Docker has introduced docker-agent, a CLI plugin that enables users to build, run, and share intelligent AI agents using declarative YAML configuration. The tool supports multi-agent orchestration and requires no coding, allowing teams of specialized agents to collaborate on complex tasks. Docker-agent is pre-installed with Docker Desktop version 4.63 and above and is also available via Homebrew.
Users define AI agents in YAML files, specifying models, descriptions, instructions, and toolsets. The plugin supports a rich ecosystem of built-in tools and any MCP server, whether local, remote, or Docker-based. It is AI provider agnostic, compatible with OpenAI, Anthropic, Gemini, AWS Bedrock, Mistral, xAI, Docker Model Runner, and others. Advanced features include built-in reasoning tools, retrieval-augmented generation (RAG) with BM25 and embeddings, and the ability to package and share agents through OCI registries.
Docker-agent’s multi-agent architecture allows automatic task delegation among specialized agents, enhancing efficiency in problem-solving workflows. Its declarative YAML approach makes configurations versionable and shareable, streamlining collaboration. This positions docker-agent as a versatile tool in the growing AI agent space, comparable to other frameworks that emphasize modularity and interoperability in AI workflows.
The docker-agent CLI plugin is integrated into Docker Desktop 4.63+ and can be installed via Homebrew, enabling users to run docker agent commands directly. The project is hosted on GitHub, where users can access documentation and contribute to its development.