目录
[1.1 什么是MCP协议?](#1.1 什么是MCP协议?)
[1.2 为什么需要MCP协议?](#1.2 为什么需要MCP协议?)
[1.3 MCP架构设计](#1.3 MCP架构设计)
[1.4 案例实战:创建MCP Server项目](#1.4 案例实战:创建MCP Server项目)
[1.5 关键配置说明](#1.5 关键配置说明)
[1.6 注意事项](#1.6 注意事项)
二、掌握@McpTool注解开发AI工具,实现计算器和天气查询工具的标准化注册
[2.1 什么是@McpTool注解?](#2.1 什么是@McpTool注解?)
[2.2 工具开发规范](#2.2 工具开发规范)
[2.3 案例实战一:计算器工具开发](#2.3 案例实战一:计算器工具开发)
[2.4 案例实战二:天气查询工具开发](#2.4 案例实战二:天气查询工具开发)
[2.5 工具自动注册流程](#2.5 工具自动注册流程)
[2.6 注意事项](#2.6 注意事项)
[三、MCP Client客户端项目创建和MCP配置实战](#三、MCP Client客户端项目创建和MCP配置实战)
[3.1 什么是MCP Client?](#3.1 什么是MCP Client?)
[3.2 整体调用链路](#3.2 整体调用链路)
[3.3 案例实战:创建MCP Client项目](#3.3 案例实战:创建MCP Client项目)
[四、使用Spring AI Alibaba Agent Framework构建ReactAgent智能体,整合MCP工具实现智能问答](#四、使用Spring AI Alibaba Agent Framework构建ReactAgent智能体,整合MCP工具实现智能问答)
[4.1 什么是ReactAgent?](#4.1 什么是ReactAgent?)
[4.2 案例实战:构建ReactAgent服务](#4.2 案例实战:构建ReactAgent服务)
[5.1 项目启动测试步骤](#5.1 项目启动测试步骤)
[5.2 ReactAgent智能体整合MCP测试验证](#5.2 ReactAgent智能体整合MCP测试验证)
[5.3 注意事项](#5.3 注意事项)
一、AI大模型MCP协议介绍和服务端项目搭建
1.1 什么是MCP协议?
-
核心定义:**MCP(Model Context Protocol)**是由Anthropic公司推出的AI工具标准化协议
-
简单说,MCP就像一个"AI工具的USB接口",让不同的AI模型可以用统一的方式调用各种工具
-
打个比方:
-
以前每个AI平台有自己的工具接口,就像手机充电线,安卓用Type-C,苹果用Lightning
-
MCP协议就像统一的充电标准,所有AI都用同一种方式调用工具

-
1.2 为什么需要MCP协议?
-
需求背景
-
协议不统一:不同AI平台的工具接口各异,适配成本高
-
工具描述混乱:JSON Schema描述不规范,AI理解困难
-
调试困难:工具调用出错时,难以定位问题
-
-
MCP协议的核心价值:
-
标准化协议:统一AI工具接口规范
-
自动发现:服务端自动注册工具,客户端自动发现
-
类型安全:强类型参数定义,减少运行时错误
-
跨平台:一次开发,多平台复用
-
1.3 MCP架构设计

1.4 案例实战:创建MCP Server项目
- 项目结构
ai/
├── pom.xml # 父工程
├── mcp-server/ # MCP服务端
│ ├── pom.xml
│ └── src/main/
│ ├── java/
│ └── resources/
└── mcp-client/ # MCP客户端
├── pom.xml
└── src/main/
- 父工程pom.xml核心配置
XML
<parent>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-parent</artifactId>
<version>3.5.3</version>
<relativePath/>
</parent>
<properties>
<maven.compiler.source>21</maven.compiler.source>
<maven.compiler.target>21</maven.compiler.target>
<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
<spring-ai.version>1.1.2</spring-ai.version>
<spring-ai-alibaba.version>1.1.2.0</spring-ai-alibaba.version>
</properties>
<repositories>
<repository>
<id>aliyun-maven</id>
<name>aliyun-maven</name>
<url>https://maven.aliyun.com/repository/central</url>
<releases>
<enabled>true</enabled>
</releases>
<snapshots>
<enabled>false</enabled>
</snapshots>
</repository>
<repository>
<id>aliyun</id>
<name>aliyun</name>
<url>https://maven.aliyun.com/repository/public</url>
<releases>
<enabled>true</enabled>
</releases>
<snapshots>
<enabled>false</enabled>
</snapshots>
</repository>
<repository>
<id>aliyun-spring</id>
<name>aliyun-spring</name>
<url>https://maven.aliyun.com/repository/spring</url>
<releases>
<enabled>true</enabled>
</releases>
<snapshots>
<enabled>false</enabled>
</snapshots>
</repository>
</repositories>
<dependencyManagement>
<dependencies>
<!-- Spring AI BOM -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-bom</artifactId>
<version>${spring-ai.version}</version>
<type>pom</type>
<scope>import</scope>
</dependency>
<!-- Spring AI Alibaba Extensions BOM(优先,覆盖 dashscope 版本为 1.1.2.1) -->
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-extensions-bom</artifactId>
<version>1.1.2.1</version>
<type>pom</type>
<scope>import</scope>
</dependency>
<!-- Spring AI Alibaba BOM -->
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-bom</artifactId>
<version>${spring-ai-alibaba.version}</version>
<type>pom</type>
<scope>import</scope>
</dependency>
</dependencies>
</dependencyManagement>
<!-- 配置插件仓库 -->
<pluginRepositories>
<pluginRepository>
<id>aliyun-plugin</id>
<name>aliyun-plugin</name>
<url>https://maven.aliyun.com/repository/public</url>
<releases>
<enabled>true</enabled>
</releases>
<snapshots>
<enabled>false</enabled>
</snapshots>
</pluginRepository>
</pluginRepositories>
- mcp-server模块pom.xml
XML
<parent>
<groupId>net.xdclass</groupId>
<artifactId>mcp-demo</artifactId>
<version>1.0-SNAPSHOT</version>
</parent>
<artifactId>mcp-server</artifactId>
<name>MCP Server</name>
<description>MCP Server - 提供工具服务的独立部署模块</description>
<dependencies>
<!-- Spring Boot Starter Web -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-web</artifactId>
</dependency>
<!-- Spring AI MCP Server WebMvc Starter (包含 @Tool 注解支持) -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-starter-mcp-server-webmvc</artifactId>
</dependency>
<!-- Spring AI Alibaba DashScope Starter -->
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-starter-dashscope</artifactId>
</dependency>
<!-- Lombok -->
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<optional>true</optional>
</dependency>
</dependencies>
<build>
<plugins>
<plugin>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-maven-plugin</artifactId>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<configuration>
<annotationProcessorPaths>
<path>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
</path>
</annotationProcessorPaths>
</configuration>
</plugin>
</plugins>
</build>
- application.yml配置
XML
server:
port: ${SERVER_PORT:8090}
spring:
application:
name: xd-mcp-server
mvc:
async:
request-timeout: -1 # SSE长连接保持,不超时
ai:
dashscope:
api-key: ${DASHSCOPE_API_KEY:sk-6xxxxxxxxxxxxxxxxxxxxff873fc}
chat:
options:
model: ${AI_MODEL:qwen-max}
mcp:
server:
enabled: true
name: xdclass-mcp-server
version: 1.0.0
type: SYNC
# 区别说明:
# STREAMABLE - 有状态,服务端仍维护session
# STATELESS - 真正无状态,请求间不维护任何会话,专为云原生/Serverless设计
protocol: STATELESS # 无状态模式,完美适配FC/Serverless环境 https://www.spring-doc.cn/spring-ai/1.1.0/api_mcp_mcp-stateless-server-boot-starter-docs.html
stateless:
mcp-endpoint: /mcp # 无状态HTTP端点
annotation-scanner:
enabled: true
- 启动类
java
@SpringBootApplication
public class McpServerApplication {
public static void main(String[] args) {
SpringApplication.run(McpServerApplication.class, args);
}
}
1.5 关键配置说明
| 配置项 | 说明 |
|---|---|
mcp.server.enabled |
启用MCP服务端 |
mcp.server.type |
SYNC同步模式 / ASYNC异步模式 |
mcp.server.protocol |
STATELESS无状态 / STREAMABLE有状态 |
annotation-scanner.enabled |
启用@McpTool注解扫描 |
1.6 注意事项
-
端口隔离:MCP Server建议独立端口部署,与业务服务分离
-
无状态模式 :
STATELESS模式适合云原生/Serverless环境,无需维护会话 -
SSE超时 :设置
request-timeout: -1避免长连接被中断
二、掌握@McpTool注解开发AI工具,实现计算器和天气查询工具的标准化注册
2.1 什么是@McpTool注解?
-
核心定义:
@McpTool是Spring AI MCP提供的工具注册注解,用于将Java方法标记为MCP工具 -
简单说,就像给方法贴一个"AI可调用"的标签,AI就能自动发现并使用这个方法
-
相关注解:
-
@McpTool:标记工具方法,定义name和description
-
@McpToolParam:标记参数,定义参数描述
-
2.2 工具开发规范
-
命名规范:
-
工具名使用下划线命名法,如
calculator_add、weather_get_current -
描述要清晰说明工具用途,AI根据描述决定是否调用
-
-
参数设计:
-
每个参数都要有description描述
-
参数类型使用基本类型或简单对象
-
-
返回值:
-
支持基本类型、Map、自定义对象
-
返回值会被自动序列化为JSON
-
2.3 案例实战一:计算器工具开发
-
需求:提供加减乘除、幂运算、开方等数学计算能力
-
代码实现
java
import lombok.extern.slf4j.Slf4j;
import org.springaicommunity.mcp.annotation.McpTool;
import org.springaicommunity.mcp.annotation.McpToolParam;
import org.springframework.stereotype.Component;
/**
* 计算器工具
* 提供基本的数学运算功能
*/
@Component
@Slf4j
public class CalculatorTool {
/**
* 加法运算
*/
@McpTool(name = "calculator_add", description = "执行两个数字的加法运算")
public double add(
@McpToolParam(description = "第一个数字") double a,
@McpToolParam(description = "第二个数字") double b) {
log.info("执行加法运算:{} + {}", a, b);
return a + b;
}
/**
* 减法运算
*/
@McpTool(name = "calculator_subtract", description = "执行两个数字的减法运算")
public double subtract(
@McpToolParam(description = "被减数") double a,
@McpToolParam(description = "减数") double b) {
log.info("执行减法运算:{} - {}", a, b);
return a - b;
}
/**
* 乘法运算
*/
@McpTool(name = "calculator_multiply", description = "执行两个数字的乘法运算")
public double multiply(
@McpToolParam(description = "第一个数字") double a,
@McpToolParam(description = "第二个数字") double b) {
log.info("执行乘法运算:{} * {}", a, b);
return a * b;
}
/**
* 除法运算
*/
@McpTool(name = "calculator_divide", description = "执行两个数字的除法运算")
public double divide(
@McpToolParam(description = "被除数") double a,
@McpToolParam(description = "除数") double b) {
if (b == 0) {
throw new IllegalArgumentException("除数不能为零");
}
log.info("执行除法运算:{} / {}", a, b);
return a / b;
}
/**
* 幂运算
*/
@McpTool(name = "calculator_power", description = "计算一个数的幂次方")
public double power(
@McpToolParam(description = "底数") double base,
@McpToolParam(description = "指数") double exponent) {
log.info("执行幂运算:{} ^ {}", base, exponent);
return Math.pow(base, exponent);
}
/**
* 平方根
*/
@McpTool(name = "calculator_sqrt", description = "计算一个数的平方根")
public double sqrt(
@McpToolParam(description = "要计算平方根的数字") double value) {
if (value < 0) {
throw new IllegalArgumentException("不能计算负数的平方根");
}
log.info("执行平方根运算:sqrt({})", value);
return Math.sqrt(value);
}
}
2.4 案例实战二:天气查询工具开发
-
需求:提供当前天气、天气预报、空气质量查询能力
-
代码实现
java
package net.xdclass.tool;
import lombok.extern.slf4j.Slf4j;
import org.springaicommunity.mcp.annotation.McpTool;
import org.springaicommunity.mcp.annotation.McpToolParam;
import org.springframework.stereotype.Component;
import java.time.LocalDate;
import java.time.LocalDateTime;
import java.util.*;
/**
* 天气工具
* 提供天气查询相关功能(模拟数据)
*/
@Slf4j
@Component
public class WeatherTool {
private final Random random = new Random();
/**
* 获取指定城市的天气信息
*/
@McpTool(name = "weather_get_current", description = "获取指定城市的当前天气信息")
public Map<String, Object> getCurrentWeather(
@McpToolParam(description = "城市名称,如:北京、上海、广州") String city) {
log.info("mcp-sever getCurrentWeather 被调用 v1-------");
Map<String, Object> weather = new HashMap<>();
weather.put("city", city);
weather.put("temperature", 15 + random.nextInt(20));
weather.put("humidity", 40 + random.nextInt(40));
weather.put("windSpeed", 5 + random.nextInt(15));
weather.put("condition", getRandomCondition());
weather.put("updateTime", LocalDateTime.now().toString());
log.info("获取当前天气信息:{}", weather);
return weather;
}
/**
* 获取未来几天的天气预报
*/
@McpTool(name = "weather_get_forecast", description = "获取指定城市未来几天的天气预报")
public Map<String, Object> getWeatherForecast(
@McpToolParam(description = "城市名称,如:北京、上海、广州") String city,
@McpToolParam(description = "预报天数,1-7天") int days) {
log.info("mcp-sever getWeatherForecast 被调用-------");
if (days < 1 || days > 7) {
throw new IllegalArgumentException("预报天数必须在1-7天之间");
}
Map<String, Object> forecast = new HashMap<>();
forecast.put("city", city);
forecast.put("days", days);
List<Map<String, Object>> dailyForecasts = new ArrayList<>();
for (int i = 0; i < days; i++) {
Map<String, Object> dayWeather = new HashMap<>();
dayWeather.put("day", i + 1);
dayWeather.put("date", LocalDate.now().plusDays(i).toString());
dayWeather.put("highTemp", 20 + random.nextInt(15));
dayWeather.put("lowTemp", 10 + random.nextInt(10));
dayWeather.put("condition", getRandomCondition());
dailyForecasts.add(dayWeather);
}
forecast.put("forecast", dailyForecasts);
log.info("获取天气预报:{}", forecast);
return forecast;
}
private String getRandomCondition() {
String[] conditions = {"晴", "多云", "阴", "小雨", "中雨", "雷阵雨"};
return conditions[random.nextInt(conditions.length)];
}
}
2.5 工具自动注册流程

2.6 注意事项
-
工具命名 :使用
模块_动作格式,如calculator_add、weather_get_current -
描述质量:描述越清晰,AI调用越准确
-
异常处理:工具内部应处理异常,返回友好错误信息
-
日志记录:每次工具调用都应记录日志,便于调试
三、MCP Client客户端项目创建和MCP配置实战
3.1 什么是MCP Client?
-
核心定义:MCP Client是连接AI模型和MCP Server的桥梁,负责工具发现和调用
-
简单说,MCP Client就像一个"翻译官":
-
向MCP Server询问有哪些工具可用
-
把AI的工具调用请求转发给Server
-
把工具执行结果返回给AI
-
3.2 整体调用链路

3.3 案例实战:创建MCP Client项目
- pom.xml依赖配置
XML
<dependencies>
<!-- Spring Boot Starter Web -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-web</artifactId>
</dependency>
<!-- Spring Boot Starter WebFlux (for HTTP Client) -->
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-webflux</artifactId>
</dependency>
<!-- Spring AI Alibaba DashScope Starter -->
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-starter-dashscope</artifactId>
</dependency>
<!-- Spring AI MCP Client Starter -->
<dependency>
<groupId>org.springframework.ai</groupId>
<artifactId>spring-ai-starter-mcp-client</artifactId>
</dependency>
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-agent-framework</artifactId>
<version>1.1.2.0</version>
</dependency>
<!-- Netty DNS Resolver for macOS -->
<dependency>
<groupId>io.netty</groupId>
<artifactId>netty-resolver-dns-native-macos</artifactId>
<classifier>osx-aarch_64</classifier>
<scope>runtime</scope>
</dependency>
<!-- Lombok -->
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<optional>true</optional>
</dependency>
</dependencies>
<build>
<plugins>
<plugin>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-maven-plugin</artifactId>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-compiler-plugin</artifactId>
<configuration>
<annotationProcessorPaths>
<path>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
</path>
</annotationProcessorPaths>
</configuration>
</plugin>
</plugins>
</build>
- application.yml配置
XML
server:
port: ${SERVER_PORT:8082}
spring:
application:
name: mcp-client
ai:
dashscope:
api-key: ${DASHSCOPE_API_KEY:sk-6fcxxxxxxxxxxxxxxxxxxd0f77}
chat:
options:
model: ${AI_MODEL:qwen-max}
retry:
max-attempts: 3
initial-interval: 2000
max-interval: 10000
multiplier: 2
mcp:
client:
enabled: true
type: SYNC
request-timeout: 60s
toolcallback:
enabled: true
# 使用Streamable HTTP连接(无状态,适合FC环境)
streamable-http:
connections:
mcp-server:
url: http://127.0.0.1:8090
endpoint: /mcp
- 启动类
java
@SpringBootApplication
public class McpClientApplication {
public static void main(String[] args) {
// 禁用JDK HTTP Client内部连接池,防止SSE重连时异常
System.setProperty("jdk.httpclient.connectionPoolSize", "0");
System.setProperty("jdk.httpclient.keepalive.timeout", "0");
SpringApplication.run(McpClientApplication.class, args);
}
}
- controller控制类
java
/**
* MCP Client REST API 控制器
* 通过 McpAgentService / ReactAgentService 驱动 AI 智能体,自动识别并调用 MCP Server 中的工具
*/
@Slf4j
@RestController
@RequestMapping("/api/mcp")
@RequiredArgsConstructor
public class ClientController {
private final ReactAgentService reactAgentService;
/**
* ReactAgent 单次调用接口(无会话记忆)
* 示例:GET /api/mcp/react?message=5加上10等于多少
*/
@GetMapping("/react")
public ResponseEntity<Map<String, Object>> reactChat(@RequestParam String message) {
log.info("收到 ReactAgent 请求 (GET): {}", message);
String response = reactAgentService.chat(message);
Map<String, Object> result = new HashMap<>();
result.put("message", message);
result.put("response", response);
return ResponseEntity.ok(result);
}
}
四、使用Spring AI Alibaba Agent Framework构建ReactAgent智能体,整合MCP工具实现智能问答
4.1 什么是ReactAgent?
-
核心定义:ReactAgent是基于ReAct(Reasoning + Acting)模式的智能体框架
-
简单说,ReactAgent会"边思考边行动":
-
Reasoning:分析用户问题,决定需要调用哪些工具
-
Acting:执行工具调用,获取结果
-
循环迭代 :根据结果继续思考,直到得出最终答案

-
4.2 案例实战:构建ReactAgent服务
- 添加依赖【已经有了】
XML
<!-- Spring AI Alibaba Agent Framework -->
<dependency>
<groupId>com.alibaba.cloud.ai</groupId>
<artifactId>spring-ai-alibaba-agent-framework</artifactId>
<version>1.1.2.0</version>
</dependency>
- ReactAgentService实现
java
package net.xdclass.service;
import com.alibaba.cloud.ai.graph.agent.ReactAgent;
import com.alibaba.cloud.ai.graph.checkpoint.savers.MemorySaver;
import com.alibaba.cloud.ai.graph.exception.GraphRunnerException;
import jakarta.annotation.PostConstruct;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
import org.springframework.ai.chat.messages.AssistantMessage;
import org.springframework.ai.chat.model.ChatModel;
import org.springframework.ai.tool.ToolCallback;
import org.springframework.ai.tool.ToolCallbackProvider;
import org.springframework.stereotype.Service;
/**
* ReactAgent 智能体服务
* 基于 Spring AI Alibaba Agent Framework,集成 MCP 工具,支持多轮会话记忆
*/
@Slf4j
@Service
@RequiredArgsConstructor
public class ReactAgentService {
private final ChatModel chatModel;
private final ToolCallbackProvider mcpToolCallbackProvider;
private ReactAgent reactAgent;
private static final String SYSTEM_PROMPT = """
你是一个集成了MCP工具的智能服务助手,拥有强大的工具调用能力。
你当前可以使用以下类型的工具:
- 数学计算:加法、减法、乘法、除法、幂运算、开方
- 天气查询:当前天气、多日天气预报、空气质量指数
工作原则:
1. 对于需要计算或查询的问题,优先调用相关工具获取准确结果
2. 对于多步骤复杂问题,请分步骤推理并依次调用工具
3. 工具调用结果需整合后再给出最终回答
4. 如果问题不需要工具,直接回答即可
""";
/**
* 初始化 ReactAgent(Spring 容器启动后执行)
* MCP 工具在此时已由 spring-ai-starter-mcp-client 自动加载完毕
*/
@PostConstruct
public void init() {
ToolCallback[] toolCallbacks = mcpToolCallbackProvider.getToolCallbacks();
log.info("===== ReactAgent 初始化开始 =====");
log.info("加载 {} 个 MCP 工具:", toolCallbacks.length);
for (ToolCallback tc : toolCallbacks) {
log.info(" - [{}] {}", tc.getToolDefinition().name(), tc.getToolDefinition().description());
}
reactAgent = ReactAgent.builder()
.name("mcp-react-agent")
.model(chatModel)
.tools(toolCallbacks)
.systemPrompt(SYSTEM_PROMPT)
.saver(new MemorySaver()) // 基于内存的会话状态存储
.build();
log.info("===== ReactAgent 初始化完成 =====");
}
/**
* 无会话单次调用
*
* @param userMessage 用户输入
* @return AI 最终响应文本
*/
public String chat(String userMessage) {
log.info("===== ReactAgent 单次调用开始 =====");
log.info("用户输入: {}", userMessage);
try {
AssistantMessage response = reactAgent.call(userMessage);
String content = response.getText();
log.info("AI 响应: {}", content);
log.info("===== ReactAgent 单次调用完成 =====");
return content;
} catch (GraphRunnerException e) {
log.error("ReactAgent 调用失败: {}", e.getMessage(), e);
throw new RuntimeException("ReactAgent 执行异常: " + e.getMessage(), e);
}
}
}
五、ReactAgent智能体整合MCP全流程测试实战
5.1 项目启动测试步骤
- 先启动Server服务端,再启动Client客户端,观察日志加载的tool工具是否齐全

5.2 ReactAgent智能体整合MCP测试验证
- 简单计算测试
- 复杂计算测试(多步骤推理)
GET http://localhost:8082/api/mcp/react?message=计算2的10次方再开平方
- 天气查询测试
GET http://localhost:8082/api/mcp/react?message=上海今天天气怎么样,未来3天天气如何
5.3 注意事项
-
初始化时机 :使用
@PostConstruct确保MCP工具加载完成后再初始化Agent -
系统提示词:清晰描述Agent能力和工作原则,提升调用准确率
-
异常处理 :捕获
GraphRunnerException,返回友好错误信息