koog
Koog is a JetBrains-backed Kotlin framework for building AI agents that run on the JVM and cross-platform targets (Android, iOS, browser). It provides fault tolerance, state persistence, LLM provider switching, and integrations with Spring Boot, Ktor, and observability tools.
Key facts
Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.
| Field | Value |
|---|---|
| Repository | JetBrains/koog |
| Owner | JetBrains |
| Primary language | Kotlin |
| License | Apache-2.0 — OSI-approved |
| Stars | 4.4k |
| Forks | 444 |
| Open issues | 155 |
| Latest release | 1.0.0 (2026-05-21) |
| Last updated | 2026-07-06 |
| Source | https://github.com/JetBrains/koog |
What koog is
A Kotlin-first agentic AI framework supporting multiplatform deployment (JVM, JS, WasmJS, Android, iOS) with built-in features: agent persistence, history compression, OpenTelemetry exporters, MCP/ACP protocol support, RAG, streaming APIs, and support for OpenAI, Anthropic, Google, DeepSeek, OpenRouter, Ollama, and Bedrock LLM providers.
Get the koog source
Clone the repository and explore it locally.
git clone https://github.com/JetBrains/koog.gitcd koog# follow the project's README for install & configurationNeed it deployed, integrated, or customized instead? DEV.co ships production installs.
Best use cases
Implementation considerations
- Requires JDK 17+; verify Kotlin version (2.3.10+) alignment in existing projects to avoid compatibility issues.
- Agent persistence and history compression require careful design of checkpoint/restore logic; review documentation on state management patterns.
- LLM provider switching is built-in but requires pre-provisioning multiple API keys/credentials; design credential management securely.
- Streaming API supports parallel tool calls; ensure downstream systems (databases, APIs) handle concurrent requests safely.
- Integration with Spring Boot or Ktor is documented; plan for embedding agents into existing middleware and ensuring proper lifecycle management.
When to avoid it — and what to weigh
- Python-native ML/data science teams — Koog is JVM/Kotlin-centric; Python teams should evaluate LangChain, AutoGen, or similar Python frameworks for closer ecosystem fit.
- Non-JVM tech stacks (Node.js, Go, C#) — Framework is JVM-only; not portable to JavaScript runtimes, Go, or .NET environments without reimplementation.
- Projects requiring guaranteed long-term backward compatibility — Framework is marked 'incubator' and relatively new (launched ~May 2025); breaking changes possible before 2.0; validate stability requirements.
- Offline-first or fully local inference without external LLM APIs — While Ollama is supported, primary design targets cloud LLM providers; local-only scenarios may require significant custom integration.
License & commercial use
Apache License 2.0 (Apache-2.0), a permissive OSI license allowing commercial use, modification, and distribution with reasonable conditions (license notice, no warranty/liability).
Apache-2.0 is permissive and generally allows commercial use. Verify no additional restrictions are imposed by JetBrains in separate agreements or terms of service; review LICENSE.txt in repo for any supplementary clauses. Use in production should be validated with legal review if required by your organization.
DEV.co evaluation signals
Editorial assessment — not user reviews. Directional, with an explicit confidence level.
| Signal | Assessment |
|---|---|
| Maintenance | Active |
| Documentation | Strong |
| License clarity | Clear |
| Deployment complexity | Moderate |
| DEV.co fit | Strong |
| Assessment confidence | High |
Review credential handling for multiple LLM providers (API keys must not leak into logs/traces); leverage Spring/Ktor security frameworks. LLM outputs may require sanitization before use in sensitive contexts. OpenTelemetry exporters should be configured to not expose sensitive conversation data. No security audit data provided; assess risk profile independently for regulated industries.
Alternatives to consider
LangChain / LangChain4j
Broader ecosystem (Python + Java), more mature, wider community adoption; LangChain4j is Java-specific but less Kotlin-idiomatic.
AutoGen / Microsoft Autogen
Multi-agent orchestration strengths; Python-native; better for teams prioritizing Python ML/research workflows.
Semantic Kernel (Microsoft)
.NET/C#/Python support; enterprise Microsoft ecosystem integration; alternative if locked into .NET or Azure.
Build on koog with DEV.co software developers
Explore Koog documentation, try the quickstart example, and join the Slack community to evaluate fit for your JVM or Kotlin Multiplatform project.
Talk to DEV.coRelated open-source tools
Surfaced by semantic similarity across the DEV.co open-source index.
Related on DEV.co
Explore the category and the services that help you build with it.
koog FAQ
Does Koog support running agents offline without external LLM APIs?
Can I switch from one LLM provider to another mid-conversation without losing context?
Is Koog production-ready?
What observability providers are supported?
Software development & web development with DEV.co
DEV.co helps companies turn open-source tools like koog into production software. Our software development services cover the full lifecycle — architecture, web development, integration, and maintenance — delivered by software developers and web developers who ship. Engage our software development agency to implement or customize it for your ai frameworks stack.
Ready to Build Enterprise AI Agents?
Explore Koog documentation, try the quickstart example, and join the Slack community to evaluate fit for your JVM or Kotlin Multiplatform project.