The new Docker‑Agent plugin enables developers to define AI agents declaratively in YAML files, work with multiple large‑language‑model providers, and distribute the agents through Docker’s registry.

Docker has unveiled a new CLI plugin that lets developers build AI agents using simple YAML definitions, expanding the platform’s reach into the fast‑growing generative‑AI space.

What the Docker‑Agent plugin does

The plugin, called Docker‑Agent, adds a set of commands to the Docker CLI that parse YAML files describing an agent’s purpose, tools, and interaction flow. Developers can then package the agent as a Docker image and push it to Docker Hub or a private registry, making deployment and versioning consistent with existing container workflows.

Supported large‑language‑model (LLM) providers include OpenAI, Anthropic, and Cohere, with the ability to switch providers by changing a single YAML field. The plugin also integrates retrieval‑augmented generation (RAG) capabilities, allowing agents to query external data sources during runtime.

Key features and workflow

  • Define the agent’s name, description, and model in a docker-agent.yaml file.
  • Specify tool bindings such as web search, database queries, or custom APIs.
  • Enable RAG by linking to vector stores or document repositories.
  • Build and push the agent image with a single docker agent build command.

The YAML schema is intentionally minimal: a model block selects the LLM, a tools array lists available actions, and an optional rag section configures knowledge bases. This declarative approach mirrors Docker’s existing Dockerfile paradigm, lowering the learning curve for developers already familiar with containerization.

Benefits for developers and enterprises

By treating AI agents as first‑class Docker artifacts, teams can leverage existing CI/CD pipelines, security scanning, and access controls. The plugin also supports multi‑arch builds, enabling agents to run on edge devices as well as cloud servers.

Enterprises gain a unified distribution channel for both traditional microservices and AI‑driven components, simplifying governance and reducing the operational overhead of managing separate AI platforms.

Getting started

To try Docker‑Agent, users install the plugin via docker plugin install docker/agent, create a docker-agent.yaml file, and run docker agent build .. Detailed documentation and example templates are available on Docker’s developer portal.

For a deeper look at the plugin’s capabilities, see the original announcement on AI Weekly’s coverage of Docker’s YAML AI agent builder.