Docker Agent is Docker’s open-source agent builder and runtime: define AI agents in simple YAML files, give them models and tools, and run or share them using the Docker workflows developers already know. Developed by Docker Engineering on GitHub, it has attracted more than 4,500 stars and is updated continuously.
Docker Agent treats an agent as a configuration artifact rather than a codebase. A short YAML file declares the model, instructions, toolsets and MCP servers an agent can use; the runtime handles the agent loop, tool execution and multi-agent delegation, where a root agent can hand work to specialized sub-agents. It connects to models from OpenAI, Anthropic, Google and local runtimes, can expose agents through APIs, and packages agents as OCI artifacts so they can be pushed to and pulled from registries like Docker Hub, versioned and shared like container images. A built-in catalog of toolsets covers filesystems, shell access, web fetching and more.
Docker Agent also supports composing teams from agents stored as separate registry artifacts, so organizations can maintain a shared library of reviewed, versioned agents. Because definitions are plain YAML kept in version control, changes to agent behavior can be code-reviewed like any other configuration change.
Docker Agent is free and open source under the Apache-2.0 license. You bring your own model API keys or local models; Docker charges nothing for the runtime itself.
Docker Agent is for developers and platform teams who want to build, run and distribute AI agents with container-native tooling, instead of adopting a heavyweight agent framework for every project.