Building Multi-Agent Systems in Java Without Leaving the JVM

You don't need Python to build multi-agent systems. I know that's a contrarian take. The entire AI agent ecosystem -- CrewAI, AutoGen, LangGraph -- is Python-first. But if your backend is Java, your CI is Gradle or Maven, and your team thinks in terms of interfaces and generics rather than decorators and duck typing, there's no reason to introduce a second language just for agent orchestration. This post is a hands-on walkthrough. We'll build four progressively more sophisticated multi-agent systems, all in Java, all running on the JVM, using AgentEnsemble . Setup Add the dependency to your build.gradle.kts : dependencies { implementation ( "net.agentensemble:agentensemble-core:2.3.0" ) } Or if you're using Maven: <dependency> <groupId> net.agentensemble </groupId> <artifactId> agentensemble-core </artifactId> <version> 2.3.0 </version> </dependency> You'll also need a LangChain4j model provider. For OpenAI: implementation ( "dev.langchain4j:langchain4j-open-ai:1.0.0-beta4" ) Build 1: