Google Agent Development Kit (ADK)

// Overview

Agent Development Kit (ADK) is an open-source, code-first Python framework from Google for building, evaluating, and deploying AI agents. You define agents in Python and stay in control of how they behave, rather than wiring everything through a UI or config files.

It is aimed at developers who want to build single agents or coordinate several agents together. ADK 2.0 adds a graph-based Workflow Runtime for composing deterministic execution flows (routing, fan-out/fan-in, loops, retry, human-in-the-loop, and nested workflows) and a Task API for structured agent-to-agent delegation.

As a general agent framework, it gives you the building blocks to define agents, connect them into workflows, run them locally through a CLI or web UI, and move them toward deployment.

// What it does

• Code-first Python API for defining agents with a name, model, and instruction
• Workflow Runtime: a graph-based engine for routing, fan-out/fan-in, loops, retry, state management, and nested workflows
• Task API for structured agent-to-agent delegation, including multi-turn task mode and human-in-the-loop
• Built-in local run tooling: interactive CLI (adk run) and a web UI (adk web)
• Native support for multi-agent setups composed through Workflow edges
• Optional integrations installable via the [extensions] extra

// Getting started

Install the package with pip (Python 3.10+ is required), define an agent in Python, then run it locally with the ADK CLI or web UI.

pip install google-adk

from google.adk import Agent

root_agent = Agent(
    name="greeting_agent",
    model="gemini-2.5-flash",
    instruction="You are a helpful assistant. Greet the user warmly.",
)

from google.adk import Agent, Workflow

generate_fruit_agent = Agent(
    name="generate_fruit_agent",
    instruction="Return the name of a random fruit. Return only the name.",
)

generate_benefit_agent = Agent(
    name="generate_benefit_agent",
    instruction="Tell me a health benefit about the specified fruit.",
)

root_agent = Workflow(
    name="root_agent",
    edges=[("START", generate_fruit_agent, generate_benefit_agent)],
)

# Interactive CLI
adk run path/to/my_agent

# Web UI
adk web path/to/agents_dir

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Build a single LLM-backed agent in Python with a defined model and instruction
• Coordinate several agents in a deterministic workflow with routing, loops, or fan-out/fan-in
• Delegate subtasks between agents using the Task API, including human-in-the-loop steps
• Prototype and test agents locally through the ADK CLI or web UI before deploying

// How Google Agent Development Kit (ADK) compares

Google Agent Development Kit (ADK) alongside other open-source agent frameworks & builders tools AI/TLDR tracks, ranked by GitHub stars.