What is Mcp Quarkus Lab

mcp-quarkus-lab is a laboratory project designed to help users learn how to create a MCP Server from scratch and extract tools from an existing application to the MCP Server.

Use cases

Use cases for mcp-quarkus-lab include developing chatbots, managing meeting summaries, and creating applications that require tool extraction and integration with AI services.

How to use

To use mcp-quarkus-lab, follow these steps: 1. Run the initial application in the ‘meeting-bot’ folder using ‘mvn quarkus:dev’. 2. Create your MCP Server in the ‘mcp-server’ folder and start it. 3. Refactor your initial app to an MCP Client by adding the MCP Client Extension. 4. Optionally, update the MCP Server to manage prompts.

Key features

Key features of mcp-quarkus-lab include the ability to create a MCP Server, extract tools from applications, manage prompts, and integrate with existing applications using the MCP Client Extension.

Where to use

mcp-quarkus-lab can be used in various fields including software development, AI applications, and any scenario where integration of tools and services is required.

MCP with Quarkus Mini Lab

Goal of the lab

In this lab you will learn how to create a MCP Server from scratch and extract Tools from an existing app to the MCP Server.

Pre-requisites

• Java 17/22
• Maven
• Ollama

Step 1 : Running the initial application

Go to the meeting-bot folder and run the app : mvn quarkus:dev or quarkus dev and open your browser to localhost:8080.

Play around with the chat bot, try to make the bot calling the Tools, add also a tool to discover how it works.

Step 2 : Creating your MCP Server

Go to the mcp-server folder and open the project in your editor. Create a new class that will contains your Tools, call it whatever you want, in the package that you want.

Check the documentation here.

Start your MCP Server : mvn quarkus:dev or quarkus dev , it will be running on port 8082.

Step 3 : Refactor your initial app to a MCP Client

Add the MCP Client Extension :

   <dependency>
        <groupId>io.quarkiverse.langchain4j</groupId>
        <artifactId>quarkus-langchain4j-mcp</artifactId>
        <version>0.27.0.CR1</version>
    </dependency>

Remove the Tools.java class and fix your Ai Service regarding the Tools annotation paramater.

Update your application.properties :

quarkus.langchain4j.mcp.workshop.transport-type=HTTP
quarkus.langchain4j.mcp.workshop.url=http://localhost:8082/mcp/sse

Try the app again and make sure that this time the Tools are called on the MCP Server side.

Step 4 : Extracting the prompt (expiremental)

MCP can also manage your prompt, let’s update our MCP Server by adding a prompt definition :

    @Prompt(name="meeting_bot")
    PromptMessage codeAssist() {
        return PromptMessage.withAssistantRole("""
       You are an AI answering questions about team's meeting.
       You will try to answer by summarizing the meeting with short sentences, when action needs to be taken you add a task to the system.
                """);
    }

Back to your meeting-bot app first modify the AI Service to add a parameter to the method signature and reference it in the prompt template annotation :

@RegisterAiService()
@SessionScoped
public interface Bot {

    @SystemMessage("""
      {systemMessage}
            """)
    String chat(String systemMessage, @UserMessage String question);
}

And finally modify the Websocket resource (ChatBotWebSocket) :

First inject the MCPClient :

  @Inject
  @McpClientName("workshop")
  McpClient mcpClient;

And finally update the method calling the AI Service :

    @OnTextMessage
    public String onMessage(String message) {
        McpGetPromptResult prompt =  mcpClient.getPrompt("meeting_bot",new HashMap<>());
        return bot.chat(prompt.messages().get(0).content().toString() ,message);
    }

Restart your apps and amke sure that the prompt is now served by the MCP Server.

Bonus tracks

Secure your MCP Server and Client