快速开始

项目设置

创建一个项目并初始化 npm。你只需要做一次。
Terminal window
```
mkdir my_project
cd my_project
npm init -y
```
安装 Agents SDK 和 Zod。SDK 使用 Zod v4 来定义工具 schema 和 structured outputs。
Terminal window
```
npm install @openai/agents zod
```
设置 OpenAI API key。如果你还没有，请按照这些说明创建一个 OpenAI API key。
Terminal window
```
export OPENAI_API_KEY=sk-...
```
或者，你也可以调用 setDefaultOpenAIKey('<api key>') 以编程方式设置 key，并使用 setTracingExportApiKey('<api key>') 进行追踪。更多细节请参阅SDK 配置指南。

创建首个智能体

智能体通过 instructions 和名称来定义。

import { Agent } from '@openai/agents';

const agent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

首次运行智能体

你可以使用 run 方法来运行智能体。触发一次运行时，需要同时传入你要启动的智能体和要传入的输入。

这会返回一个运行结果，其中包含最终输出以及本次运行期间执行的所有操作。

import { Agent, run } from '@openai/agents';

const agent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

const result = await run(agent, 'When did sharks first appear?');

console.log(result.finalOutput);

在第二轮中，你可以将 result.history 传回 run()，绑定一个会话，或通过 conversationId / previousResponseId 复用 OpenAI 服务器管理的状态。运行智能体指南对比了这些方法。

为智能体添加工具

你可以为智能体提供工具，用于查询信息或执行操作。

import { Agent, tool } from '@openai/agents';
import { z } from 'zod';

const historyFunFact = tool({
  // The name of the tool will be used by the agent to tell what tool to use.
  name: 'history_fun_fact',
  // The description is used to describe when to use the tool by telling it what it does.
  description: 'Give a fun fact about a historical event',
  // This tool takes no parameters, so we provide an empty Zod object.
  parameters: z.object({}),
  execute: async () => {
    // The output will be returned back to the agent to use.
    return 'Sharks are older than trees.';
  },
});

const agent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
  // Add the tool to the agent.
  tools: [historyFunFact],
});

添加更多智能体

你也可以类似地定义更多智能体，将问题拆分为更小的部分，让智能体更专注于当前任务。此外，通过在智能体上定义模型，你还可以针对不同问题使用不同模型。

import { Agent } from '@openai/agents';

const historyTutorAgent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

const mathTutorAgent = new Agent({
  name: 'Math Tutor',
  instructions:
    'You provide help with math problems. Explain your reasoning at each step and include examples',
});

定义交接

为了在多个智能体之间进行编排，你可以为智能体定义 handoffs。这样智能体就能将对话传递给下一个智能体。这一过程会在运行期间自动发生。

import { Agent } from '@openai/agents';

const historyTutorAgent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

const mathTutorAgent = new Agent({
  name: 'Math Tutor',
  instructions:
    'You provide help with math problems. Explain your reasoning at each step and include examples',
});

// Use Agent.create() to keep handoff output types aligned.
const triageAgent = Agent.create({
  name: 'Triage Agent',
  instructions:
    "You determine which agent to use based on the user's homework question",
  handoffs: [historyTutorAgent, mathTutorAgent],
});

运行结束后，你可以通过查看结果中的 lastAgent 属性，确认是哪个智能体生成了最终响应。

运行智能体编排

runner 会处理单个智能体执行、交接以及工具调用。

import { Agent, run } from '@openai/agents';

const historyTutorAgent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

const mathTutorAgent = new Agent({
  name: 'Math Tutor',
  instructions:
    'You provide help with math problems. Explain your reasoning at each step and include examples',
});

const triageAgent = Agent.create({
  name: 'Triage Agent',
  instructions:
    "You determine which agent to use based on the user's homework question",
  handoffs: [historyTutorAgent, mathTutorAgent],
});

async function main() {
  const result = await run(triageAgent, 'What is the capital of France?');
  console.log(result.finalOutput);
}

main().catch((err) => console.error(err));

整体集成

现在把这些整合成一个完整示例。将其放入你的 index.js 文件并运行。如果你的应用已经配置为 TypeScript，也可以改用 index.ts。

import { Agent, run } from '@openai/agents';

const historyTutorAgent = new Agent({
  name: 'History Tutor',
  instructions:
    'You provide assistance with historical queries. Explain important events and context clearly.',
});

const mathTutorAgent = new Agent({
  name: 'Math Tutor',
  instructions:
    'You provide help with math problems. Explain your reasoning at each step and include examples',
});

const triageAgent = Agent.create({
  name: 'Triage Agent',
  instructions:
    "You determine which agent to use based on the user's homework question",
  handoffs: [historyTutorAgent, mathTutorAgent],
});

async function main() {
  const result = await run(triageAgent, 'What is the capital of France?');
  console.log(result.finalOutput);
}

main().catch((err) => console.error(err));

查看追踪

Agents SDK 会自动为你生成追踪。这让你可以查看智能体如何运行、调用了哪些工具，以及交接到了哪个智能体。

要查看你的智能体运行期间发生了什么，请前往 OpenAI Dashboard 中的 Trace 查看器。

后续步骤

了解如何构建更复杂的智能体流程：

了解如何配置智能体。
了解运行智能体和会话。
了解工具、护栏和模型。