[Workshop]
Demystifying the Agentic Loop: Building Safe, Server-Driven Agents with Spring
Adib Saikali - Broadcom
In this workshop, you’ll learn how an agentic loop is built from first principles. Using MoneyMate—a Spring Boot application in a realistic financial domain—we’ll walk through how a simple request/response interaction evolves into a full agentic system that can reason, act, observe results, and decide whether to continue or stop.
Rather than treating agents as black boxes, the workshop makes the loop explicit in code. You’ll see how decision making, action invocation, and state observation fit together, and how common agentic architecture patterns—loop orchestration, termination conditions, and error handling—shape reliable behavior. Along the way, we’ll examine why naïve, prompt-driven agents are fragile and how an explicit control loop makes them more robust.
The workshop then focuses on HATEOAS as a key pattern for simplifying agentic loops. You’ll learn how server-driven hypermedia moves workflow and constraints out of prompts and back into the system, allowing the API to guide what an agent is allowed to do at each step. Using Spring HATEOAS alongside Spring AI and MCP, you’ll see how these pieces work together: MCP provides a stable execution surface for the agentic loop, while hypermedia dynamically constrains and directs agent behavior.
By the end of the session, you’ll have a clear mental model of the agentic loop, a concrete understanding of how server-driven APIs improve agent safety, and practical architectural patterns you can apply in real Spring applications—whether or not hypermedia is part of your final design.