Koog, JetBrains' agent orchestration framework, now integrates with Spring AI. The integration lets developers keep their existing Spring AI LLM providers, vector databases, and chat memory backends while layering Koog's advanced agent runtime on top. Key additions include type-safe graph-based workflows, built-in persistence for fault-tolerant execution, sophisticated history compression for cost control, and native parallel execution via coroutines. Setup requires adding three Koog starter dependencies and auto-configured beans bridge Spring AI interfaces to Koog equivalents. A customer support agent example demonstrates progressively adding intent classification, subgraph routing, durable execution with PostgreSQL checkpoints, and token-based history summarization.
Table of contents
Spring AIKoog’s Spring AI integrationWhen a pure Spring AI agent reaches its limitWhat does Koog’s Spring AI integration enable?What can you add on top?The bottom lineSort: