Kotlin AI Agents with Koog: Build Agentic AI Apps on the JVM

Most AI agent tutorials assume you'll learn Python. But if you build on the JVM — Android, Spring, Ktor, or any Kotlin backend — switching languages just to add AI features makes no sense. This course takes a different path: you'll build real, production-ready AI agents entirely in Kotlin, using Koog, the open-source agent framework from JetBrains.

Koog was built specifically for the JVM, with a type-safe Kotlin DSL instead of the loosely-typed Python frameworks most courses cover. That means the AI agents you build here drop straight into the applications you already work on, with the tooling, structure, and reliability the JVM world expects. And because Koog 1.0 ships with a stable core, what you learn is built to last, not to break with the next release.

You won't just watch slides. From the very first lecture, you're writing code. We start by demystifying what an AI agent actually is — how it differs from a single LLM call, how the agent loop works, and where autonomy comes from. Then you build your first working agent in just a few lines, and grow it, lecture by lecture, into something you could genuinely deploy.

Along the way you'll learn how to:

• Create your own tools and let the agent call them to take real actions

• Get structured, typed output back from a model instead of fragile text parsing

• Design multi-step workflows using Koog's graph strategies, with full control over the flow

• Make agents reliable with persistence, history compression, and cross-session memory

• Build a RAG system so your agent answers questions from your own documents, not just its training data

• Connect to external tools through MCP, and switch between cloud and local models mid-conversation

• Deploy your agent as a real web service with Ktor, then add observability, guardrails, and basic evaluation

A point I care about: you can take this entire course without spending a cent on API fees. Koog integrates with Ollama, so you can run capable models locally and for free while you learn. An OpenAI key is useful for a few lectures, but it's optional, not a paywall.

Everything is built around a single project — a developer assistant called Dev Helper — that grows with you. Each section comes with the full source code at checkpoints, so you can follow along or jump straight to any stage. There are quizzes to reinforce the concepts and a final project to make the agent your own.

By the end, you won't just understand AI agents in theory. You'll have built one from scratch, given it tools and memory, connected it to the outside world, and deployed it — all in Kotlin, on the stack you already know.

If you're a Kotlin or Android developer, a JVM engineer, or someone who's tired of every AI tutorial defaulting to Python, this is the course that meets you where you already are.