Building AI Agents in Kotlin • Anton Arhipov • YOW! 2025

A conference talk introducing Koog, JetBrains' Kotlin-based AI agent framework. The speaker demonstrates building agents with tool calling, graph-based strategies, subgraphs, and structured output using local LLMs via Ollama. Key topics include why Kotlin on the JVM is preferred over Python for JetBrains' use case, the challenges of reliable tool calling with smaller local models, type-safe graph DSL for defining agent workflows, and mapping LLM responses to domain objects. A simple coding agent achieving ~50% on SWE-bench with 50 lines of code is also discussed.