assistant generates the agent configuration in 15-20 seconds. The visual canvas updates with a node representing the workout planner agent. Click on the save button to save it.

Step 3: Test the agent with different scenarios

After the agent is generated, test it with different fitness scenarios to see how it adapts to recommendations. The testing panel appears at the bottom of the interface.

Click inside the chat input field and enter the first test prompt:

I'm completely new to exercise. I have 30 minutes and no equipment at home. I want to build general fitness. 

Copy to clipboard
Copy to clipboard

Press Enter to send the message. The agent processes the request and generates a workout plan tailored for beginners with no equipment.

The agent outputs a structured 30-minute workout plan including:

• 5-minute warm-up with light cardio and dynamic stretches
• 20-minute main exercise sequence with bodyweight squats, modified push-ups, alternating lunges, planks, and glute bridges
• Specific sets and reps for each exercise (3 sets of 10-15 reps, 60-90 second rest periods)
• 5-minute cool-down with static stretches
• Form cues and beginner-specific guidance

Now test with an advanced scenario. Enter the second prompt:

I've been training for 5 years. I have 90 minutes in a fully equipped gym. My goal is maximal strength development. 

Copy to clipboard
Copy to clipboard

The agent generates a 90-minute maximal strength program including:

• 10-minute warm-up with cardio and dynamic stretches
• 75-minute main sequence with compound movements: barbell back squats (5 sets, pyramid structure with 1-3 rep maximal attempts), bench press and bent-over rows (4 sets of 4-6 reps in supersets), overhead press (4 sets of 3-5 reps), Romanian deadlifts and weighted pull-ups (3 sets, 6-8 reps)
• Heavy loads with 2-3 minute rest periods between strength sets
• Accessory work with shorter rest (60-90 seconds)
• Progressive overload strategies and form-focused guidance
• 5-minute cool-down with static stretches

The agent adapts volume, intensity, and exercise selection based on training experience and available equipment.

Conclusion

This tutorial walked through building a workout planner agent using Google ADK Visual Agent Builder. The process involved creating a project, writing a natural language specification, and testing the agent with different fitness scenarios.

The key results:

• Visual Agent Builder converted English descriptions into working agents in seconds
• Agent adapted output based on inputs: 30-minute bodyweight routines for beginners, 90-minute strength programs for advanced lifters
• Testing interface validated behavior immediately without deployment

The platform eliminates manual coding while generating production-ready YAML configurations that support version control and deployment workflows.

For a deeper understanding of agentic design patterns, prompt engineering techniques, and LLM components, check out Codecademy’s How to Build AI Agents course.

Frequently asked questions

1. What is ADK Google?

ADK (Agent Development Kit) is Google’s open-source framework for building AI agent systems. Released at Google Cloud NEXT 2025, it provides Python libraries, CLI tools, and deployment utilities. Google uses ADK internally for Agentspace and Customer Engagement Suite products.

2. What are the benefits of Google ADK?

ADK provides three main benefits: flexible orchestration through workflow agents (Sequential, Parallel, Loop) and LLM-driven routing, integrated development tools including web UI and CLI for testing and debugging, and deployment flexibility supporting containerized deployments or Google’s Agent Engine service.

3. What is the difference between Genkit and ADK?

Genkit focuses on embedding AI capabilities into existing applications through API access and prompt templating for single-agent use cases. ADK specializes in multi-agent systems where multiple specialized agents coordinate complex workflows. ADK includes orchestration features like workflow agents and state management that Genkit lacks.

4. What does ADK mean?

ADK stands for Agent Development Kit. The name reflects its focus on agent-centric development where systems use language models for reasoning and decision-making. The “kit” includes Python SDK, CLI tools, web UI, evaluation framework, and deployment utilities.