Data & AIBuilding AI Agents on the JVM
AI agents don’t have to be black boxes—or written in Python. In this session, we’ll build one from scratch in Kotlin using Koog, a lightweight framework for constructing tool-using LLM agents.
We’ll start by breaking down the fundamentals:
• How an agent communicates with a language model
• How tools are registered and invoked
• How to build robust interaction loops for both local LLMs and remote APIs like OpenAI or Google
From there, we’ll layer in Model Control Protocol (MCP) integrations and show how Koog’s DSL lets us define agent workflows as readable, reproducible graphs.
The session concludes with a live implementation of a coding agent—capable of generating and testing small programs—by wiring together tools such as test runners, linters, and file systems.
talkDetail.whenAndWhere
We’ll start by breaking down the fundamentals:
• How an agent communicates with a language model
• How tools are registered and invoked
• How to build robust interaction loops for both local LLMs and remote APIs like OpenAI or Google
From there, we’ll layer in Model Control Protocol (MCP) integrations and show how Koog’s DSL lets us define agent workflows as readable, reproducible graphs.
The session concludes with a live implementation of a coding agent—capable of generating and testing small programs—by wiring together tools such as test runners, linters, and file systems.
Anton Arhipov
talkDetail.shareFeedback
talkDetail.feedbackNotYetAvailable
talkDetail.feedbackAvailableAfterStart
talkDetail.signInRequired
talkDetail.signInToFeedbackDescription
occupancy.title
occupancy.votingNotYetAvailable
occupancy.votingAvailableBeforeStart
talkDetail.signInRequired
occupancy.signInToVoteDescription
comments.speakerNotEnabledComments