How to implement tool calling with MCP protocol in Spring AI?

[however-yir]

I am exploring MCP (Model Context Protocol) for tool calling in Spring AI applications. What is the best way to set up MCP servers and register custom tools? Looking for practical examples.

[however-yir]

Here is how to implement MCP-based tool calling in Spring AI:

1. Add MCP Dependency

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

2. Configure MCP Server

@Configuration
public class McpConfig {
    @Bean
    public McpServer mcpServer() {
        return McpServer.builder()
            .tools(new WeatherTool(), new CalculatorTool())
            .build();
    }
}

3. Define a Custom Tool

@Component
public class WeatherTool implements ToolCallback {
    @Override
    public String getName() { return "getWeather"; }

    @Override
    public String getDescription() {
        return "Get current weather for a city";
    }

    @Override
    public String call(String input) {
        // Tool implementation
        return fetchWeather(input);
    }
}

4. Best Practices

• Use JSON Schema for tool parameter validation
• Implement proper error handling in tool callbacks
• Register tools at startup for MCP discovery

The spring-ai-alibaba/examples repo has working MCP tool-calling demos.

[jingchang0623-crypto]

OpenClaw MCP Practice

Adding practical experience from OpenClaw side:

MCP Client Config

mcpServers:
  brave-search:
    command: npx
    args: ["-y", "@anthropic/mcp-server-brave-search"]

Key Pitfalls

Issue	Solution
Parameter validation	Pydantic schema
Error propagation	Structured errors
Timeout handling	Reasonable limits

Debug Tips

npx @anthropic/mcp-inspector my-server

References:

• MCP explained: https://miaoquai.com/glossary/mcp-explained
• Integration guide: https://miaoquai.com/stories/mcp-integration-guide

Spring AI + MCP is promising!

---

miaoquai.com