Repolode
Submit
google

google/adk-python

Code-first Python framework from Google for building, evaluating, and deploying AI agents, optimized for Gemini but model-agnostic.

View on GitHubWebsite
Stars
21,921
Forks
4,409
Open issues
358
Last commit
7 hours ago

What is it?

Agent Development Kit (ADK) is a modular Python framework that applies software development practices to AI agent creation. You define agent logic, tools, and orchestration directly in Python, which keeps agents testable and versionable. It is optimized for Gemini but works with other models and frameworks.

The framework covers composing multiple specialized agents into hierarchies, a graph-based Workflow runtime with routing, loops, retries, and human-in-the-loop steps, and a Task API for agent-to-agent delegation. Agents can use built-in tools, custom functions, OpenAPI specs, and MCP tools. A tool confirmation flow can require explicit approval before a tool runs.

ADK includes a command-line interface and a development web UI for testing and debugging agents, an evaluation command that runs agents against evalset files, and deployment commands for Docker and Google Cloud Run.

Key features

  • 01Graph-based Workflow runtime with routing, loops, retries, and state management
  • 02Task API for structured agent-to-agent delegation
  • 03Multi-agent hierarchies built from composed specialized agents
  • 04Tool ecosystem including custom functions, OpenAPI specs, and MCP tools
  • 05Tool confirmation flow for human approval before execution
  • 06Built-in dev web UI, CLI, and evaluation command

Use cases

  • Building a multi-agent application where specialized agents handle different subtasks
  • Creating Gemini-based agents that call external APIs through OpenAPI or MCP tools
  • Testing and evaluating agent behavior with repeatable evalset files
  • Deploying an agent as a container to Cloud Run or running it locally with Docker

Quick start

  1. 1

    Install the stable release from PyPI

    pip install google-adk
  2. 2

    Create an agent file with a root_agent (copy the Agent example from the README)

  3. 3

    Run the agent in the interactive CLI

    adk run path/to/my_agent

Things to know

  • Optimized for Gemini, and some features expect Google Cloud or Gemini credentials.
  • Requires Python 3.11 or newer.
  • The development version on main may contain experimental changes or bugs.

Topics

#agent#multi-agent#llm#gemini#python#agent-framework#evaluation#agentic#agentic-ai#agents

FAQ

What is adk-python?
Code-first Python framework from Google for building, evaluating, and deploying AI agents, optimized for Gemini but model-agnostic. Agent Development Kit (ADK) is a modular Python framework that applies software development practices to AI agent creation. You define agent logic, tools, and orchestration directly in Python, which keeps agents testable and versionable. It is optimized for Gemini but works with other models and frameworks.
Is adk-python free to use? What license does it have?
adk-python is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is adk-python still maintained?
Actively maintained. The last commit to adk-python was on Oct 20, 2026.
What can I use adk-python for?
Building a multi-agent application where specialized agents handle different subtasks Creating Gemini-based agents that call external APIs through OpenAPI or MCP tools Testing and evaluating agent behavior with repeatable evalset files Deploying an agent as a container to Cloud Run or running it locally with Docker
How do I install adk-python?
Install the stable release from PyPI: pip install google-adk Create an agent file with a root_agent (copy the Agent example from the README) Run the agent in the interactive CLI: adk run path/to/my_agent
Where can I get adk-python?
The source code and installation instructions are on GitHub: https://github.com/google/adk-python

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects