Rod Johnson, creator of the Spring Framework, has launched Embabel — an open-source, Apache-licensed agentic AI framework for the JVM built on Spring Boot and written in Kotlin. Embabel addresses the enterprise AI predictability problem by using GOAP (Goal-Oriented Action Planning), a non-LLM pathfinding algorithm borrowed from game development, to dynamically select execution paths at runtime rather than relying on LLMs or static state machines. Java's strong type system is a core feature: structured return types, validation annotations, and domain models are surfaced directly to LLMs as prompt context, making the model a participant in the application's type system rather than an opaque black box. The framework supports per-step model selection across providers like OpenAI, Anthropic, and Llama, and integrates with coding agents like Claude Code and GitHub Copilot. Johnson argues this approach gives Java developers a meaningful edge over Python-based frameworks like LangChain and CrewAI for enterprise AI workloads.
Table of contents
Where Embabel comes fromThe determinism problemJava’s type system as a featureModel choice as a design principleGetting startedSort: