souzamarcos/mcp-server

MCP Server Example

This is an example implementation of an MCP (Model Context Protocol) server using Spring AI. The application demonstrates how to create tools that can be called through the MCP protocol.

Prerequisites

• Java 21 or higher
• Gradle 8.x (or use the wrapper included in the project)

Project Structure

src/
├── main/
│   └── java/
│       └── com/
│           └── marcos/
│               └── mcpserver/
│                   ├── McpServerApplication.java
│                   └── core/
│                       └── service/
│                           └── WeatherService.java
└── test/
    └── java/
        └── com/
            └── marcos/
                └── mcpserver/
                    └── core/
                        └── service/
                            └── WeatherServiceTest.java

Features

The project includes an example service (WeatherService) that exposes three MCP tools:

1. getWeather: Returns weather information for a specific city
2. getWeatherForecastByLocation: Returns weather forecast for specific coordinates
3. getAlerts: Returns weather alerts for a US state

How to Run

Using Gradle Wrapper

# Run the application
./gradlew bootRun

Using the JAR

# Generate the JAR
./gradlew build

# Run the JAR
java -jar build/libs/mcp-server-0.0.1-SNAPSHOT.jar

Tests

To run unit tests:

# Run all tests
./gradlew test

# Run a specific test
./gradlew test --tests "com.marcos.mcpserver.core.service.WeatherServiceTest"

Configuration

The MCP server configuration is in the application.yml file:

spring:
  ai:
    mcp:
      server:
        enabled: true
        stdio: true
        name: mcp-server
        version: 1.0.0

Using MCP Server with Copilot

To use this MCP server with GitHub Copilot, you need to configure the mcp.json file located at:

• Windows: %LOCALAPPDATA%\github-copilot\intellij\mcp.json
• macOS: ~/Library/Application Support/github-copilot/intellij/mcp.json
• Linux: ~/.config/github-copilot/intellij/mcp.json

Add the following configuration to the file:

{
    "servers": {
        "weather-mcp-server": {
            "type": "stdio",
            "command": "java -jar",
            "args": [".\\build\\libs\\mcp-server-0.0.1-SNAPSHOT.jar"],
            "cwd": "C:/Repository/mcp-server"
        }
    }
}

Alternatively, you can use Gradle to run the server:

{
    "servers": {
        "weather-mcp-server": {
            "type": "stdio",
            "command": "C:/Repository/mcp-server/gradlew.bat",
            "args": ["bootRun"],
            "cwd": "C:/Repository/mcp-server"
        }
    }
}

Configuration Details

• weather-mcp-server: Unique name to identify your server
• type: Server type (use "stdio" for standard input/output communication)
• command: Command to run the server (can be java -jar or the gradlew path)
• args: Arguments to run the application (JAR path or bootRun)
• cwd: Working directory where the server will run

Note: Adjust the paths according to your project location.

Build

To perform a complete project build:

# Clean and rebuild the project
./gradlew clean build

# Build without running tests
./gradlew build -x test

Development

To add new tools to the MCP server:

1. Create a new service class
2. Annotate methods with @Tool
3. Configure the tool name and description
4. Spring AI will handle exposing these tools through the MCP protocol

Example:

@Service
public class MyService {
    @Tool(name = "myTool", description = "Tool description")
    public String myTool(String input) {
        return "Result";
    }
}

Technologies Used

• Spring Boot 3.5.0
• Spring AI (MCP Server) 1.0.0-M6
• JUnit 5 for testing
• Gradle as build system

References

• Spring AI Documentation
• Model Context Protocol Specification