spring-ai-starter-mcp-client小试牛刀

本文主要研究下如何使用spring-ai-starter-mcp-client

步骤

pom.xml

xml 复制代码
		<dependency>
			<groupId>org.springframework.ai</groupId>
			<artifactId>spring-ai-starter-mcp-client</artifactId>
		</dependency>

		<dependency>
            <groupId>org.springframework.ai</groupId>
            <artifactId>spring-ai-ollama-spring-boot-starter</artifactId>
        </dependency>

application.yml

yaml 复制代码
spring:
  application:
    name: mcp-client-example
  main:
    web-application-type: none
  ai:
    ollama:
      base-url: http://localhost:11434
      chat:
        model: qwen2.5:latest
    mcp:
      client:
        enabled: true
        name: my-mcp-client
        version: 1.0.0
        request-timeout: 30s
        type: SYNC  # or ASYNC for reactive applications
        sse:
          connections:
            server1:
              url: http://localhost:8080
#            server2:
#              url: http://otherserver:8081
        toolcallback:
          enabled: 'true'
logging:
  level:
    io:
      modelcontextprotocol:
        client: WARN
        spec: WARN

示例

less 复制代码
	@Bean
	public CommandLineRunner predefinedQuestions(ChatClient.Builder chatClientBuilder,
			ConfigurableApplicationContext context, SyncMcpToolCallbackProvider toolCallbackProvider) {

		return args -> {

			var chatClient = chatClientBuilder
					.defaultTools(toolCallbackProvider)
					.defaultAdvisors(new MessageChatMemoryAdvisor(new InMemoryChatMemory()))
					.build();

			System.out.println("\n>>> ASSISTANT: " + chatClient.prompt("帮我调用tool查一下xxx").call().content());
			System.out.println("\n>>> ASSISTANT: " + chatClient.prompt("取出create_time的最大值").call().content());

			context.close();
		};
	}

源码

McpClientAutoConfiguration

org/springframework/ai/mcp/client/autoconfigure/McpClientAutoConfiguration.java

less 复制代码
@AutoConfiguration(after = { StdioTransportAutoConfiguration.class, SseHttpClientTransportAutoConfiguration.class,
		SseWebFluxTransportAutoConfiguration.class })
@ConditionalOnClass({ McpSchema.class })
@EnableConfigurationProperties(McpClientCommonProperties.class)
@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX, name = "enabled", havingValue = "true",
		matchIfMissing = true)
public class McpClientAutoConfiguration {

	/**
	 * Create a dynamic client name based on the client name and the name of the server
	 * connection.
	 * @param clientName the client name as defined by the configuration
	 * @param serverConnectionName the name of the server connection being used by the
	 * client
	 * @return the connected client name
	 */
	private String connectedClientName(String clientName, String serverConnectionName) {
		return clientName + " - " + serverConnectionName;
	}

	/**
	 * Creates a list of {@link McpSyncClient} instances based on the available
	 * transports.
	 *
	 * <p>
	 * Each client is configured with:
	 * <ul>
	 * <li>Client information (name and version) from common properties
	 * <li>Request timeout settings
	 * <li>Custom configurations through {@link McpSyncClientConfigurer}
	 * </ul>
	 *
	 * <p>
	 * If initialization is enabled in properties, the clients are automatically
	 * initialized.
	 * @param mcpSyncClientConfigurer the configurer for customizing client creation
	 * @param commonProperties common MCP client properties
	 * @param transportsProvider provider of named MCP transports
	 * @return list of configured MCP sync clients
	 */
	@Bean
	@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX, name = "type", havingValue = "SYNC",
			matchIfMissing = true)
	public List<McpSyncClient> mcpSyncClients(McpSyncClientConfigurer mcpSyncClientConfigurer,
			McpClientCommonProperties commonProperties,
			ObjectProvider<List<NamedClientMcpTransport>> transportsProvider) {

		List<McpSyncClient> mcpSyncClients = new ArrayList<>();

		List<NamedClientMcpTransport> namedTransports = transportsProvider.stream().flatMap(List::stream).toList();

		if (!CollectionUtils.isEmpty(namedTransports)) {
			for (NamedClientMcpTransport namedTransport : namedTransports) {

				McpSchema.Implementation clientInfo = new McpSchema.Implementation(
						this.connectedClientName(commonProperties.getName(), namedTransport.name()),
						commonProperties.getVersion());

				McpClient.SyncSpec syncSpec = McpClient.sync(namedTransport.transport())
					.clientInfo(clientInfo)
					.requestTimeout(commonProperties.getRequestTimeout());

				syncSpec = mcpSyncClientConfigurer.configure(namedTransport.name(), syncSpec);

				var syncClient = syncSpec.build();

				if (commonProperties.isInitialized()) {
					syncClient.initialize();
				}

				mcpSyncClients.add(syncClient);
			}
		}

		return mcpSyncClients;
	}

	//......
}	

McpClientAutoConfiguration依赖McpClientCommonProperties,它根据McpClientCommonProperties和List<NamedClientMcpTransport>构建List<McpSyncClient>List<McpAsyncClient>

McpToolCallbackAutoConfiguration

org/springframework/ai/mcp/client/autoconfigure/McpToolCallbackAutoConfiguration.java

less 复制代码
@AutoConfiguration(after = { McpClientAutoConfiguration.class })
@EnableConfigurationProperties(McpClientCommonProperties.class)
@Conditional(McpToolCallbackAutoConfiguration.McpToolCallbackAutoconfigurationCondition.class)
public class McpToolCallbackAutoConfiguration {

	public static class McpToolCallbackAutoconfigurationCondition extends AllNestedConditions {

		public McpToolCallbackAutoconfigurationCondition() {
			super(ConfigurationPhase.PARSE_CONFIGURATION);
		}

		@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX, name = "enabled", havingValue = "true",
				matchIfMissing = true)
		static class McpAutoConfigEnabled {

		}

		@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX + ".toolcallback", name = "enabled",
				havingValue = "true", matchIfMissing = false)
		static class ToolCallbackProviderEnabled {

		}

	}

	/**
	 * Creates tool callbacks for all configured MCP clients.
	 *
	 * <p>
	 * These callbacks enable integration with Spring AI's tool execution framework,
	 * allowing MCP tools to be used as part of AI interactions.
	 * @param syncMcpClients provider of MCP sync clients
	 * @return list of tool callbacks for MCP integration
	 */
	@Bean
	@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX, name = "type", havingValue = "SYNC",
			matchIfMissing = true)
	public ToolCallbackProvider mcpToolCallbacks(ObjectProvider<List<McpSyncClient>> syncMcpClients) {
		List<McpSyncClient> mcpClients = syncMcpClients.stream().flatMap(List::stream).toList();
		return new SyncMcpToolCallbackProvider(mcpClients);
	}

	@Bean
	@ConditionalOnProperty(prefix = McpClientCommonProperties.CONFIG_PREFIX, name = "type", havingValue = "ASYNC")
	public ToolCallbackProvider mcpAsyncToolCallbacks(ObjectProvider<List<McpAsyncClient>> mcpClientsProvider) {
		List<McpAsyncClient> mcpClients = mcpClientsProvider.stream().flatMap(List::stream).toList();
		return new AsyncMcpToolCallbackProvider(mcpClients);
	}

}

McpToolCallbackAutoConfiguration在spring.ai.mcp.client.enabledspring.ai.mcp.client.toolcallback.enabled都为true时才生效,它会根据List<McpSyncClient>去创建SyncMcpToolCallbackProvider或AsyncMcpToolCallbackProvider

小结

spring ai提供了spring-ai-starter-mcp-client用于快速集成mcpClient到chatClient中;McpClientAutoConfiguration依赖McpClientCommonProperties,它根据McpClientCommonProperties和List<NamedClientMcpTransport>构建List<McpSyncClient>List<McpAsyncClient>;McpToolCallbackAutoConfiguration在spring.ai.mcp.client.enabledspring.ai.mcp.client.toolcallback.enabled都为true时才生效,它会根据List<McpSyncClient>去创建SyncMcpToolCallbackProvider或AsyncMcpToolCallbackProvider

doc

相关推荐
你是一个铁憨憨2 小时前
从 GIS 到 Spatial Agent:MCP 如何重新定义 GIS 的 AI 入口
arcgis·ai·agent·mcp·spatial
Java牛马5 小时前
AI Agent 技术栈梳理(Skill / 蒸馏 / MCP / Harness)
人工智能·ai agent·蒸馏·skill·mcp·harness
alwaysmavs7 小时前
实测 OpenConnector:给 Agent 接 SaaS,别再把 Token 塞进环境变量
mcp
ryan_9967 小时前
一次讲清 A2A 协议与 MCP 边界:从 Agent Card 到 Task 生命周期
agent·mcp·a2a·json-rpc·agent通信
xrlfreedom10 小时前
大厂 MCP 面试实录:本地 Server 远程化改造中的技术选型与落地实践
docker·json schema·mcp
五度易链-区域产业数字化管理平台1 天前
WorkBuddy 实战:将带 MD5 签名的第三方 API 封装为 MCP 服务(企业模糊搜索接口案例)
大数据·人工智能·mcp
栩栩云生1 天前
别再硬记 AWS 和 k8s 命令了!一行命令把十几个云平台的 CLI 全接进 AI
kubernetes·agent·mcp
枝枝在Coding1 天前
我把 Codex 接进 SSH 后,才发现 Xterminal MCP 最难的是敢不敢让 Agent 动手
ssh·mcp
可乐ea2 天前
Anthropic 的 CI/CD 值班智能体:Claude Tag 当一线响应者的架构拆解与踩坑复盘
ci/cd·架构·claude·devops·ai智能体·mcp
xrlfreedom2 天前
大厂 MCP 面试实录:内部 REST API 封装为可审计 MCP Tools 的落地方案
docker·mcp·提示注入防护