
中文版
chrome-devtools-mcp(GitHub:https://github.com/ChromeDevTools/chrome-devtools-mcp ,2025-09-11 创建,今晨抓取时已 56,931★)是 Chrome DevTools 团队官方出品的 MCP 服务器和命令行工具:让你的编程助手(Antigravity、Claude、Cursor、Copilot 等)直接操控并检查一个真实的 Chrome 浏览器。
按 README 的说法,它把 Chrome DevTools 的全部能力交给了 Agent:录制性能轨迹并提炼出可执行的优化建议;高级浏览器调试——分析网络请求、截图、查看带源码映射堆栈的控制台报错;用 Puppeteer 做可靠的自动化,自动等待操作结果。装法是把它配进 Agent 的 MCP 服务器列表(或直接用 CLI),之后 Agent 就能在浏览器里复现 Bug、测性能、跑自动化,不用你再手动点一遍。Apache-2.0 开源。
需要注意:它会把浏览器实例里的内容暴露给 MCP 客户端,Agent 能查看、调试甚至修改浏览器里的数据——别在里面放敏感信息;另外这个项目是 9 月创建的“老面孔”,今天冲上 GitHub 日榜正是因为越来越多 Agent 工作流把它当成了标配。
为什么值得了解
- 官方出品的分量:Chrome DevTools 团队自己做的,不是第三方封装,能力和稳定性都有保障
- 调试方式变了:以前是人对着 DevTools 面板点,现在是 Agent 自己开浏览器、复现问题、读报错、量性能——“让 Agent 帮你调前端 Bug”终于有了趁手的官方工具
- 覆盖全流程:从性能分析、网络请求到截图和自动化,一个工具打通
- 56,000+ 星、日榜热门:说明它正在变成 AI 编程工作流里的标准件,现在接入正合适
English Version
chrome-devtools-mcp (GitHub: https://github.com/ChromeDevTools/chrome-devtools-mcp , created 2025-09-11, 56,931 stars at this morning’s fetch) is the official MCP server and CLI from the Chrome DevTools team: it lets your coding agent (Antigravity, Claude, Cursor, Copilot and others) control and inspect a live Chrome browser.
Per the README, it hands the full power of Chrome DevTools to agents: record performance traces and extract actionable insights; advanced browser debugging — analyze network requests, take screenshots, and inspect console errors with source-mapped stack traces; and reliable automation through Puppeteer, which waits for each action’s result automatically. You add it to your agent’s MCP server list (or use the CLI directly), and from then on the agent can reproduce bugs in the browser, measure performance and run automations without you clicking through everything by hand. Apache-2.0 licensed.
Two notes: it exposes the browser instance’s contents to MCP clients, so the agent can inspect, debug and even modify data in the browser — keep sensitive information out of it. And this is a September project resurfacing on today’s daily trending list precisely because more and more agent workflows are adopting it as standard kit.
Why it’s worth knowing
- Official weight: built by the Chrome DevTools team themselves, not a third-party wrapper, so capability and stability are solid
- Debugging has changed shape: where a human used to click through the DevTools panel, an agent can now open the browser, reproduce the bug, read the error and measure performance itself — “let the agent fix your frontend bug” finally has a proper official tool
- The whole flow in one tool: performance analysis, network requests, screenshots and automation, all covered
- 56,000+ stars and trending today: it is becoming a standard component of AI coding workflows, so now is the right time to adopt it

评论(0)
暂无评论