docker-agent is a Docker CLI plugin that enables users to define, execute, and share AI agents through a declarative YAML configuration. Agents are specified in files such as agent.yaml, where a root agent can be assigned a model (e.g., openai/gpt-5-mini), a natural‑language instruction, and toolsets—illustrated by an MCP‑type tool referencing docker:duckduckgo. The plugin supports multi‑agent orchestration, allowing teams of specialized agents to delegate tasks automatically. A rich tool ecosystem includes built‑in reasoning utilities (think, todo, memory) and a pluggable retrieval‑augmented generation (RAG) system that offers BM25, embedding‑based, hybrid search, and reranking capabilities. Agent execution is invoked via docker agent run, with options to run a local YAML file, pull an agent from an OCI registry (docker agent run myorg/agent:tag), or generate a new agent interactively (docker agent new). Installation paths include the pre‑bundled plugin in Docker Desktop 4.63+, Homebrew (brew install docker-agent), manual binary placement in ~/.docker/cli-plugins, or direct execution from GitHub releases. At least one API key must be set (e.g., OPENAI_API_KEY) or the Docker Model Runner can be used for local models. The tool also collects anonymous usage data for improvement.