Introduction

docker-agent is a powerful, customizable multi-agent system that lets you build, run, and share AI agents using simple YAML configuration.

What is docker-agent?

docker-agent is an open-source tool by Docker that orchestrates AI agents with specialized capabilities and tools. Instead of writing code to wire up LLMs, tools, and workflows, you declare your agents in YAML — their model, personality, tools, and how they collaborate — and docker-agent handles the rest.

Why docker-agent?

After spending years building AI agents using various frameworks, the Docker team kept asking the same questions:

• How do we make building agents less of a hassle? — Most agents use the same building blocks. docker-agent provides them out of the box.
• Can we reuse those building blocks? — Declarative YAML configs mean you can mix and match agents, models, and tools without rewriting code.
• How can we share agents easily? — Push agents to any OCI registry and run them anywhere with a single command.

docker-agent is built in the open so the community can make use of this work and contribute to its future.

How it Works

At its core, docker-agent follows a simple loop:

1. You define agents in YAML — their model, instructions, tools, and sub-agents.
2. You run an agent via the TUI, CLI, or API.
3. The agent processes your request — calling tools, delegating to sub-agents, and reasoning step by step.
4. Results stream back in real-time via an event-driven architecture.

# A minimal agent definition
agents:
  root:
    model: openai/gpt-4o
    description: A helpful assistant
    instruction: You are a helpful assistant.
    toolsets:
      - type: think

# Run it
$ docker agent run agent.yaml

What’s Next?