Puppeteer 能让程序自己打开浏览器，像人一样点鼠标、输文字、抓页面内容，干活根本不用你盯着。

GitHub 上有 95k Star，是浏览器自动化领域绝对的标杆项目。

GitHub：https://github.com/puppeteer/puppeteer…

以前要每天手动登录某个网站抓数据、填表、测某个网页功能，重复点来点去特别烦，用录屏宏工具还很脆弱，网页一改版就失效。换

成 Puppeteer 之后，写几行代码就能控制 Chrome 或 Firefox，模拟点击搜索框、输入关键词、点结果、把页面文字抓下来，整套流程默认无界面运行，可以丢给后台自己跑完，人不用管。

更有意思的是它还有 MCP：

Puppeteer 做的 chrome-devtools-mcp，让 AI agent 能直接通过 MCP 协议操控浏览器、调试网页。

也就是说现在不止程序员在用它写自动化脚本，AI agent 也能拿它当手去点网页。

puppeteer/puppeteer

Source: https://github.com/puppeteer/puppeteer

Puppeteer

Puppeteer is a JavaScript library which provides a high-level API to control Chrome or Firefox over the DevTools Protocol or WebDriver BiDi. Puppeteer runs in the headless (no visible UI) by default

Get started | API | FAQ | Contributing | Troubleshooting

Installation

npm i puppeteer # Downloads compatible Chrome during installation.
npm i puppeteer-core # Alternatively, install as a library, without downloading Chrome.

:::note

Modern package managers (including npm (see the RFC), pnpm, Yarn, Bun, and Deno) block dependency install scripts by default. If the install script is blocked, Puppeteer will not download the browser during installation, leading to runtime errors.

You can manually download the required browsers after installation by running:

npx puppeteer browsers install

Alternatively, you can configure your package manager to allow the install script to run (for example, with npm, by adding "puppeteer" to "allowScripts" in your package.json).

:::

MCP

Install chrome-devtools-mcp, a Puppeteer-based MCP server for browser automation and debugging.

Puppeteer also supports the experimental WebMCP API.

Example

import puppeteer from 'puppeteer';
// Or import puppeteer from 'puppeteer-core';

// Launch the browser and open a new blank page.
const browser = await puppeteer.launch();
const page = await browser.newPage();

// Navigate the page to a URL.
await page.goto('https://developer.chrome.com/');

// Set the screen size.
await page.setViewport({width: 1080, height: 1024});

// Open the search menu using the keyboard.
await page.keyboard.press('/');

// Type into search box using accessible input name.
await page.locator('::-p-aria(Search)').fill('automate beyond recorder');

// Wait and click on first result.
await page.locator('.devsite-result-item-link').click();

// Locate the full title with a unique string.
const textSelector = await page
  .locator('::-p-text(Customize and automate)')
  .waitHandle();
const fullTitle = await textSelector?.evaluate(el => el.textContent);

// Print the full title.
console.log('The title of this blog post is "%s".', fullTitle);

await browser.close();