Browser Use
Browser agents and stealth cloud browsers for AI: run web tasks, fill forms, extract data.
作者提供的说明(英文原文)
Web search, Google SERP, scraping, screenshots, PDFs, brand data, and website diagnostics.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
API_KEY标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
适用于网页版、桌面版和手机版 Claude:
https://mcp.geekflare.com/{API_KEY}/mcp地址中有只有你知道的 {…} 部分(例如你的账户),请复制后自行填写。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"geekflare": {
"command": "npx",
"args": [
"-y",
"@geekflare/mcp"
],
"env": {
"API_KEY": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add --transport http geekflare 'https://mcp.geekflare.com/{API_KEY}/mcp'claude mcp add geekflare --env API_KEY=your-value -- npx -y @geekflare/mcp运行前请先填写空着的值。
地址中有只有你知道的 {…} 部分(例如你的账户),请复制后自行填写。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"geekflare": {
"url": "https://mcp.geekflare.com/{API_KEY}/mcp"
}
}
}粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"geekflare": {
"command": "npx",
"args": [
"-y",
"@geekflare/mcp"
],
"env": {
"API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
地址中有只有你知道的 {…} 部分(例如你的账户),请复制后自行填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"geekflare": {
"type": "http",
"url": "https://mcp.geekflare.com/{API_KEY}/mcp"
}
}
}粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"geekflare": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@geekflare/mcp"
],
"env": {
"API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
ChatGPT 只能连接互联网上的服务器:
https://mcp.geekflare.com/{API_KEY}/mcp地址中有只有你知道的 {…} 部分(例如你的账户),请复制后自行填写。
菜单名称可能因套餐和版本不同而略有差异。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
这个地址需要填入你自己的值,所以我们无法替你读取工具列表。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日
注册名称已通过域名 geekflare.com 验证(注册中心会核实域名所有权)
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
这个地址需要填入你自己的账户信息,所以我们无法替你检查
这个地址需要填入你自己的账户信息,所以我们无法替你检查
尚未检查;我们会在检查端点时读取工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:API_KEY
两种方式都可以:安装在你的电脑上,或者连接云端版本。
GitHub 仓库最后更新于–,最新版本(0.5.0)于12天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
同一分类中有:Browser Use, Chrome DevTools MCP, MCP Server Browser
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Browser agents and stealth cloud browsers for AI: run web tasks, fill forms, extract data.
MCP server for Chrome DevTools
MCP server for browser use access
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Official MCP (Model Context Protocol) server for the Geekflare. Connect Geekflare's web intelligence tools directly to Claude, Cursor, Windsurf, and other AI assistants.
Sign up at geekflare.com/api and copy your API key from the dashboard.
Add this to your `claude_desktop_config.json`:
{
"mcpServers": { "geekflare": { "command": "npx", "args": ["-y", "@geekflare/mcp"], "env": { "API_KEY": "your-api-key-here" } } } }
Add to your MCP settings:
{
"mcpServers": { "geekflare": { "command": "npx", "args": ["-y", "@geekflare/mcp"], "env": { "API_KEY": "your-api-key-here" } } } }
{
"mcpServers": { "geekflare": { "command": "docker", "args": ["run", "-i", "--rm", "-e", "API_KEY=your-api-key-here", "geekflare/mcp"] } } }
webScrapeScrape full page content from any URL. Returns HTML, Markdown, JSON, or LLM-optimised text — including structured data via ready-made extraction templates, custom CSS/XPath schemas, or AI-powered extraction.
| Parameter | Type | Default | Description | |||
|---|---|---|---|---|---|---|
url \* | string | — | Target URL | |||
device | desktop | mobile | desktop | Device to emulate | ||
format | array | ["markdown"] | Output formats (up to 3): html, markdown, json, markdown-llm, html-llm, text, text-llm | |||
proxyMode | boolean | "auto" | false | false never uses a proxy, auto retries through one if the site blocks the request, true always uses one | ||
proxyCountry | string | — | Route through a country ISO code (e.g. "us"), used when a proxy is active | |||
renderJS | boolean | auto | Execute JavaScript before extracting. If omitted, resolved automatically based on whether the page needs it | |||
fileOutput | boolean | false | Return a download URL instead of inline content | |||
blockAds | boolean | true | Block ads during scrape | |||
stealth | boolean | false | Bypass CAPTCHAs (slower) | |||
waitTime | number | 0 | Seconds to wait after page load before capturing content | |||
extractionMode | default | cssSchema | xpathSchema | template | default | Only used when format includes json |
template | product | contact | — | Ready-made extraction template when extractionMode is template | ||
extractionSchema | object | — | Custom field-extraction schema for cssSchema/xpathSchema modes | |||
aiPrompt | object | — | AI-powered extraction/analysis of the scraped page. Supports prompt, schema, listing, summary, sentiment, and keywords modes. Adds +6 credits |
metaScrapeScrape meta tags — title, description, Open Graph, Twitter cards, and more.
| Parameter | Type | Default | Description | |
|---|---|---|---|---|
url \* | string | — | Target URL | |
device | desktop | mobile | desktop | Device to emulate |
format | json | markdown | json | Response format |
proxyCountry | string | — | Country ISO code | |
renderJS | boolean | true | Whether to execute JavaScript | |
fileOutput | boolean | false | Return a download URL instead of inline content | |
blockAds | boolean | true | Block ads during scrape |
brandGet structured brand information for a website domain, including brand identity, logos, colors, typography, social profiles, links, page metadata, and company information.
The brand tool supports two levels of brand intelligence:
standard — Returns structured brand information available from the website.enriched — Includes additional LLM-synthesized company intelligence when available.| Parameter | Type | Default | Description | |
|---|---|---|---|---|
url \* | string | — | Target URL | |
refresh | boolean | false | Force an on-demand fetch and refresh the cached brand data | |
mode | standard | enriched | standard | Depth of brand data to return. enriched includes LLM-synthesized company intelligence |
{
"url": "https://example.com", "mode": "enriched" }
To bypass cached data and fetch fresh information:
{
"url": "https://example.com", "refresh": true, "mode": "enriched" }
The response may include:
Enriched company information depends on the publicly available information for the domain, so individual fields may be omitted.
screenshotCapture a screenshot of any website. Supports full-page, element-only, Retina, dark mode, transparent backgrounds, and AI-friendly link highlighting.
| Parameter | Type | Default | Description | ||
|---|---|---|---|---|---|
url \* | string | — | Target URL | ||
device | desktop | mobile | desktop | Device to emulate | |
type | png | jpeg | webp | png | Image format |
proxyCountry | string | — | Country ISO code | ||
fullPage | boolean | false | Capture the full page | ||
selector | string | — | CSS selector for the element to capture | ||
fallbackToFullPage | boolean | false | Fall back to a full-page capture if selector isn't found | ||
blockAds | boolean | true | Block ads | ||
hideCookie | boolean | true | Remove cookie banners | ||
skipCaptcha | boolean | true | Bypass anti-bot challenges | ||
addTimestamp | boolean | false | Add a timestamp | ||
highlightLinks | boolean | false | Draw borders around links/buttons — useful for AI vision models | ||
pageHeight | number | — | Custom page height in pixels | ||
viewportWidth | number | — | Viewport width | ||
viewportHeight | number | — | Viewport height | ||
captureBeyondViewport | boolean | — | Capture content beyond the configured viewport | ||
delay | number | — | Seconds to wait after page load | ||
quality | number | 90 | Image quality for JPEG/WEBP | ||
scaleFactor | number | — | Device pixel ratio | ||
theme | light | dark | auto | auto | Color scheme |
removeBackground | boolean | false | Remove page background (PNG only) | ||
disableAnimations | boolean | false | Freeze CSS animations before capture | ||
inline | boolean | false | Return image data inline instead of a CDN URL |
searchSearch the web and return clean, structured results. Supports web, news, and image search with optional AI-grounded answers.
| Parameter | Type | Default | Description |
|---|---|---|---|
query \* | string | — | Search query |
limit | number | 10 | Number of results |
time | string | — | Time filter: any, d, w, m, y, d7, h6 |
location | string | — | Country ISO code to localise results |
| `so
只显示了开头部分;其余内容请在仓库中阅读