chrome-devtools-mcp
Enables AI to interact with Chrome DevTools Protocol to inspect and debug live web pages, including DOM, styles, console logs, and network requests.
README
chrome-devtools-mcp
让 AI 通过 MCP 协议直接调用 Chrome DevTools Protocol(CDP),获取浏览器运行时 DOM 结构、计算样式、控制台日志、网络请求等信息,彻底解决 AI 只能读源码、无法获取真实运行时上下文的问题。
前置条件
- Node.js 18+
- Chrome 浏览器
安装
npm install
npm run build
启动 Chrome(调试模式)
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome \
--remote-debugging-port=9222 \
--user-data-dir=/tmp/chrome-debug-profile
推荐添加 alias 到 ~/.zshrc,之后直接用 chrome-debug 启动:
alias chrome-debug='/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-debug-profile'
OpenCode 集成
在 ~/.config/opencode/opencode.json 中添加:
{
"mcp": {
"chrome-devtools": {
"type": "local",
"command": ["node", "/path/to/chrome-devtools-mcp/dist/index.js"],
"env": {
"CHROME_DEBUGGING_PORT": "9222"
},
"enabled": true
}
}
}
注意:
command必须是数组格式,不能是字符串。
查看实时日志
MCP server 的所有 CDP 调用日志写入 ~/chrome-devtools-mcp.log,在任意终端运行以下命令实时查看:
tail -f ~/chrome-devtools-mcp.log
工具清单(11 个)
DOM 工具
| 工具 | 说明 |
|---|---|
list_tabs |
列出所有可调试的 Tab(type=page) |
query_selector |
按 CSS 选择器查找元素,返回 tagName/id/classList |
get_dom_tree |
获取 DOM 子树,selector 可选,depth 最大 10 |
get_ancestors |
获取元素完整祖先链(含指定样式 + 自动诊断),sticky/overflow/z-index 问题根因定位专用 |
CSS 工具
| 工具 | 说明 |
|---|---|
get_computed_styles |
获取元素计算样式,返回 summary(布局关键属性)+ all(完整属性) |
get_matched_styles |
获取元素匹配的 CSS 规则列表(含来源文件、行号) |
get_box_model |
获取元素盒模型尺寸(width/height/margin/padding/border/content) |
Runtime 工具
| 工具 | 说明 |
|---|---|
get_layout_metrics |
获取页面视口/内容尺寸(viewportWidth/viewportHeight/contentWidth/contentHeight) |
evaluate_js |
在页面上下文执行 JS 表达式,自动拦截 cookie/localStorage/fetch 等敏感 API |
get_console_logs |
获取控制台日志,支持按 level 过滤(log/info/warn/error/debug) |
Network 工具
| 工具 | 说明 |
|---|---|
get_network_requests |
获取网络请求记录,支持 url_filter 关键词过滤,返回 method/status/duration/headers |
注意:
get_console_logs和get_network_requests只收集 MCP 连接后的事件,连接前的记录无法追溯。
智能诊断(Phase 3)
get_ancestors 会自动在返回结果里附带 diagnosis 字段,无需 AI 额外推理即可定位根因:
{
"ancestors": [...],
"diagnosis": {
"hints": [
{
"depth": 2,
"tagName": "section",
"classList": ["arco-layout"],
"issue": "position:sticky 失效 — 祖先节点 overflow 非 visible/unset 会阻断 sticky",
"property": "overflow",
"value": "hidden",
"impact": "blocking"
}
],
"summary": "发现 1 个阻断性问题。第2层 <section> overflow:hidden"
}
}
impact 三级:blocking(直接阻断)/ likely(大概率影响)/ possible(可能有影响)
检测覆盖:
overflow: hidden/scroll/auto/clip→ sticky 失效transform非 none → fixed/sticky 失效will-change: transform/opacity/filter→ 层叠上下文异常contain: layout/paint/strict/content→ fixed 失效filter/backdrop-filter→ 层叠上下文异常
典型用法
诊断 sticky 失效
1. 调用 get_computed_styles(".sticky-header")
→ summary.position: sticky(确认已设置)
2. 调用 get_ancestors(".sticky-header")
→ diagnosis.summary: "发现 1 个阻断性问题。第3层 <div.el-scrollbar__wrap> overflow:hidden"
→ 根因定位完成,无需进一步推理
排查 JS 错误
1. 调用 get_console_logs(level="error")
→ 获取页面 error 级别日志
2. 调用 evaluate_js("document.querySelector('.btn').disabled")
→ 获取按钮当前的 disabled 状态
排查接口问题
1. 调用 get_network_requests(url_filter="/api/cooperation")
→ 查看合作相关接口的请求状态和响应时间
安全说明
- 调试端口绑定
127.0.0.1,禁止公网暴露 evaluate_js自动拦截document.cookie、localStorage、sessionStorage、fetch、XMLHttpRequest- MCP Server 仅本地运行,数据不出 localhost
- 仅供开发环境使用,勿集成到生产流程
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.