MCP Tools Reference

BrowseHand MCP server provides the following tools. When you make natural language requests to Claude Desktop, the appropriate tools are automatically invoked.

Tool Summary

Tool	Description
ping_extension	Check extension connection status
read_browser_content	Read text content from current tab
execute_script	Execute JavaScript code
extract_structured_data	Extract structured data from repeating elements
click_element	Click element by CSS selector
scroll_page	Scroll page or specific element
wait_for_element	Wait for element to appear
navigate_to	Navigate to a URL
get_current_url	Get current URL
get_dom_snapshot	Get DOM structure snapshot
save_to_csv	Save data to CSV file
save_to_json	Save data to JSON file

---

Connection

ping_extension

Check the connection status with Chrome Extension.

Parameters: None

Example:

Check the extension connection status

---

Page Reading

read_browser_content

Read HTML content from the currently active browser tab.

Parameter	Type	Required	Description
selector	string	❌	DOM selector to extract (default: body)

Examples:

Read the current page content
Read only the .main-content area using read_browser_content

get_dom_snapshot

Get the DOM structure (main tags and text) for AI analysis. Unnecessary tags (script, style, noscript, etc.) are automatically removed.

Parameters: None

Example:

Analyze this page's DOM structure

get_current_url

Get the current browser tab’s URL.

Parameters: None

---

Script Execution

execute_script

Execute JavaScript code in the browser.

Parameter	Type	Required	Description
code	string	✅	JavaScript code to execute

Examples:

Use execute_script to run document.title
Use execute_script to count all links on this page

Page Analysis Use Case

Useful for identifying current selectors on sites with dynamic class names (like Google Maps).

Use execute_script to find the business list container selector

---

Data Extraction

extract_structured_data

Extract structured data from repeating elements.

Parameter	Type	Required	Description
containerSelector	string	✅	Selector for each repeating item container (e.g., .business-item)
fields	object	✅	Field definitions. Key is field name, value is relative selector within container
limit	number	❌	Maximum items to extract (default: all)

Example:

Use extract_structured_data on .product-card with fields:
{ title: '.product-name', price: '.product-price', link: 'a' }

Result Example:

[
  { "title": "Product A", "price": "$10.00", "link": "/product/1" },
  { "title": "Product B", "price": "$20.00", "link": "/product/2" }
]

---

Page Manipulation

click_element

Click an element by CSS selector.

Parameter	Type	Required	Description
selector	string	✅	CSS selector of element to click
waitAfter	number	❌	Milliseconds to wait after click (default: 1000)

Examples:

Click the .next-button
Use click_element to click #submit-btn and wait 2 seconds

scroll_page

Scroll the browser page or a specific element.

Parameter	Type	Required	Description
selector	string	❌	CSS selector of element to scroll (scrolls entire page if not specified)
direction	string	✅	Scroll direction: down, up, bottom, top
amount	number	❌	Pixels to scroll (only used with down/up)

selector Option (Important)

To scroll only the sidebar on Google Maps, Naver Maps, etc., you MUST use the selector option.

// Scroll entire page (default)
scroll_page direction: "down"

// Scroll only specific container (Google Maps sidebar)
scroll_page selector: "div[role='feed']" direction: "down"

Without specifying selector, the entire map scrolls and the list won’t load new items.

Examples:

Scroll to the bottom of the page
Use scroll_page to scroll div[role="feed"] down by 500 pixels

wait_for_element

Wait until a specific element appears.

Parameter	Type	Required	Description
selector	string	✅	CSS selector of element to wait for
timeout	number	❌	Maximum wait time in milliseconds (default: 10000)

Examples:

Wait until .loading disappears
Use wait_for_element to wait for .results to appear with 5 second timeout

navigate_to

Navigate the browser to a specific URL.

Parameter	Type	Required	Description
url	string	✅	URL to navigate to

Examples:

Go to Google Maps
Use navigate_to to go to https://www.google.com/maps

---

File Storage

save_to_csv

Save data to a CSV file.

Parameter	Type	Required	Description
filename	string	✅	Filename to save (e.g., leads.csv)
data	array	✅	Array of data to save. Each item must be an object
append	boolean	❌	If true, append to existing file; if false, overwrite (default: false)

Save Location: If no path is specified, files are saved to the Desktop.

Examples:

Save the extracted data to cafes.csv
Use save_to_csv to append data to results.csv (append: true)

save_to_json

Save data to a JSON file.

Parameter	Type	Required	Description
filename	string	✅	Filename to save (e.g., data.json)
data	object/array	✅	Data to save

Example:

Save the extracted data to results.json

---

Workflow Example

Scraping Data from Google Maps

1. navigate_to("https://www.google.com/maps/search/gangnam+cafe")
2. wait_for_element("[data-index]", timeout=10000)
3. Loop:
   a. extract_structured_data(containerSelector="[data-index]", fields={...})
   b. scroll_page(selector="div[role='feed']", direction="down", amount=500)
   c. wait_for_element("[data-index]:last-child")
4. save_to_csv("gangnam_cafes.csv", collected_data)

As a natural language request to Claude:

Scrape "Gangnam cafe" search results from Google Maps.
Scroll to collect 30 business listings and save to CSV.
Make sure to scroll only the sidebar (selector: div[role="feed"]).