Docker Agent is a CLI plugin and runtime from Docker Engineering that lets users build, run, and share intelligent AI agents using a declarative YAML configuration. It supports multi-agent teams that can delegate tasks, a rich tool ecosystem including MCP servers and built-in utilities, and provider-agnostic model support (OpenAI, Anthropic, Google Gemini, AWS Bedrock, Mistral, xAI, Docker Model Runner, etc.). Key capabilities include advanced reasoning primitives (think, todo, memory), pluggable retrieval-augmented generation options (BM25, embeddings, hybrid search, reranking), and the ability to package agents as OCI artifacts for distribution. Agents are defined with model, instruction, and toolset entries and can be run locally or pulled from registries; examples and interactive agent generation are provided.
Installation and usage are straightforward: the plugin ships with Docker Desktop (4.63+), is available via Homebrew or binary releases, and runs with commands like docker agent run, docker agent new, or by referencing registry images. The project is documented with a full docs site, examples, and a contributing guide, and collects anonymous telemetry. It is open-source under Apache-2.0, actively maintained (10,703 commits) with community channels and notable adoption (about 3.4k stars and 481 forks).
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.