Koog
ActiveDescription
A JVM framework by JetBrains for building predictable, fault-tolerant, enterprise-ready AI agents across all platforms — from backend services to Android, iOS, and in-browser environments, with built-in MCP and multi-provider LLM support.
Key Features
- Multiplatform deployment across JVM, JS, WasmJS, Android, and iOS via Kotlin Multiplatform
- Built-in fault-tolerance with automatic retries and agent state persistence for recovery
- Intelligent history compression to optimize token usage in long conversations
- Enterprise integrations with Spring Boot, Ktor, and OpenTelemetry observability exporters
- LLM switching without losing conversation history across OpenAI, Anthropic, Google, and more
- Flexible graph-based workflow engine for designing complex agent behaviors
Use Cases
💡 Building enterprise-grade AI agents for backend services with full observability
💡 Creating cross-platform mobile AI assistants that work on Android and iOS simultaneously
💡 Developing multi-agent systems with MCP protocol integration for tool orchestration
💡 Implementing RAG-powered agents with vector embeddings and knowledge retrieval
💡 Running AI agents in browser environments with WasmJS support
Categories
Quick Start
Add Maven Central dependencies (ai.koog:koog-agents:1.0.0) to your Gradle or Maven project, set your LLM API key as an environment variable, then create an AIAgent with a prompt executor and system prompt. Run agent.run() to execute tasks. Requires JDK 17+ and Kotlin 2.3.10+.