Docker Agent

Run configurable AI agent teams from a CLI, connect local or cloud models, use MCP tools, and share agent definitions through OCI registries.

Docker Agent runs individual AI agents or teams from configuration files. Developers can use it for coding assistants, research tasks and automation workflows.

Agent definitions: Describe instructions, models, tools and teammates in YAML or HCL.
Coordinated teams: Give specialists different roles and let a coordinator delegate tasks such as code changes and reviews.
Tools and retrieval: Let agents edit files, execute commands, retain memory and retrieve local documents, with hybrid search, reranking and MCP tools.
Terminal and API access: Work interactively in a terminal or connect applications through HTTP, an OpenAI-compatible API, MCP or A2A.
Agent sharing: Distribute agent definitions through OCI registries, including Docker Hub.

The Apache 2.0 runtime supports local models through Docker Model Runner or Ollama. OpenAI, Anthropic, Gemini and AWS Bedrock backends send requests to the selected cloud provider and require its credentials. MCP tools can run locally, remotely or in Docker containers. The tool collects anonymous usage telemetry.