Rod Johnson (로드 존슨), creator of the Spring Java development framework, unveiled Embabel, a Java-based agent AI framework, at Microsoft’s JDConf developer conference.

According to The New Stack, Embabel is an Apache-licensed open-source framework and has been released on GitHub. It is based on Spring Boot, written in Kotlin and fully compatible with Java. Johnson said Java developers familiar with the Spring ecosystem can adapt quickly.

Johnson’s goal with Embabel is to solve predictability problems that companies face in AI.

Generative AI such as ChatGPT is powerful, but results are hard to predict if used as-is in corporate business processes. Existing Java code is predictable but lacks flexibility. Embabel focuses on finding a balance in between.

A core element of Embabel is GOAP, a non-LLM AI pathfinding algorithm.

Existing approaches have been either to let an LLM decide on its own or to have humans prewrite every case. The former makes outcomes hard to predict, and the latter reduces flexibility. With GOAP, if only the possible action units are defined, it can find the best path on its own to meet a goal and explain why it chose that path, enabling verification and auditing.

Embabel supports multiple LLMs by default, including OpenAI, Anthropic and Meta Llama. It allows different models to be assigned at each step of a workflow. It also supports integration with coding agents such as Claude Code and GitHub Copilot.

Johnson said his goal is to demonstrate through Embabel that Java is not inferior to Python-based solutions in agent-based systems, particularly in enterprise systems, and is instead ahead. "Java is driving innovation in generative AI," Johnson said.