Build Production-Ready AI Agents

A flexible, modular framework to take your agents from prototype to planet-scale.

Get Started View on GitHub

# Define an agent in a few lines
from google.adk import Agent, tool

@tool
def get_weather(city: str) -> str:
    """Gets the weather for a city."""
    return f"The weather in {city} is sunny."

agent = Agent(
    instruction="You are a helpful assistant.",
    tools=[get_weather]
)

🐍 Python

pip install google-adk

🔷 TypeScript

npm install @google/adk

☕ Java

com.google.adk:google-adk

🐹 Go

go get google.golang.org/adk

Seamlessly Integrate with Your Tools

Use ADK Skills and the Model Context Protocol (MCP) to give agents secure, reliable access to any tool or API, accelerating your workflow.

$ claude 'Build me a weather agent with ADK'

Using ADK skill + MCP server...

from google.adk import Agent, tool
from my_api.weather import fetch_forecast

@tool
def get_weather(city: str) -> str:
    """Gets the weather for a specific city."""
    return fetch_forecast(city=city)

weather_agent = Agent(
    name="weather_checker",
    model="gemini-2.0-flash",
    instruction="You provide weather forecasts.",
    tools=[get_weather],
)

✔ Agent created with tool bindings

Test, Trace, and Debug

Use the interactive web UI for deep diving into agent behavior. Test prompts, inspect tool calls, and trace multi-step interactions to debug and refine your agents locally.

Evaluate Agent Quality

Go beyond vibes. The built-in evaluation framework lets you systematically test your agents against golden datasets. Measure not just final responses, but entire agent trajectories for more reliable results.