Building AI agents today requires juggling SDKs, APIs, and orchestration code — and switching models means rewriting everything. Docker cagent and GitHub Models fix that.

Docker cagent is an open-source multi-agent runtime that lets you define agents, tools, and delegation rules declaratively in YAML, then run with a single command: docker cagent run. No frameworks, no boilerplate — just portable, reproducible agent orchestration.

GitHub Models provides unified access to 40+ LLMs (OpenAI, Anthropic, Meta, DeepSeek) with one token and endpoint. Swap GPT-5 for Claude or Llama by changing one line — no code changes, no lock-in.

This session comes from real-world experience: the speaker collaborated with Docker’s product team, contributed the GitHub Models provider integration for cagent, and authored Docker’s official blog introducing it.

How to define multi-agent teams in YAML, connect MCP tools for APIs and databases, and package agents as OCI artifacts deployable anywhere.
Live Demo: Will build a multi-agent system, push it to Docker Hub, and run it on a new machine — proving full portability.

Ideal for developers and platform engineers building scalable, AI systems.