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)
Any website as MCP tools: a real Chrome signed in as you, and shortcuts the agent saves as tools.
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。
ANYSITE_HEADLESS, ANYSITE_CHROME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 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"anysite": {
"command": "npx",
"args": [
"-y",
"anysite-mcp"
],
"env": {
"ANYSITE_HEADLESS": "your-value",
"ANYSITE_CHROME": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add anysite --env ANYSITE_HEADLESS=your-value --env ANYSITE_CHROME=your-value -- npx -y anysite-mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"anysite": {
"command": "npx",
"args": [
"-y",
"anysite-mcp"
],
"env": {
"ANYSITE_HEADLESS": "your-value",
"ANYSITE_CHROME": "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。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"anysite": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"anysite-mcp"
],
"env": {
"ANYSITE_HEADLESS": "your-value",
"ANYSITE_CHROME": "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。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
ANYSITE_HEADLESS envANYSITE_CHROME env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。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
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10日
注册名称已通过 GitHub 账户 DwytMCL 验证,该账户拥有此仓库
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
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
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。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.6.0)于今天进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
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:↓ 每周下载量i★ 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.
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。
A real Chrome, signed in as you, on your own computer.
claude mcp add anysite -- npx -y anysite-mcp
No API keys. No cloud browser. No extra LLM. Your logins never leave your machine.
Quick start · Why it's different · Tools · Shortcuts · Shortcut Library · Packs · Safety · FAQ
anysite-mcp is an MCP server that drives a real Chrome on your computer. Every site gets its own browser profile that you sign in to once, by hand. The agent then works the site the way a screen reader does, by role and name, so small redesigns don't break it.
When the agent gets a job done, it can save the steps as a shortcut, and that shortcut becomes a new tool it can call in one step next time.
Your accounts, kept apart
One Chrome profile per site, signed in by hand once. Google sign-in works.
Gets faster with use
A job the agent has figured out becomes a one-call tool that's there next session. Shortcuts are tested before they're saved and fail loudly if a page stops looking right.
Repairs itself
When a site renames a button or swaps it for a link, the shortcut finds the new one and updates itself. No extra LLM needed.
Learns by watching
Run anysite-mcp record github.com my_job, do the job by hand, and it's a tool.
You set the limits
Per site: read-only, ask before sending, buying or deleting, and which domains it may open.
Light on tokens
After each action the agent sees only what changed, and points at elements by ref. web_extract turns lists and tables into JSON.
A shared, health-checked library
Community shortcuts are tested against the live sites every week, and you can search them from the terminal.
Packs for popular sites
Gemini and NotebookLM are included. A new pack is one file.
You: Ask Gemini to summarise this PDF, then add the summary as a source to my "Thesis" notebook in NotebookLM.
Claude: gemini_ask { prompt: "Summarise this PDF in 10 bullet points", files: ["~/Downloads/paper.pdf"] } → - The study compares … (Chat: https://gemini.google.com/app/…) notebooklm_add_source { notebook: "Thesis", text: "- The study compares …" } → Added pasted text.
You: Check my GitHub notifications.
Claude: web_open { site: "github.com", url: "/notifications" } → This looks like a sign-in page. Call web_login for "github.com" … web_login { site: "github.com" } "A Chrome window opened on GitHub. Sign in there, then quit that window and tell me." You: done Claude: web_open { site: "github.com", url: "/notifications" } → (accessibility tree of your inbox) web_save_shortcut { site: "github.com", name: "notifications", description: "List unread GitHub notifications", steps: [{ "action": "open", "url": "/notifications" }, { "action": "read", "role": "main" }] } → Saved github_com_notifications.
(Next week, in a new session) You: Anything new on GitHub? Claude: github_com_notifications {} → 3 unread: …
Most browser tools make the agent rediscover a site every session. anysite-mcp keeps what it learned.
| anysite-mcp | Playwright MCP | browser-use / Stagehand | |
|---|---|---|---|
| Signed in as you | Yes, a separate profile per site, signed in by hand once | One browser for everything (its extension can use your Chrome) | Usually cloud or fresh browsers |
| Google sign-in (Gemini, NotebookLM, Gmail) | Works: sign-in happens in a plain, non-automated window | Often refused as an "insecure browser" | Often refused |
| Gets faster with use | The agent saves working steps as new tools | No | Some cache actions |
| Survives redesigns | Saved steps repair themselves | No | With an LLM call |
| Learn by watching you | `record` mode | No | No |
| Per-site permissions | Read-only, confirm, allowed domains | No | No |
| Site packs and shared shortcuts | Gemini, NotebookLM, a health-checked library | No | No |
| Needs an LLM API key | No | No | Yes |
[!NOTE]
Profiles are kept apart on purpose. Copying a signed-in profile so two browsers share one session gets that
session revoked by Google (and other careful sites), which logs you out everywhere.
Requirements: Node 22+ and Chrome (or
npx playwright install chromium). Works on macOS, Windows and Linux.
1. Sign in to the sites you want, once each. A normal Chrome window opens; sign in, then quit it (Cmd+Q on a Mac, close all its windows on Windows and Linux).
npx -y anysite-mcp login gemini
npx -y anysite-mcp login notebooklm npx -y anysite-mcp login github.com
2. Add the server to your agent.
Claude Code
claude mcp add anysite -- npx -y anysite-mcp
Claude Desktop, Cursor, Windsurf and other MCP clients
Add this to claude_desktop_config.json or the client's MCP settings:
{ "mcpServers": { "anysite": { "command": "npx", "args": ["-y", "anysite-mcp"] } } }
3. Ask your agent to do something on one of those sites. That's it.
With no arguments the server offers every site you have signed in to. To pick sites, list them: npx -y anysite-mcp gemini github.com. The agent can also add a new site at any time by calling web_login.
anysite-mcp is listed in the MCP Registry as io.github.DwytMCL/anysite-mcp, so clients that browse the registry can find and install it.
Each tool takes site: a pack name like gemini, or a domain like github.com.
| Tool | What it does |
|---|---|
web_snapshot | The page's accessibility tree: roles, names, text, and a ref per element ([ref=e12]). Give ref or role/name to see only part of it, depth for an outline. |
web_open | Go to a URL (absolute or relative to the site) |
web_click, web_type, web_select, web_press | Act on an element found by ref, by role and name, or by visible text (add within: "dialog" and so on to look only inside one part of the page). Returns only what changed; show: "full" for the whole page. |
web_read | Text of the page or of one element |
web_extract | Rows as JSON, one per matching element (role: "row", "listitem", "article"…), with optional fields looked up in each |
web_wait | Wait, or wait until some text appears |
web_upload, web_download | Give files to a file picker; save a download |
web_screenshot | PNG of the page, for when the tree is not enough |
web_login | Open a plain window so you can sign in |
web_save_shortcut, web_list_shortcuts, web_delete_shortcut | Turn a working sequence into a tool named _ (tried once before it's saved) |
web_record | Watch the user do a job in the site's window and save it as a shortcut |
Every tool carries MCP annotations for clients that use them, for example to ask before a destructive tool runs. Tools that only read are marked read-only, and tools that can change something on a page (clicking, typing, choosing, pressing keys, uploading) are marked destructive. Shortcut tools are marked by what their steps do.
| Pack | Tools |
|---|---|
gemini | gemini_ask (prompt, files, model, tools such as Create image or Deep Research; returns the reply as markdown), gemini_list_chats |
notebooklm | notebooklm_ask, notebooklm_add_source (URLs, text, files), notebooklm_create_notebook, notebooklm_list_notebooks |
A shortcut is a list of steps, each one the arguments of a web_* tool plus its action. {{param}} is filled in from the tool's arguments, and the shortcut returns the text of its read steps.
expect is text that must be on the page afterwards, so a shortcut fails instead of returning the wrong thing.test_args (set verify: false for steps that must not happen twice, like sending).~/.anysite-mcp/sites/ .json, which you can edit, share or commit.{
"search": { "description": "Search Hacker News and return the results", "params": { "query": "What to search for" }, "steps": [ { "action": "open", "url": "https://hn.algolia.com/?q={{query}}" }, { "action": "wait", "seconds": 5, "until_text": "results", "expect": "{{query}}" }, { "action": "read", "role": "main" } ] } }
Params can also be { "description", "type": "number" | "boolean", "optional", "default" }.
A step whose element is gone is repaired before it fails:
Only a single clear match counts. The shortcut file is updated and the agent is told what changed. If nothing fits, the error says which step and why, and the agent can redo the job and save the shortcut again.
npx -y anysite-mcp record github.com star_repo
Chrome opens on the site; click and type as usual, then press Enter in the terminal. Each typed value becomes a parameter, with what you typed as its default. The agent can do the same with web_record.
npx -y anysite-mcp doctor
doctor follows each shortcut while its steps only open and read pages, checks that the element of the first step that would act is there, and repairs what it can. It never clicks or types.
Each parameter gets its default if it has one. Otherwise an optional parameter is left empty, so the check sees the p
只显示了开头部分;其余内容请在仓库中阅读