[Workshop]
From Completions to Goals: AI Agents with Spring AI
Mark Pollack - Broadcom
AI coding agents have moved beyond novelty into practical utility. Tools like Claude Code and Gemini CLI can now meaningfully participate in real development workflows—classifying issues, reviewing pull requests, and improving test coverage. But how do you integrate these capabilities into your Java applications? And how do you know if they’re actually working?
This hands-on workshop introduces a paradigm shift: from using Spring AI’s ChatClient for chat completions to using AgentClient for achieving goals. You’ll climb the agentic stack—from invoking off-the-shelf CLI agents through unified APIs, to building custom agentic loops with curated tools, to verifying results with the Judge framework. You’ll also explore emerging protocols: ACP for standardized client-to-agent communication and A2A for agent-to-agent collaboration.