AI tools, scored for your job
Get paid to train AI
🤝 You're in the Free & Open Source library — community-built tools you can use without paying. A different concept from our scored AI directory.
Home › Free & Open Source › Engineering › adk-python
✦ Free & open sourceEngineering

adk-python

Code-first Python framework for designing, evaluating, and running multi-agent workflows and task hierarchies.

Build and orchestrate multi-agent Python workflows

A modular framework with a graph execution engine, human-in-the-loop controls, and built-in testing UIs.

Free forever · open source · Apache-2.0
✓ Light setup✓ Browser-based✓ Real-time collaboration
8.4
ZEKAI
Powerful Python agent framework
Free & open source · Engineering
Ease of setup7
Usefulness8
Maintenance9.5
Community9.2
★ 21,690
GITHUB
Oct 2026
UPDATED
Apache-2.0
LICENSE
🛠 Don't want to self-host? We can set it up →
⚡ Quick answer

Agent Development Kit (ADK) is a Python framework for software engineers building complex, multi-agent AI applications. It helps you design graph-based agent execution flows, handle multi-turn delegation, and run interactive web evaluations locally.

Runs on
Cross-platform
Hosting
Cloud or self-host
Languages
Python
License
Apache-2.0

What it is

ADK gives you core abstractions like Agent and Workflow to manage agent instructions, state, and delegation loops. It includes a built-in development web UI, CLI evaluation commands, and human-in-the-loop tool execution safeguards. You can run agents locally or export them directly to Docker containers and cloud infrastructure.

Who it's for

This toolkit suits software engineers and computational design teams looking to write custom agent logic in Python rather than drag-and-drop builders. It is not designed for non-technical users seeking a ready-to-use no-code assistant.

Key features

Graph execution engine

Compose execution flows with support for branching, retry logic, loops, and nested sub-workflows.

Human-in-the-loop confirmation

Require manual approval or custom inputs before agents run sensitive tools.

Built-in development UI

Inspect agent decision trees, test interactions, and run evaluation datasets locally.

Multi-agent delegation

Organize specialized task agents into structured hierarchies with single or multi-turn turns.

How to start

Up and running in 3 steps

1
Install the package using pip install google-adk.
2
Define your primary logic by instantiating Agent and Workflow objects in Python.
3
Launch the local development UI using adk web path/to/agents_dir to test and evaluate interactions.
Real talk

Our honest verdict

What's great

  • Includes a built-in local web UI and CLI for debugging and evaluating agent behavior.
  • Supports code-first workflows, allowing version control and standard Python testing.
  • Provides explicit human-in-the-loop guards for tool execution.
  • Offers flexible deployment paths to Docker containers or Google Cloud Run.

Where it falls short

  • Requires Python 3.10+ programming knowledge to define agent logic and custom tools.
  • Running live models requires external LLM API keys, incurring pay-per-token API costs.
  • Cloud deployments require ongoing maintenance of cloud infrastructure and container registries.
Cost

Pricing & license

ADK is free to use under the Apache-2.0 license. Operating costs depend on the LLM API tokens consumed and the compute resources used to host your deployment.

Paid work

Open remote AI-training roles

896 open remote AI-training roles, updated November 16, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.

See all open roles →
Compare

Alternatives

Get more

Use it, or have us set it up

📄 Free PDF: adk-python for Engineering
A plain-language guide to getting your first result. Sent to your inbox.
🔔 64 people want a full video tutorial — add your vote
🛠 Don't want to set it up yourself?
We'll install, host & configure adk-python for you — and train your team.
Get it done for you ↗
Questions

FAQ

Do I have to use Google Gemini models with ADK?

While optimized for Gemini, ADK is model-agnostic and supports alternative model providers.

How do I ensure an agent does not execute dangerous code?

You can implement tool confirmation flows that pause execution and prompt a human operator for approval before running a function.

Radar

Latest AI news

✉️ The weekly free toolkit
The best new free & open-source tools for Engineering — one short email a week, in your language.
Free · 1 email/week · unsubscribe anytime
Get paid to train AI →

See Zekai first in Google