Langchain4j is a library that enables developers to easily integrate language models and AI workflows into Java applications, gaining traction within the Java and enterprise AI communities.

With the langchain4j-agentic module, you can combine AI (and non-AI) agents into powerful but controlled workflows. In this session, Lize explores the core patterns: sequential, looping, conditional, and parallel, plus the supervisor pattern where agents decide for themselves which tasks to run. She also covers human validation strategies that keep your agents in check. Compound agents wrap entire workflows into a single building block, while AgenticScope provides control over context and a clear view of the call chain.

Through playful demos, this presentation shows agent systems that scale from small tasks to complex automation. Whether you are just curious about AI or ready to experiment in your own codebase, you grasp what is possible today, how to keep it under control, and how Java developers shape the next rise of the agents.

Recorded at Devoxx Belgium 2025.