# Quarkus Insights #247: Agentic Orchestration with LangChain4j and Kubernetes

DevFeed: [Quarkus Insights #247: Agentic Orchestration with LangChain4j and Kubernetes](<https://devfeed.tech/articles/quarkus-insights-247-agentic-orchestration-with-langchain4j-and-kubernetes-43875.md>)

Original publisher: [Read original article](<https://quarkus.io/blog/quarkus-insights-247-agentic-orchestration/>)

Author: James Cobb (https://twitter.com/insectengine)

Published: 2026-05-12T00:00:00Z

Content type: article

Language: en

Sources: [Quarkus - Supersonic Subatomic Java](<https://devfeed.tech/sources/quarkus-supersonic-subatomic-java.md>)

Topics: [Quarkus](<https://devfeed.tech/topics/quarkus.md>), [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [langchain4j](<https://devfeed.tech/topics/langchain4j.md>), [Orchestration](<https://devfeed.tech/topics/orchestration.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Messaging](<https://devfeed.tech/topics/messaging.md>), [Infrastructure as code](<https://devfeed.tech/topics/infrastructure-as-code.md>), [Homelab](<https://devfeed.tech/topics/homelab.md>), [opentofu](<https://devfeed.tech/topics/opentofu.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [infrastructure-as-code](<https://devfeed.tech/tags/infrastructure-as-code.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [langchain4j](<https://devfeed.tech/tags/langchain4j.md>), [messaging](<https://devfeed.tech/tags/messaging.md>), [ollama](<https://devfeed.tech/tags/ollama.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [quarkus](<https://devfeed.tech/tags/quarkus.md>), [spring-ai](<https://devfeed.tech/tags/spring-ai.md>), [spring-boot](<https://devfeed.tech/tags/spring-boot.md>)

## AI overview

A Quarkus Insights episode presents a distributed agentic AI system built with Quarkus, LangChain4j, and Kubernetes. The system uses independent agent pods, NATS messaging, MCP tools, ConfigMap-based prompts, and a custom Kubernetes operator, supported by local GPU inference in a home lab.

## Source excerpt

This summary was generated using AI, reviewed by humans - watch the video for the full story. Quarkus Insights #247: Agentic Orchestration with LangChain4j and Kubernetes In episode 247 of Quarkus Insights, Jonathan Johnson, an independent software architect and adjunct professor at Trinity College (Connecticut), shared his innovative work on distributed agentic AI systems running on Kubernetes with Quarkus and LangChain4j. The Journey: From Spring Boot to Quarkus Jonathan's journey began with a home lab project using Spring Boot and Spring AI. After attending a Connecticut Java Users Group talk by Eric Deandrea (from the Quarkus team), he was inspired to explore LangChain4j with Quarkus. Using Claude AI for code conversion, he successfully migrated his Spring Boot application to Quarkus with LangChain4j in approximately one hour. Key migration benefits observed: - Faster startup times - Reduced memory footprint - Lower CPU usage - Native compilation with GraalVM support The Home Lab Setup Jonathan built an impressive home lab infrastructure to support his distributed AI experiments: Three custom-built rigs running Proxmox as the hypervisor GPU resources: NVIDIA 4090 (24GB) and 5090 (32GB) for local model inference Total memory: 480GB RAM across the cluster Storage: 70TB NAS using TrueNAS Kubernetes: Vanilla Kubernetes deployed via kubeadm Infrastructure as Code: OpenTofu scripts for reproducible deployments The entire setup runs Ollama on Kubernetes with GPU passthrough, enabling local LLM inference without cloud dependencies. Distributed Agentic Architecture: A Different Approach Rather than using traditional LangChain4j annotations for sequential or parallel agent pipelines within a single application, Jonathan implemented a distributed round-table architecture using NATS messaging: Traditional Approach (Monolithic) @SequenceAgent @ParallelAgent @SupervisorAgent // All agents in one compiled application Distributed Approach (Kubemoot) Agents as Kubernetes pods: E