How to Seamlessly Integrate MCP Protocol with Spring AI for Powerful LLM Tool Calls

Why MCP Is Needed

Before MCP, each tool had a different integration method, causing high client development cost (e.g., getWeather as HTTP, getLocation as HSF) and server-side inconsistencies where agents needed multiple interfaces.

MCP standardizes client‑server interaction in AI development, simplifying tool calls.

Using MCP with Spring‑AI

Framework Overview

Spring‑AI provides a unified way to call large models and integrate MCP tools.

Dependency

<dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-mcp-client</artifactId>
  <version>1.0.0-M7</version>
</dependency>
<dependency>
  <groupId>com.alibaba.cloud.ai</groupId>
  <artifactId>spring-ai-alibaba-starter</artifactId>
  <version>1.0.0-M6.1</version>
</dependency>

Configure OpenAI Model

spring.ai.openai.base-url=https://idealab.alibaba-inc.com
spring.ai.openai.api-key=YOUR_KEY
spring.ai.openai.chat.completions-path=api/openai/v1/chat/completions
spring.ai.openai.chat.options.model=gpt-4o-0513-global
spring.ai.openai.chat.options.temperature=0.1

Register Chat Model

@Configuration
public class ChatClientConfig {
    @Autowired private ToolCallbackProvider tools;
    @Autowired private OpenAiChatModel chatModel;
    @Bean
    public CommandLineRunner predefinedQuestions(ConfigurableApplicationContext context) {
        return args -> {
            var chatClient = ChatClient.builder(chatModel).build();
            String userInput = "帮我将这个网页内容进行抓取 https://www.shuaijiao.cn/news/view/68320.html";
            System.out.println("
>>> QUESTION: " + userInput);
            System.out.println("
>>> ASSISTANT: " + chatClient.prompt().user(userInput).call().content());
            context.close();
        };
    }
}

Inject MCP Tool into ChatModel

spring.ai.mcp.client.name=ai-demo
spring.ai.mcp.client.type=SYNC
spring.ai.mcp.client.toolcallback.enabled=true
spring.ai.mcp.client.request-timeout=30000
spring.ai.mcp.client.enabled=true
spring.ai.mcp.client.sse.connections.server1.url=https://mcp-09724909-442f-4b85.api-inference.modelscope.cn

After injection, the chat model can automatically select and call MCP tools based on user intent.

Native MCP SDK Usage

Dependency

<dependency>
  <groupId>io.modelcontextprotocol.sdk</groupId>
  <artifactId>mcp</artifactId>
  <version>0.9.0</version>
</dependency>

Initialize Client

@PostConstruct
public void init() {
    McpClientTransport transport = new HttpClientSseClientTransport("https://mcp-09724909-442f-4b85.api-inference.modelscope.cn");
    mcpClient = McpClient.sync(transport)
        .requestTimeout(Duration.ofSeconds(20))
        .capabilities(ClientCapabilities.builder().roots(true).sampling().build())
        .build();
    mcpClient.initialize();
}

@PreDestroy
public void destroy() {
    if (mcpClient != null) {
        mcpClient.closeGracefully();
    }
}

List Tools

@Bean
public String getToolList() {
    ListToolsResult toolsResult = mcpClient.listTools();
    for (Tool tool : toolsResult.tools()) {
        System.out.println(tool.name());
        System.out.println(tool.description());
        System.out.println(tool.inputSchema());
    }
    return null;
}

Call a Tool (e.g., fetch webpage)

Map<String,Object> params = new HashMap<>();
params.put("url","https://www.shuaijiao.cn/news/view/68320.html");
CallToolResult result = mcpClient.callTool(new CallToolRequest("fetch", params));
System.out.println(extractTextContent(result));

private String extractTextContent(CallToolResult toolResult) {
    StringBuilder sb = new StringBuilder();
    toolResult.content().forEach(c -> {
        if (c instanceof TextContent tc) sb.append(tc.text());
    });
    return sb.toString();
}

Pitfalls & Solutions

1.0.0‑M6 SSE error due to duplicate tool names – exclude SseHttpClientTransportAutoConfiguration or upgrade to M7.

SSE URL rewriting issue – upgrade to M7 and add io.modelcontextprotocol.sdk 0.9.

Connection timeout – ensure MCP service URLs are publicly reachable.

Conclusion

The Spring‑AI approach offers quick integration with minimal code, ideal for prototypes, while the native MCP SDK gives fine‑grained control suitable for platform‑level development. Choose the method that matches your project’s scalability and flexibility requirements.