Browserless Agent 🌐
A comprehensive web automation skill for OpenClaw that provides 30+ browser actions including navigation, data extraction, form filling, screenshot capture, PDF generation, file handling, and advanced web scraping capabilities.
🚀 Features
- - Navigation: Full control over page navigation, redirects, and history
- Data Extraction: Get text, attributes, HTML, computed styles, and structured data
- Form Automation: Type text, click buttons, select options, upload files
- Visual Capture: Screenshots (full page, element-only, viewport)
- Content Generation: Save pages as PDF with custom options
- Advanced Interactions: Hover, drag-drop, keyboard shortcuts, scrolling
- Multi-tab Support: Manage multiple pages and windows
- Network Control: Intercept requests, modify headers, block resources
- Storage Access: Manage cookies, localStorage, sessionStorage
- Dynamic Content: Wait for selectors, network idle, custom conditions
- iFrames: Interact with nested frame content
- Browser State: Emulate devices, set geolocation, handle dialogs
🔧 Configuration
This skill requires the BROWSERLESS_URL environment variable to be configured in OpenClaw.
Optionally, you can also set BROWSERLESS_TOKEN for authentication.
To set it up:
- 1. Open OpenClaw settings
- Navigate to Skills → browserless-agent
- Enter your Browserless base URL in the API Key field
- (Optional) Add BROWSERLESS_TOKEN in the env section for token authentication
Configuration Examples:
Cloud Service (with token):
CODEBLOCK0
Local Service (no token):
CODEBLOCK1
Custom Endpoint:
CODEBLOCK2
The skill will automatically:
- - Add
/playwright/chromium if endpoint is not specified - Append token as query parameter if
BROWSERLESS_TOKEN is set - Work with or without authentication token
Get your Browserless service at: browserless.io
Get your Browserless service at: browserless.io
📖 Available Actions
Navigation & Page Control
navigate
Navigate to a URL.
CODEBLOCK3
go_back
Navigate to previous page in history.
CODEBLOCK4
go_forward
Navigate to next page in history.
CODEBLOCK5
reload
Reload the current page.
CODEBLOCK6
wait_for_load
Wait for page to finish loading.
CODEBLOCK7
Data Extraction
get_text
Extract text content from element(s).
CODEBLOCK8
get_attribute
Get attribute value from element(s).
CODEBLOCK9
get_html
Get inner or outer HTML of element(s).
CODEBLOCK10
get_value
Get input value from form element(s).
CODEBLOCK11
get_style
Get computed CSS style property.
CODEBLOCK12
get_multiple
Extract multiple pieces of data at once.
CODEBLOCK13
Interaction & Input
type_text
Type text into an element.
CODEBLOCK14
click
Click on an element.
CODEBLOCK15
double_click
Double-click on an element.
CODEBLOCK16
right_click
Right-click (context menu) on an element.
CODEBLOCK17
hover
Move mouse over an element.
CODEBLOCK18
focus
Focus on an element.
CODEBLOCK19
select_option
Select option(s) in a dropdown.
CODEBLOCK20
check
Check a checkbox or radio button.
CODEBLOCK21
uncheck
Uncheck a checkbox.
CODEBLOCK22
upload_file
Upload file(s) to file input.
CODEBLOCK23
press_key
Press keyboard key(s).
{"key": "Enter"}
Common keys: Enter, Tab, Escape, ArrowDown, Control+A, etc.
keyboard_type
Type text with keyboard (supports shortcuts).
CODEBLOCK25
Scrolling & Position
scroll_to
Scroll to specific position.
CODEBLOCK26
scroll_into_view
Scroll element into viewport.
CODEBLOCK27
scroll_to_bottom
Scroll to bottom of page.
CODEBLOCK28
scroll_to_top
Scroll to top of page.
CODEBLOCK29
Visual & Capture
screenshot
Take screenshot of page or element.
CODEBLOCK30
pdf
Generate PDF from current page.
CODEBLOCK31
Evaluation & Execution
evaluate
Execute JavaScript in page context.
CODEBLOCK32
evaluate_function
Execute JavaScript function with arguments.
CODEBLOCK33
Waiting & Timing
wait_for_selector
Wait for element to appear.
{"selector": ".dynamic-content", "timeout": 10000, "state": "visible"}
States: visible, hidden, attached, detached
wait_for_timeout
Wait for specified milliseconds.
CODEBLOCK35
wait_for_function
Wait for JavaScript expression to return truthy.
CODEBLOCK36
wait_for_navigation
Wait for navigation to complete.
{"timeout": 30000, "wait_until": "networkidle"}
wait_until options: load, domcontentloaded, networkidle
Element State Checking
is_visible
Check if element is visible.
CODEBLOCK38
is_enabled
Check if element is enabled.
CODEBLOCK39
is_checked
Check if checkbox/radio is checked.
CODEBLOCK40
element_exists
Check if element exists in DOM.
CODEBLOCK41
element_count
Count elements matching selector.
CODEBLOCK42
Storage & Cookies
get_cookies
Get all cookies or specific cookie.
CODEBLOCK43
set_cookie
Set a cookie.
CODEBLOCK44
delete_cookies
Delete cookies.
{"name": "session_id"}
Omit name to delete all cookies.
get_local_storage
Get localStorage item.
CODEBLOCK46
set_local_storage
Set localStorage item.
CODEBLOCK47
clear_local_storage
Clear all localStorage.
CODEBLOCK48
Network & Requests
set_extra_headers
Set extra HTTP headers for all requests.
CODEBLOCK49
block_resources
Block specific resource types.
{"types": ["image", "stylesheet", "font"]}
Types: document, stylesheet, image, media, font, script, xhr, fetch, other
get_page_info
Get comprehensive page information.
{}
Returns: title, url, html (optional), viewport size, etc.
iFrame Handling
get_frame_text
Get text from element inside iframe.
CODEBLOCK52
click_in_frame
Click element inside iframe.
CODEBLOCK53
Multi-Page/Tab
new_page
Open a new page/tab.
CODEBLOCK54
close_page
Close a specific page.
CODEBLOCK55
switch_page
Switch to a different page.
CODEBLOCK56
list_pages
List all open pages.
CODEBLOCK57
Browser Context
set_viewport
Set viewport size.
CODEBLOCK58
emulate_device
Emulate mobile device.
{"device": "iPhone 12"}
Common devices: iPhone 12, iPad Pro, Galaxy S21, Pixel 5
set_geolocation
Set geolocation.
CODEBLOCK60
set_user_agent
Set custom user agent.
CODEBLOCK61
Advanced Automation
drag_and_drop
Drag element and drop on target.
CODEBLOCK62
fill_form
Fill multiple form fields at once.
CODEBLOCK63
extract_table
Extract data from HTML table.
CODEBLOCK64
extract_links
Extract all links from page.
CODEBLOCK65
handle_dialog
Set how to handle JavaScript dialogs (alert/confirm/prompt).
{
"action": "accept",
"text": "Optional prompt response"
}
Actions: accept, dismiss
💡 Usage Examples
Example 1: Web Scraping
CODEBLOCK67
Example 2: Form Automation
CODEBLOCK68
Example 3: Screenshot with Element Highlight
CODEBLOCK69
Example 4: PDF Generation
CODEBLOCK70
🎯 OpenClaw Integration
To use this skill from OpenClaw, the agent can automatically invoke these actions. Examples:
User: "Take a screenshot of example.com"
Agent: Executes screenshot action with the URL
User: "What's the title of wikipedia.org?"
Agent: Navigates to Wikipedia and extracts text from the title element
User: "Search for 'Python' on Google and get the first result link"
Agent: Navigates to Google, types in search, clicks search, extracts first result
🔒 Security Notes
- - Browserless connection uses WebSocket over TLS (wss://)
- Credentials are never logged or exposed in responses
- All browser actions are isolated in the Browserless container
- No local browser installation required
🐛 Troubleshooting
Connection fails:
- - Verify
BROWSERLESS_WS URL is correct - Check if token is valid and not expired
- Ensure network allows WebSocket connections
Timeout errors:
- - Increase timeout values for slow-loading pages
- Use
wait_for_selector before interacting with dynamic content - Consider using
wait_until: "networkidle" for AJAX-heavy sites
Element not found:
- - Verify selector using browser DevTools
- Wait for element to load with INLINECODE72
- Check if element is inside an iframe
📚 Resources
Browserless Agent 🌐
一个为OpenClaw设计的全面网页自动化技能，提供30多种浏览器操作，包括导航、数据提取、表单填写、截图捕获、PDF生成、文件处理以及高级网页抓取功能。
🚀 功能特性
- - 导航：完全控制页面导航、重定向和历史记录
- 数据提取：获取文本、属性、HTML、计算样式和结构化数据
- 表单自动化：输入文本、点击按钮、选择选项、上传文件
- 视觉捕获：截图（全页、仅元素、视口）
- 内容生成：使用自定义选项将页面保存为PDF
- 高级交互：悬停、拖放、键盘快捷键、滚动
- 多标签支持：管理多个页面和窗口
- 网络控制：拦截请求、修改标头、阻止资源
- 存储访问：管理Cookie、localStorage、sessionStorage
- 动态内容：等待选择器、网络空闲、自定义条件
- iFrame：与嵌套框架内容交互
- 浏览器状态：模拟设备、设置地理位置、处理对话框
🔧 配置
此技能需要在OpenClaw中配置BROWSERLESS_URL环境变量。
可选地，您还可以设置BROWSERLESS_TOKEN进行身份验证。
设置步骤：
- 1. 打开OpenClaw设置
- 导航至技能 → browserless-agent
- 在API密钥字段中输入您的Browserless基础URL
- （可选）在env部分添加BROWSERLESS_TOKEN用于令牌认证
配置示例：
云服务（带令牌）：
BROWSERLESS_URL=wss://chrome.browserless.io
BROWSERLESS_TOKEN=your-token-here
本地服务（无令牌）：
BROWSERLESS_URL=ws://localhost:3000
自定义端点：
BROWSERLESS_URL=wss://your-host.com/playwright/chromium
BROWSERLESS_TOKEN=optional-token
该技能将自动：
- - 如果未指定端点，则添加/playwright/chromium
- 如果设置了BROWSERLESS_TOKEN，则将其作为查询参数附加
- 无论是否使用认证令牌均可工作
获取您的Browserless服务：browserless.io
获取您的Browserless服务：browserless.io
📖 可用操作
导航与页面控制
navigate
导航到URL。
json
{url: https://example.com}
go_back
导航到历史记录中的上一页。
json
{}
go_forward
导航到历史记录中的下一页。
json
{}
reload
重新加载当前页面。
json
{hard: false}
waitforload
等待页面加载完成。
json
{timeout: 30000}
数据提取
get_text
提取元素的文本内容。
json
{selector: h1, all: false}
get_attribute
获取元素的属性值。
json
{selector: img, attribute: src, all: false}
get_html
获取元素的内部或外部HTML。
json
{selector: article, outer: false, all: false}
get_value
获取表单元素的输入值。
json
{selector: input[name=email]}
get_style
获取计算后的CSS样式属性。
json
{selector: .box, property: background-color}
get_multiple
一次性提取多条数据。
json
{
extractions: [
{name: title, selector: h1, type: text},
{name: image, selector: img, type: attribute, attribute: src},
{name: price, selector: .price, type: text}
]
}
交互与输入
type_text
在元素中输入文本。
json
{selector: input[type=search], text: hello world, delay: 0, clear: true}
click
点击元素。
json
{selector: button.submit, force: false, delay: 0}
double_click
双击元素。
json
{selector: .item}
right_click
右键点击（上下文菜单）元素。
json
{selector: .context-target}
hover
将鼠标悬停在元素上。
json
{selector: .menu-item}
focus
聚焦到元素。
json
{selector: input}
select_option
在下拉菜单中选择选项。
json
{selector: select, values: [option1, option2]}
check
勾选复选框或单选按钮。
json
{selector: input[type=checkbox]}
uncheck
取消勾选复选框。
json
{selector: input[type=checkbox]}
upload_file
上传文件到文件输入框。
json
{selector: input[type=file], files: [path/to/file.pdf]}
press_key
按下键盘键。
json
{key: Enter}
常用键：Enter、Tab、Escape、ArrowDown、Control+A等。
keyboard_type
使用键盘输入文本（支持快捷键）。
json
{text: Hello World}
滚动与定位
scroll_to
滚动到指定位置。
json
{x: 0, y: 500}
scrollintoview
将元素滚动到视口中。
json
{selector: .footer}
scrolltobottom
滚动到页面底部。
json
{}
scrolltotop
滚动到页面顶部。
json
{}
视觉与捕获
screenshot
截取页面或元素的屏幕截图。
json
{
path: screenshot.png,
full_page: true,
selector: null,
quality: 90,
type: png
}
pdf
从当前页面生成PDF。
json
{
path: page.pdf,
format: A4,
landscape: false,
margin: {top: 1cm, right: 1cm, bottom: 1cm, left: 1cm},
print_background: true
}
评估与执行
evaluate
在页面上下文中执行JavaScript。
json
{expression: document.title}
evaluate_function
使用参数执行JavaScript函数。
json
{
function: (x, y) => x + y,
args: [5, 10]
}
等待与定时
waitforselector
等待元素出现。
json
{selector: .dynamic-content, timeout: 10000, state: visible}
状态：visible、hidden、attached、detached
waitfortimeout
等待指定的毫秒数。
json
{timeout: 2000}
waitforfunction
等待JavaScript表达式返回真值。
json
{
expression: () => document.readyState === complete,
timeout: 10000
}
waitfornavigation
等待导航完成。
json
{timeout: 30000, wait_until: networkidle}
wait_until选项：load、domcontentloaded、networkidle
元素状态检查
is_visible
检查元素是否可见。
json
{selector: .modal}
is_enabled
检查元素是否启用。
json
{selector: button}
is_checked
检查复选框/单选按钮是否被选中。
json
{selector: input[type=checkbox]}
element_exists
检查元素是否存在于DOM中。
json
{selector: .optional-element}
element_count
计算匹配选择器的元素数量。
json
{selector: .list-item}
存储与Cookie
get_cookies
获取所有Cookie或特定Cookie。
json
{name: session_id}
set_cookie
设置Cookie。
json
{
name: user_preference,
value: dark_mode,
domain: example.com,
path: /,
expires: 1735689600,
httpOnly: false,
secure: true,
sameSite: Lax
}
delete_cookies
删除Cookie。
json
{name: session_id}
省略名称将删除所有Cookie。
getlocalstorage
获取localStorage项。
json
{key: user_data}
setlocalstorage
设置local
