"Agentic Development with Koog: Building AI Agents on the JVM and Kotlin Multiplatform"
For experienced Kotlin, Java, and platform engineers, building AI agents is no longer about wrapping prompts around an API call. It is about designing systems that can reason, act, recover, and integrate cleanly with production software. This book introduces Koog as a serious engineering framework for agentic development, showing advanced readers how to construct reliable, inspectable, and extensible agents across the JVM and Kotlin Multiplatform ecosystems.
Across the book, readers move from Koog fundamentals and runtime choices into the core agent model, provider abstractions, tool integration, and type-safe graph workflows. It then expands into planning strategies, memory and retrieval, checkpointing, observability, reliability, and performance engineering, before addressing real-world integration with Spring Boot, Ktor, Java services, and multiplatform application boundaries. The emphasis is on architectural decisions, operational trade-offs, and production-ready implementation patterns rather than toy demos or oversimplified examples.
The book assumes confidence with Kotlin or Java, modern backend architecture, and practical software delivery concerns such as concurrency, tracing, and fault tolerance. Its distinguishing strength is that it treats agent systems as software systems first: typed, testable, observable, and deployable. Readers will come away able to design Koog-based agents that fit demanding enterprise and cloud-native environments without sacrificing flexibility or
