spring-ai-mcp-server-patterns

This skill provides production-ready patterns for building Model Context Protocol (MCP) servers with Spring Boot and Spring AI. It solves the problem of exposing Java application capabilities as AI-callable tools, resources, and prompt templates by documenting the canonical annotations, transport configuration, and architectural patterns for AI function/tool calling using Spring's official AI framework.

The skill covers core annotations (@EnableMcpServer, @Tool, @ToolParam, @PromptTemplate, @PromptParam), the three transport types (stdio for local/Claude Desktop, http, and sse for streaming), and Maven/Gradle dependency setup. Step-by-step instructions show project setup with API keys pulled from environment variables, defining @Tool methods inside @Component beans, and building reusable prompt templates. A rich set of reference files goes further into advanced patterns — dynamic runtime tool registration, multi-model selection across OpenAI and Anthropic, caching with Spring's cache abstraction, and security. Notably, the security-oriented examples demonstrate a SecureToolExecutor that authenticates and authorizes callers via Spring Security before running a tool, Bean Validation on tool inputs, and defensive argument checks, alongside examples.md, migration-guide.md, api-reference.md, and testing-guide.md.

It targets Java and Spring developers building MCP servers to extend AI assistants with custom tools, implementing function calling, or integrating MCP clients. The advanced dynamic-registration examples include dispatching to Groovy scripts, Spring beans, or HTTP endpoints, which are legitimate extensibility patterns but warrant care when the registration inputs are externally controlled; the reference pairs them with authentication, authorization, and input-validation guidance.