Build a Spring Boot MCP Server to Let AI Safely Call Your App

What is Spring AI MCP?

Spring AI MCP framework exposes data and services as standardized tools for AI models, communicating via STDIO instead of traditional Web APIs, providing a secure and efficient way for native AI clients to interact with code.

In this guide we build a lightweight Spring Boot application that lets AI directly invoke business logic to query CRM user information.

Step 1: Create the Spring Boot project

Generate a project at start.spring.io with the dependency Spring AI MCP Server (STDIO transport) or Spring AI MCP Server WebMVC (SSE transport) and download the zip.

<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-mcp-server</artifactId>
</dependency>

Step 2: Understand the project structure

Customer.java : data model for customer records.

CustomerService.java : service class where AI‑exposed methods are defined.

TestMcpServerApplication.java : main application entry point.

application.properties : server configuration.

Step 3: Define the data model

package com.example.testmcpserver;

public record Customer(String id, String name, String email, String phone, String company) {}

Step 4: Implement MCP tools

In CustomerService.java annotate methods with @Tool to expose them to the MCP framework.

@Service
public class CustomerService {
    private static final Logger log = LoggerFactory.getLogger(CustomerService.class);
    private List<Customer> customers = new ArrayList<>();

    @Tool(name = "get_customers", description = "Get a list of customers from the CRM system")
    public List<Customer> getCustomers() {
        return customers;
    }

    @Tool(name = "get_customer_by_name", description = "Get a single customer from the CRM system by name")
    public Customer getCustomerByName(String name) {
        return customers.stream()
                .filter(customer -> customer.name().equals(name))
                .findFirst()
                .orElse(null);
    }

    @PostConstruct
    public void init() {
        customers.addAll(List.of(
            new Customer("1","张三","[email protected]","13800138001","阿里巴巴"),
            new Customer("2","李四","[email protected]","13800138002","腾讯科技"),
            new Customer("3","王五","[email protected]","13800138003","字节跳动"),
            new Customer("4","赵六","[email protected]","13800138004","华为技术"),
            new Customer("5","钱七","[email protected]","13800138005","百度在线")
        ));
    }
}

Step 5: Register the tools

In TestMcpServerApplication.java create a bean that registers all @Tool methods.

@SpringBootApplication
public class TestMcpServerApplication {
    public static void main(String[] args) {
        SpringApplication.run(TestMcpServerApplication.class, args);
    }

    @Bean
    public List<ToolCallback> crmTools(CustomerService customerService) {
        return List.of(ToolCallbacks.from(customerService));
    }
}

Running the MCP client

Build the executable JAR with mvn clean package, locate it in the target/ directory, and run it.

Option 1: Connect via Cursor IDE

Configure a new MCP server in Cursor with a JSON like:

{
  "name":"crm-demo-mcp",
  "command":"java",
  "args":["-jar","path/to/crm-mcp-server-0.0.1-SNAPSHOT.jar"]
}

Then issue commands such as “Get all CRM customer information” or “Find customer named ‘张三’”.

Option 2: Use SSE transport

Add the WebMVC dependency:

<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-mcp-server-webmvc</artifactId>
</dependency>

Running the Spring Boot application will expose an /sse MCP endpoint. Configure the client with:

{
  "mcpServers": {
    "crm-demo-mcp": {
      "url": "http://localhost:8080/sse"
    }
  }
}

Conclusion

The Spring Boot MCP service enables AI to safely and efficiently access application data, providing a foundation for building smarter, AI‑driven applications.