Dice MCP Server (Java)

A minimal Model Context Protocol server written in Java, built as a first MCP project. It exposes two tools over the stdio transport:

Built with the official MCP Java SDK 2.0.0.

Requirements

• Java 17 or higher (tested on Java 25)
• Maven 3.9+

Build

mvn clean package

This produces a self-contained runnable jar at target/dice-mcp-server.jar.

All dependencies come from Maven Central, so a stock Maven install builds it as-is.

Run

The server speaks JSON-RPC over stdin/stdout, so you normally don't run it by hand — an MCP client (like Cursor) launches it. To run it directly:

java -jar target/dice-mcp-server.jar

It will print dice-mcp-server started (stdio). Waiting for client... to stderr and then wait for a client. Press Ctrl+C to stop.

Note: stdout is reserved for the JSON-RPC protocol, so all logging goes to stderr.

Connect to Cursor

Add the server to your Cursor MCP config (~/.cursor/mcp.json for global, or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "dice": {
      "command": "java",
      "args": ["-jar", "/absolute/path/to/dice-mcp-server/target/dice-mcp-server.jar"]
    }
  }
}

Then reload Cursor. You should see the dice server with its roll_dice and add tools. Try asking: "Roll a 20-sided die".

Quick manual test (stdio)

You can drive the protocol by hand to verify it works:

{ printf '%s\n' \
  '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' \
  '{"jsonrpc":"2.0","method":"notifications/initialized"}' \
  '{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
  '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"roll_dice","arguments":{"sides":20}}}'; \
  sleep 2; } | java -jar target/dice-mcp-server.jar

You should see JSON-RPC responses including a dice roll result.

Project layout

pom.xml                                              Maven build (MCP BOM + shade plugin)
src/main/java/com/example/dice/DiceMcpServer.java    The server + tools

Where to go next

• Add a third tool that calls a real public API (e.g. weather or a joke API).
• Add input validation and richer error messages.
• Try the add tool returning structuredContent instead of plain text.
• Rebuild the same server in Go using the official Go SDK to compare.