Control the browser via the browser-use command-line tool, with support for persistent sessions (~50ms latency), multiple parallel sessions, cloud browsers, local tunnels, and more.
Installation
The installation script will automatically install a Python virtual environment,browser-use package, Playwright, and Chromium, with no need to manually install Playwright or Selenium.
macOS / Linux
curl -fsSL https://browser-use.com/cli/install.sh | bash
Windows(Git Bash must be installed first:winget install Git.Git）
# 在 Git Bash 中运行
curl -fsSL https://browser-use.com/cli/install.sh | bash
Verify after installation
browser-use doctor
For detailed installation steps for the three platforms, prerequisites, and troubleshooting, see references/setup.md。
Standard Workflow
The browser runs persistently between commands, with no need to restart each time.
1. Open a web page
browser-use open https://example.com
2. View page state (get interactive elements and indices)
browser-use state
# 返回 URL、标题、所有可点击元素及其索引号
3. Interact
# 点击元素（使用 state 返回的索引号）
browser-use click 5
# 在输入框中输入文字（点击 + 输入）
browser-use input 3 "搜索内容"
# 通用打字（输入到当前焦点元素）
browser-use type "Hello World"
# 滚动页面
browser-use scroll down
browser-use scroll up
browser-use scroll down --amount 1000
# 悬停 / 双击 / 右键
browser-use hover 7
browser-use dblclick 4
browser-use rightclick 2
# 键盘按键
browser-use keys Enter
browser-use keys "Control+a"
# 下拉框选择
browser-use select 4 "option-value"
# 上传文件
browser-use upload 6 /path/to/file.pdf
4. Get data
# 截图
browser-use screenshot page.png
browser-use screenshot --full fullpage.png
# 执行 JavaScript 获取数据
browser-use eval "document.title"
browser-use eval "Array.from(document.querySelectorAll('h2')).map(h=>h.textContent)"
# 获取元素信息
browser-use get title
browser-use get html
browser-use get html --selector "main"
browser-use get text 5
browser-use get value 3
browser-use get attributes 2
browser-use get bbox 4
5. Wait for conditions
browser-use wait selector ".loading" --state hidden
browser-use wait text "Success"
browser-use wait selector "h1" --timeout 5000
6. Page navigation
browser-use back
browser-use tab list
browser-use tab new https://example.com
browser-use tab switch 2
browser-use tab close
7. End
browser-use close
browser-use close --all
Global Options
| Option | Description |
|---|---|
--headed |
Show the browser window (headless mode by default) |
--profile [NAME] |
Use the real Chrome profile (preserve login state, cookies) |
--connect |
Automatically connect to an already running Chrome |
--cdp-url <url> |
Connect to the browser via a CDP URL |
--session NAME |
Named session (supports multiple parallel browsers) |
--json |
Output in JSON format |
--mcp |
Run in MCP server mode |
Use the real Chrome profile
# 使用 Default 配置，已登录的网站无需重新登录
browser-use --profile open https://mail.google.com
# 使用指定 Chrome 配置
browser-use --profile "Profile 1" open https://example.com
Multiple parallel sessions
# 启动两个独立的浏览器会话
browser-use --session work open https://app.example.com
browser-use --session personal open https://social.example.com
# 分别操作
browser-use --session work click 3
browser-use --session personal type "hello"
# 查看活跃会话
browser-use sessions
Cookie management
browser-use cookies get
browser-use cookies get --url https://example.com
browser-use cookies set token abc123 --domain .example.com --secure
browser-use cookies export cookies.json
browser-use cookies import cookies.json
browser-use cookies clear
Python automation scripts
# 执行 Python（持久会话，变量跨命令保留）
browser-use python "
for i in range(5):
browser.scroll('down')
browser.wait(0.5)
"
# 运行 Python 文件
browser-use python --file script.py
# 查看已定义变量
browser-use python --vars
Cloud browser
# 保存 API key
browser-use cloud login <browser-use-api-key>
# 启动云浏览器
browser-use cloud connect
# 创建自动化任务
browser-use cloud v2 POST /tasks '{"task":"搜索 AI 新闻","url":"https://google.com"}'
Local tunnel (expose local services)
# 将本地 3000 端口暴露到公网
browser-use tunnel 3000
# → 返回公网 URL，可供云浏览器访问
browser-use tunnel list
browser-use tunnel stop 3000
Typical Use Cases
Web data collection
browser-use open https://news.ycombinator.com
browser-use eval "JSON.stringify([...document.querySelectorAll('.titleline a')].map(a=>({title:a.textContent,href:a.href})).slice(0,10))"
browser-use close
Form filling
browser-use open https://example.com/signup
browser-use state # 找到表单元素索引
browser-use input 2 "user@example.com"
browser-use input 3 "password123"
browser-use click 5 # 点击提交
browser-use screenshot result.png
browser-use close
Search and screenshot
browser-use open https://www.google.com
browser-use state
browser-use input 2 "browser-use github"
browser-use keys Enter
browser-use screenshot search.png
browser-use close
Troubleshooting
| Issue | Solution |
|---|---|
| Chromium not found | browser-use install |
| Page load timeout | Use the wait command or check the network |
| Incorrect element index | Re-run state to get the latest index |
| Login state required | Use the --profile option |
| Windows ARM64 | Install the x64 version of Python (via emulation) |
| Daemon error | taskkill /IM python.exe /F（Windows） |