Get started with the Live API using ADK

This tutorial shows you how to create an agent and use the Agent Development Kit (ADK) Streaming to enable voice and video communication. You install the ADK, set up a basic agent that uses Google Search, and run the agent with the adk web tool.

Before you begin

This guide assumes you have experience using a terminal in Windows, macOS, or Linux environments.

Set up your environment and install the ADK

This section shows you how to prepare your local environment.

  1. Create and activate a virtual environment. Using a virtual environment is a recommended practice.

    # Create the environment
python -m venv .venv

# Activate the environment in each new terminal
# For macOS or Linux:
source .venv/bin/activate
# For Windows CMD:
.venv\Scripts\activate.bat
# For Windows PowerShell:
.venv\Scripts\Activate.ps1

  2. Install the ADK.

    pip install google-adk

Create the project structure

Create the necessary directories and files for your agent.

  1. Create the following folder structure with empty files:

    Diagram of project structure: adk-streaming folder contains app folder, which contains .env file and google_search_agent folder, which contains __init__.py and agent.py files.

  2. Add the following code to the app/google_search_agent/agent.py file. This file defines your agent's logic. You must define a root_agent.

    from google.adk.agents import Agent
from google.adk.tools import google_search  # Import the tool

root_agent = Agent(
  # A unique name for the agent.
  name="basic_search_agent",
  # The Large Language Model (LLM) that agent will use.
  # Please fill in the latest model id that supports live from
  # https://google.github.io/adk-docs/get-started/streaming/quickstart-streaming/#supported-models
  model="...",  # for example: model="gemini-live-2.5-flash-preview-native-audio-09-2025"
  # A short description of the agent's purpose.
  description="Agent to answer questions using Google Search.",
  # Instructions to set the agent's behavior.
  instruction="You are an expert researcher. You always stick to the facts.",
  # Add google_search tool to perform grounding with Google search.
  tools=[google_search]
)

  3. Add the following code to the app/google_search_agent/__init__.py file:

    from . import agent

Set up the platform

To run the agent, configure it to use Google Cloud Vertex AI.

  1. Open the .env file located in the app/ directory.

  2. Add the following content to the file. Replace PROJECT_ID with your Google Cloud project ID and replace LOCATION with your Google Cloud location.

    GOOGLE_CLOUD_PROJECT=PROJECT_ID
GOOGLE_CLOUD_LOCATION=LOCATION
GOOGLE_GENAI_USE_VERTEXAI=True

Run the agent with the dev UI

Launch the development user interface to interact with your agent.

  1. Change your current directory to app.

    cd app

  2. Set the SSL_CERT_FILE environment variable. This step is required for voice and video tests.

    macOS/Linux

    export SSL_CERT_FILE=$(python -m certifi)

    Windows

    $env:SSL_CERT_FILE = (python -m certifi)

  3. Run the dev UI.

    adk web

  4. Open the URL provided in the terminal, which is typically http://localhost:8000 or http://127.0.0.1:8000.

  5. Select google_search_agent.

The following diagram shows how user input flows to the agent, how the agent uses the Google Search tool, and how the agent returns a response:

Diagram showing user input going to agent, agent using Google Search tool to get information, and agent returning a response to the user.

Interact with the agent

After you launch the dev UI, you can interact with your agent by using text, voice, or video.

Use text input

Enter the following prompts in the UI to test the agent's text-based responses. The agent uses the google_search tool to get the latest information to answer these questions.

  • What is the weather in New York?
  • What is the time in New York?
  • What is the weather in Paris?
  • What is the time in Paris?

Use voice and video input

To use voice input, reload the web browser and click the microphone button. Ask a question, and you hear the answer in real time.

To use video input, reload the web browser and click the camera button. Ask a question like "What do you see?", and the agent describes what it sees from the video input.

Stop the dev UI

To stop the adk web tool, press Ctrl+C in the terminal where it is running.

What's next