TLDR: JetBrains has introduced Koog, an open-source, Kotlin-based framework designed to empower developers in building sophisticated and reliable AI agents within the JVM ecosystem. Announced at KotlinConf 2025, Koog aims to make Kotlin a premier language for AI development by offering intuitive, composable, and powerful tools for creating intelligent, autonomous agents.
At KotlinConf 2025, JetBrains made significant strides in the realm of artificial intelligence, prominently featuring the unveiling of Koog, an innovative open-source framework for developing AI agents in Kotlin. This initiative underscores JetBrains’ dual commitment: enabling developers to build AI applications with Kotlin and providing Kotlin developers with AI-powered tools to enhance their development process.
Koog is designed to facilitate the creation of intelligent, autonomous agents that can understand context, take initiative, and operate independently. The framework offers a modern Kotlin DSL (Domain-Specific Language) and comes equipped with pre-built workflows and patterns that developers can utilize directly or combine to suit their needs. This allows for the integration of Large Language Model (LLM) intelligence with Kotlin’s renowned expressiveness.
Prior to Koog, a comprehensive Kotlin-native agentic framework solution was not available. JetBrains’ internal experience in agentic development led to the refinement of approaches now integrated into Koog, making it intuitive, composable, and capable of addressing real-world AI challenges. Key features of Koog include its pure Kotlin implementation, eliminating external service dependencies, a modular feature system for extending agent capabilities, and robust tool integration for agents to access external systems. It also provides out-of-the-box solutions for common use cases like streaming from LLMs, invoking multiple tools from a single LLM request, generating structured outputs, and creating history compression strategies.
Beyond Koog, JetBrains is also enhancing developer experience with AI-powered tools deeply embedded in their Integrated Development Environments (IDEs). These tools offer functionalities such as AI-powered chat for codebase interaction and context-aware answers to project-specific questions. The code completion feature is driven by Mellum, JetBrains’ proprietary LLM, specifically optimized for code-related tasks. Mellum is noted for its speed and accuracy, with a version fine-tuned for Kotlin, and is also being open-sourced.
Furthermore, JetBrains showcased Junie, an AI coding agent capable of handling complex development tasks across various Kotlin applications, including server-side, Android, and Kotlin Multiplatform.
These AI advancements were part of a broader set of announcements at KotlinConf 2025, which also included updates to Kotlin 2.2, the K2 compiler becoming the default in IntelliJ IDEA 2025.1 (resulting in over 40% faster compilation times for large codebases), and the progression of Amper, an experimental build tool for Kotlin and the JVM. The Kotlin Foundation also saw growth with new members like Block, Inc. and Meta, and its Grants Programme was reopened to support open-source contributions.
Also Read:
- PlatformCon 2025: Three Pivotal Shifts Towards AI-First Platform Engineering
- Advancements in AI Memory Frameworks Promise More Efficient and Robust Agents
Overall, KotlinConf 2025 highlighted Kotlin’s evolution into a comprehensive ecosystem, not just a language, with a clear vision for the future of software development, particularly in the burgeoning field of AI.
