Docker Releases Open-Source AI Agent Builder and Runtime

Docker has released docker-agent, an open-source tool designed to build, run, and share AI agents without requiring custom code. The project functions as a docker CLI plugin, enabling users to define agent behavior, tools, and orchestration logic through declarative YAML configurations.

Core Capabilities

The platform supports a multi-agent architecture where specialized agents can automatically delegate tasks to one another. Key features include:

• Provider Agnosticism: Supports major AI providers including OpenAI, Anthropic, Gemini, AWS Bedrock, Mistral, xAI, and Docker Model Runner for local inference.
• Tool Ecosystem: Integrates built-in tools with any MCP (Model Context Protocol) server, whether local, remote, or Docker-based.
• Advanced Reasoning: Includes built-in tools for thinking, todo management, and memory.
• RAG Support: Offers pluggable retrieval mechanisms using BM25, embeddings, hybrid search, and reranking.

Distribution and Installation

Agents can be packaged and shared via OCI registries, allowing them to be pulled and run anywhere. Installation options include:

• Docker Desktop: Pre-installed in version 4.63+.
• Homebrew: Installable via brew install docker-agent.
• Binary Releases: Available from GitHub Releases.

The tool is released under the Apache-2.0 license and is actively maintained by Docker Engineering, with over 3.4k stars on GitHub.