Horacio Gonzalez

Large Language Models (LLMs) are revolutionizing how we build software, but harnessing their full potential requires more than just API calls. In this talk, we will introduce MCP (Model Context Protocol), a powerful approach for connecting LLMs with your applications and data, making them smarter and more context-aware. We will begin with an introduction to MCP, explaining how it enables LLMs to interact seamlessly with your tools, APIs, and data sources. You’ll discover the difference between generic and domain-specific MCP servers, learning when to choose each approach and how to design them effectively. Through practical examples in Java and JavaScript, we will demonstrate how to build both generic and domain-specific MCP servers. We will explore using LangChain4j and Quarkus for Java-based solutions, while also showcasing how to implement MCP servers using JavaScript for a broader range of use cases. You will learn how to structure your MCP servers, design efficient prompts, and optimize tool interactions to enhance LLM performance. Finally, we will cover deployment strategies, including cloud deployment, ensuring your MCP servers can scale and perform reliably. Whether you are a Java or JavaScript developer looking to leverage LLMs in your applications or simply curious about MCP’s potential, this session offers a hands-on, code-focused exploration of the concepts and best practices for connecting LLMs to your existing tools and data.