Chrome MCP Server — AI Agent 浏览器自动化与 DevTools
安装 Chrome MCP 服务器，让 AI Agent 完全控制浏览器——导航页面、提取内容、执行 JavaScript、操作 DevTools、自动化 Web 工作流。多种实现适配不同场景。
安装 Chrome MCP Server
{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-puppeteer"]
}
}
}Chrome MCP Server — Extension-Based Browser MCP
mcp-chrome turns Chrome into an MCP server via an extension + bridge. Install mcp-chrome-bridge, load the extension, then connect to 127.0.0.1:12306/mcp.
Codex Chrome MCP Proxy v3
Sanitized public Chrome MCP bundle for Codex: persistent CDP proxy, real Chrome login-state control, background tabs, focus protection, multi-agent isolation, and cached chrome-devtools-mcp startup fallback.
Multi-Platform Social Media Growth Skill for Developer Communities
Production-tested Claude Code skill for automating authentic developer engagement across 10+ platforms (Twitter, Bluesky, Dev.to, HN, Reddit, GitHub, Product Hunt, Hashnode, Medium, Quora). Includes rotation queue, per-platform cooldowns, action diversity rules, Chrome MCP toolchain with JS injection patterns, anti-ban safety, and continuous optimization framework. Battle-tested across 30+ cycles.
Chrome DevTools MCP — Browser Debugging for AI Agents
Give your AI coding agent full access to Chrome DevTools for browser automation, debugging, and performance analysis. Works with Claude, Cursor, Copilot, and 15+ AI tools.
Puppeteer MCP — Headless Chrome Server for AI Agents
MCP server wrapping Google Puppeteer for headless Chrome automation. Navigate pages, screenshot, generate PDFs, scrape SPAs, and test web apps through AI agent commands. 2,000+ stars.
Chrome MCP 完整运维指南
Chrome MCP 完整运维知识库 + 一键安装：持久代理防弹窗/防抢焦点/多 Agent 隔离。三个脚本全文内联，复制本文给 AI Agent 即可完成安装
Chrome MCP 进程安全清理脚本
安全清理 Chrome MCP 旧进程脚本，自动保留最新进程只杀旧的
Chrome MCP 代理脚本（防抢焦点）
Chrome MCP 代理启动脚本，解决多 Agent 抢焦点问题。可移植版：路径全部 $HOME 化，头部含依赖与安装说明
MCP Config: chrome
MCP server configuration for chrome (portable, $HOME-based via bash -c). Requires chrome-mcp-proxy.sh + cdp-proxy.mjs in ~/scripts — see Chrome MCP 完整运维指南 asset for one-shot install
Multi-Browser MCP Proxies — Arc Browser & Chrome Beta Variants
Companion to 'CDP WebSocket 代理' for running parallel, isolated MCP fleets against Arc Browser and Chrome Beta on top of the same cdp-proxy.mjs. Portable $HOME paths + install steps included. Arc proxy auto-discovers the WebSocket path from /json/version; Chrome Beta proxy points at Beta's own DevToolsActivePort. Lets you run mcp__chrome__*, mcp__beta__*, and mcp__arc__* side-by-side with independent client state and no cross-talk.
Chrome Fleet — Multi-Agent Browser Pool with Shared Login State
Multi-agent control plane for chrome-devtools-mcp. Portable $HOME paths + install steps included. Two modes: (1) shared main Chrome — N CDP proxies on 9401/9402/9403... all multiplexing onto one logged-in Chrome :9222 with focus protection and ID isolation handled by cdp-proxy.mjs; (2) isolated agent Chromes — dedicated Chrome instance per agent on :930N with its own user-data-dir. Includes a status tool to inspect the running fleet.
Surf CLI — Control Chrome for AI Agents (No MCP)
Surf CLI lets agents control Chrome via a local extension + native host, offering agent-agnostic browser control without running an MCP server.
bb-browser — Browser-as-API CLI + MCP Server
Use your real Chrome login state as an API: bb-browser provides a CLI + MCP server with 103 commands across 36 platforms (Twitter/Reddit/YouTube/etc.).
MCP Config: Chrome Beta (Proxy Mode)
Chrome Beta browser MCP proxy configuration (portable, $HOME-based via bash -c). Bridges CDP WebSocket on port 9222 to MCP on port 9402. Requires chrome-beta-mcp-proxy.sh + cdp-proxy.mjs in ~/scripts (see Multi-Browser MCP Proxies asset)
AI Agent 的完整浏览器控制
The Chrome MCP server ecosystem gives AI agents the ability to see and interact with the web. There are multiple implementations optimized for different workflows: Puppeteer MCP (headless automation — scraping, testing, screenshots), Chrome DevTools MCP (connect to your live browser session for debugging), and Chrome MCP Proxy (background operation without focus-stealing popups).
For developers, the most transformative use case is AI-assisted debugging. Point the DevTools MCP at your running app, describe the bug, and the agent can inspect the DOM, read console errors, check network requests, analyze performance profiles, and suggest fixes — all without you touching the Chrome inspector. For QA teams, Puppeteer MCP enables AI-driven test generation: the agent navigates your app, identifies interactive elements, and writes end-to-end tests.
For web scraping and data extraction, Chrome MCP servers handle JavaScript-heavy sites that static scrapers can't reach. The agent can navigate SPAs, wait for dynamic content, extract structured data, and handle pagination. Browse the full browser automation tools directory for Playwright alternatives, and see web scraping tools for complementary extraction frameworks. For testing workflows, check AI testing tools on TokRepo.
The web is the world's largest API. Chrome MCP is how AI agents call it.
常见问题
What Chrome MCP servers are available?+
Three main options: (1) Puppeteer MCP (@modelcontextprotocol/server-puppeteer) — headless Chrome for automation, scraping, and testing. (2) Chrome DevTools MCP — connects to a running Chrome instance via the DevTools Protocol for live debugging. (3) Chrome MCP Proxy — runs Chrome in the background without focus-stealing, ideal for long-running agent tasks. Choose based on your use case.
Which Chrome MCP should I install first?+
Start with Puppeteer MCP — it's the official reference implementation, requires zero setup (downloads its own Chromium), and handles the most common use cases: navigating pages, taking screenshots, extracting content, filling forms, and clicking elements. Upgrade to DevTools MCP only if you need to debug a live browser session.
Can Chrome MCP handle JavaScript-heavy single-page apps?+
Yes. Unlike static HTTP scrapers, Chrome MCP runs a real browser engine. It executes JavaScript, renders React/Vue/Angular apps, waits for API calls to complete, and interacts with the fully rendered DOM. The agent can wait for specific elements to appear, handle infinite scroll, and navigate client-side routing.
Does it work with sites that require login?+
Yes. The agent can fill login forms, handle OAuth redirects, and maintain cookies across page navigations. For DevTools MCP connected to your real browser, it uses your existing session — no login needed. For Puppeteer, you can pass cookies or have the agent log in at the start of each session. Avoid storing credentials in MCP configs — use environment variables.
Can AI agents write browser tests with Chrome MCP?+
Yes. A common pattern: install Puppeteer MCP, point the agent at your app URL, and ask it to 'write end-to-end tests for the checkout flow.' The agent navigates the app, identifies forms and buttons, writes Playwright or Puppeteer test scripts, and can even run them to verify they pass. Much faster than writing browser tests manually.