JetBrains has launched Koog for Java, bringing its AI agent framework to the JVM with a fully idiomatic Java API. The framework lets Java teams build AI agents directly in existing backends without Python microservices. Key features include multiple workflow strategies (functional, graph-based, and GOAP planning), Spring Boot integration, support for major LLM providers (OpenAI, Anthropic, Google, DeepSeek, Ollama), fault-tolerant persistence (Postgres, S3, disk) that resumes from the exact failure point, OpenTelemetry observability with Langfuse and W&B Weave support, history compression to manage token costs, and fine-grained Java thread pool control via ExecutorService configuration.

9m read timeFrom blog.jetbrains.com
Post cover image
Table of contents
What you get with Koog for JavaSimple setupPredictable workflows with custom strategiesPersistence for fault toleranceObservability with OpenTelemetryHistory compressionManaging Java threadsTry Koog for Java
1 Comment

Sort: