Spring Founder Returns with a New AI Agent Framework

Rod Johnson, the founder of Spring, has launched a new AI project called Embabel, which has already attracted about 3.8k stars and 381 forks on GitHub. He describes it as the most important project he has created since founding Spring.

Embabel addresses the most challenging issue of AI agents—lack of control—by adopting GOAP (Goal‑Oriented Action Planning) from game AI. Developers declare goals, actions, preconditions, and outcomes, and a deterministic planner selects the execution path, re‑planning after each step based on the updated state.

The framework separates the planning logic from the large language model (LLM). The LLM only handles reasoning‑heavy parts, while the code‑driven planner determines "how to get there," making complex agent workflows more explainable, debuggable, and auditable.

Built on Spring AI and written in Kotlin, Embabel is fully Java‑friendly. It integrates type‑safe model definitions, supports multiple models and tools, and includes human‑in‑the‑loop capabilities within a unified framework.

For Java teams, Embabel eliminates the need to switch to Python or rely on extensive prompt engineering. Existing Spring‑based business systems can be directly incorporated into agent workflows.

If Embabel can replicate Spring’s historic success in standardizing enterprise development, its true value lies not merely in wrapping model APIs but in transforming "uncontrollable AI demos" into deployable engineering systems.