Blog About Book

A Fast Way to Get Started with Agent Development Kit on Google Cloud

July 3, 2025 • Wietse Venema

The Agent Development Kit (ADK) by Google is an open-source framework for creating AI agents. I'll show you how to get started with it using Gemini on Google Cloud and run the local web interface for development.

You can also use Gemini without Google Cloud. To learn more, go to ai.google.dev. In this guide, however, I'll focus on using Google Cloud.

Prerequisites

Before you begin, you'll need a Google Cloud project with billing enabled.

This guide requires the following tools:

  • uv: An extremely fast Python package and project manager.
  • gcloud CLI: The command-line interface for Google Cloud. Make sure you have authenticated (gcloud auth login) and set a default project (gcloud config set project [YOUR_PROJECT_ID]).

Enable the AI Platform API

Before you can use Gemini, you need to enable the AI Platform API for your project.

gcloud services enable aiplatform.googleapis.com

Create an Agent

First, create a new directory for your project and initialize it with uv.

mkdir agent-project
cd agent-project
uv init

Next, add the google-adk package to your project.

uv add google-adk

Now, create a new agent named demo-agent.

uv run adk create demo-agent

The adk create command walks you through a few setup questions. You'll be asked to:

  1. Choose a model: Select gemini-2.5-flash. You can always change this later in the generated agent.py file.
  2. Choose a backend: Select Vertex AI to access the model through Google Cloud.
  3. Confirm your Google Cloud project: If you have gcloud configured, it suggests your default project and the global region.

The command then creates a new directory demo-agent with a few files, including agent.py with this agent implementation:

from google.adk.agents.llm_agent import Agent

root_agent = Agent(
    model='gemini-2.5-flash',
    name='root_agent',
    description='A helpful assistant for user questions.',
    instruction='Answer user questions to the best of your knowledge',
)

Finally, start the ADK web UI.

uv run adk web

This starts a local web server and opens the ADK web UI in your browser. You can now chat with your agent and explore the ADK's features.

Modifying Your Agent's Configuration

You can change your agent's configuration after it has been created:

  • Model: The model is defined in demo-agent/agent.py. You can edit the model parameter in the Agent constructor to use a different Gemini model.
  • Google Cloud Project and Region: The connection to Vertex AI is configured in the demo-agent/.env file. Here you can change the GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION.

Using the Global Endpoint

Using the global region provides a single, highly available endpoint. This means that instead of being tied to a specific geographic location, your requests are dynamically routed to the most available resources to ensure uptime. You don't control which region handles the request; the system optimizes for availability. You can read more about this in the Vertex AI documentation.

Wrap Up

You've now created a basic agent with the Google ADK, configured it to use Gemini on Google Cloud, and launched the web UI. You also know how to modify the agent's configuration by editing the agent.py and .env files.

As a next step, you can start adding tools to your agent to give it more capabilities. I wrote another post to show you how to add Python functions as custom tools to your ADK agent.

* * *

More on: Google ADK: Google's open-source framework for creating AI agents →