Currently developers face critical production challenges when building intelligent agent-based applications: managing agent lifecycles, handling authentication, and maintaining consistency across AI clients. During this demonstration, we will show how to use Docker's latest tools to build AI agents in containers, allowing you to build, share, and scale intelligent systems with the same reliability as containerized applications. This talk will feature Docker's cagent project which brings declarative YAML to multi-agent systems, letting developers create specialized AI agents that collaborate like microservices, and Docker's mcp-gateway to provide agents with scalable access to external APIs and data sources through a unified MCP gateway.