Building A2A Agent using Langchain4J and SpringBoot

Engineering Blog · April 2026
📦 Spring Boot 4 · LangChain4j · SpringAI A2A Server
🧠 Ollama / Qwen3
🔗 A2A Protocol 1.0

Table of Contents

• Spring A2A Server AutoConfiguration
• LangChain4j Agent
• LangChain4j Agent Executor and Handler

---

Overview

graph LR
    classDef clientContainer fill:#1a1a1a,stroke:#4CAF50,stroke-width:2px,color:#fff;
    classDef serverContainer fill:#1a1a1a,stroke:#4CAF50,stroke-width:2px,color:#fff;
    classDef component fill:#fff,stroke:#333,stroke-width:1px,color:#000;
    classDef greenBox fill:#e8f5e9,stroke:#4CAF50,stroke-width:1px,color:#000;
    classDef blueBox fill:#e3f2fd,stroke:#2196F3,stroke-width:1px,color:#000;
    classDef internalBox fill:#e3f2fd,stroke:#333,stroke-width:1px,color:#000;

    subgraph Client ["A2A Client"]
        direction TB
        ClientSDK[A2A Client SDK]:::blueBox
        ChatClient(ChatClient):::greenBox
        Tools[tools]:::greenBox
        
        ChatClient <--> Tools
        ChatClient <--> ClientSDK
    end

    subgraph Server ["LangChain4j A2A Server"]
        direction TB
        AgentCardController[AgentCardController]:::component
        
        subgraph MC_Group ["MessageController"]
            direction TB
            MC_Node[MessageController]:::component
            
            subgraph ServerSDK ["A2A Server SDK"]
                direction TB
                Langchain4jAgentExecutor[Langchain4jAgentExecutor]:::greenBox
                Langchain4jAgent[Langchain4jAgent]:::greenBox
                
                Langchain4jAgentExecutor --> Langchain4jAgent
            end
        end
    end

    %% Step 1
    ClientSDK -- "1. .well-known/card" --> AgentCardController
    
    %% Step 2
    AgentCardController -- "AgentCard" --> ClientSDK
    
    %% Step 3 (External)
    ClientSDK -- "2. sendMessage" --> MC_Node
    
    %% Step 3 (Internal Execution)
    MC_Node -- "3. execute" --> Langchain4jAgentExecutor
    
    %% Step 4
    Langchain4jAgentExecutor -- "4. handle" --> Langchain4jAgent
    
    %% Step 5
    Langchain4jAgent -- "5. response" --> ClientSDK

    class Client,Server clientContainer;
    class MC_Group serverContainer;
    class ServerSDK internalBox;

Spring A2A Server AutoConfiguration

Spring A2A Server AutoConfiguration allows to declare AgentCard, AgentExecutor for LangChain4j Agents.However the DefaultAgentExecutor of Spring AI A2A Server implementation by design expects ChatClient and ChatClientExecutorHandler which are specific to Spring’s way of managing the A2A Server implementation.

LangChain4j Agent

📦📦 Full Code @Langchain4j-Spring-A2A-Server

Lets build a simple LangChain4j Agent using AgenticServices which allows to build agents using Builder pattern. GeneralQAAgent is a simple agent capable of managing Q&A queries from the customer backed by a language model.

GeneralQAAgent generalQAAgent = AgenticServices
                .agentBuilder(GeneralQAAgent.class)
                .outputKey("answer")
                .chatModel(chatModel)
                .build();

To make this GeneralQAAgent to participate in A2A protocol, we can define custom Langchain4jAgentExecutor and Langchain4jAgentExecutorHandler

LangChain4j Agent Executor and Handler

Lets define two Functional Interface for handling the execution request from the AgentExecutor

✨ One which accepts ChatModel and RequestContext
✨ One which accepts Agent and RequestContext

Langchain4jChatModelExecutorHandler

@FunctionalInterface
public interface Langchain4jChatModelExecutorHandler {
    String execute(ChatModel chatModel,RequestContext requestContext);
}

Langchain4jAgentExecutorHandler

@FunctionalInterface
public interface Langchain4jAgentExecutorHandler {
    String execute(Agent langchain4jAgent, RequestContext requestContext);
}

Lets define two custom AgentExecutor for handling the A2A client request

✨ Langchain4jAgentExecutor
✨ Langchain4jChatModelExecutor

Now these Executors can be used to integrate with GeneralQAAgent to make it particiapate in A2A protocol negotiation.

Final Outcome

Using Spring AI A2A Server , we are successfully able to expose any Agent/AI Service applications built using Langchain4j as A2A Server.