Using MCP in Practice: From Quick Experiments to Multi-Agent Systems

MCP is quickly becoming the standard way developers connect AI applications and agents to tools and services. From automating everyday development workflows to building sophisticated AI agents, teams are increasingly using MCP servers to make AI more useful with access to real systems and data. But with so many ways to use MCP and a growing ecosystem of tools and frameworks it can be difficult to know where to start or how to scale effectively.

In this session, we’ll explore three practical approaches to using MCP servers: 

1. Docker’s MCP Toolkit and Catalog for quick experimentation 
2. Declarative multi-agent applications with cagent
3. Full agent frameworks such as LangGraph and CrewAI

You’ll learn when to use each approach, what level of control each provides, and how to apply them in practice through concrete guidance and real-world examples.