Koog 1.0 Stabilizes Kotlin AI Agent Framework
Koog 1.0 locks APIs for one year in JetBrains' Kotlin AI agent framework, adding local Android LiteRT inference and OpenTelemetry. JVM teams can build production agents without breaking changes. Adopt stable core first, then add beta features as needed.
Why it matters
JetBrains has declared Koog 1.0 stable, guaranteeing no breaking changes in core modules for at least one year. For teams building AI agents in Kotlin or Java, this removes the biggest blocker to production adoption: framework churn. Koog runs on JVM and Kotlin Multiplatform, so you can embed agentic workflows complete with tools, memory, and persistence directly into existing backend services or Android apps without maintaining separate Python microservices or bridging across languages.
What changed
The 1.0 release adds local Android AI through LiteRT model support, letting you run inference on-device instead of routing every request to the cloud. Java interop received a full redesign with a cleaner and more consistent API, and HTTP transport is now fully decoupled so you can plug in your own client instead of fighting the default. OpenTelemetry support spans all targets including Kotlin Multiplatform, giving you unified observability across JVM, Android, and native builds. Persistence and memory improved for long-running agents, and Anthropic prompt caching is now supported to cut latency and token costs on repeated prompts. The release also cleans up APIs and adds migration helpers to ease the upgrade path from earlier versions.
What to watch
Koog splits features into stable and beta modules. Stable modules carry the one-year guarantee, while beta modules cover still-evolving functionality. If you are already using Koog, migrate your core dependencies to the stable channel and isolate beta usage to specific features. If you are evaluating agent frameworks, compare Koog against Spring AI or LangChain4j on JVM interop and multiplatform reach. The local Android support also makes Koog a rare first-class option for on-device agents in the Kotlin ecosystem, especially if you need offline inference or strict data residency.
Interview question
When migrating to Koog 1.0, how should teams handle the stable and beta module split to align with JetBrains' support guarantees?
- a.Adopt beta modules for core agent logic and stable modules for peripheral utilities
- b.Migrate all code to stable modules and defer beta features until they mature
- c.Move core dependencies to stable modules while isolating beta features to specific use casesCorrect
- d.Keep all modules on the beta channel to lock in the one-year API guarantee
Why? this is the answer
Koog 1.0 guarantees one year of API stability only for stable core modules, so teams should migrate core dependencies there while isolating beta features to specific use cases. Option B is tempting but wrong because the framework explicitly supports adding beta features as needed rather than avoiding them entirely.
Just read this? Test yourself on what you have been reading.
Read the original → blog.jetbrains.com
- #kotlin
- #ai-agents
- #jvm
- #android
- #observability
You just looked this up. Could you explain it out loud?
That is the part interviews actually test. Tezvyn takes questions like this one and gives you what the interviewer is really checking, the answer that lands, and the mistake that ends the conversation, in the four minutes before your next meeting.
The iPhone app is on the way
We are building it. Until it lands, nothing here is held back from you: every interview card, your saved cards, streaks and the job board all work in Safari, plus hundreds of free practice quizzes of thirty questions each. Sign in and it all carries over to the app the day it arrives.
Want it as an icon? Tap Share at the bottom of Safari, then Add to Home Screen. It opens full screen and the cards you have read stay available offline.
We are hiring for this. Open roles that interview on kotlin — each one lists the topics its interview covers.
See open roles