Scrapling MCP Server: Setup for Claude Code, Cursor and Other Apps

• Downloads a web page's text quickly, with no JavaScript, for simple sites

  Downloads a page without opening a browser or running its JavaScript. Use it for plain sites that do not block scrapers. You can set the HTTP method, headers, cookies and a CSS selector to keep only part of the page.

  Try: “Pull the plain text of the install page at docs.example.com/install.”

  Network · make_request

• Opens a page in a hidden browser, lets it load, and returns its content

  Loads a page in a stealth browser, runs its JavaScript so the content appears, and returns what it finds. Built for sites that only show content after scripts run, and for some pages behind Cloudflare checks. This one-off call does not keep the browser open for later requests.

  Try: “Read the job listings on this careers page, which only load with JavaScript.”

  Runs code · browser_fetch_once

• Reads the page you have open in a browser as text, Markdown or HTML

  Pulls the content of the page already open in a browser session, or one element on it, without reloading. Output can be text, Markdown, HTML or a snapshot listing the page's elements. A search pattern can find matching lines in the snapshot.

  Try: “Copy the price table on the page I have open into Markdown.”

  Read · browser_extract

• Takes a picture of the open page or one section of it that the assistant can view

  Captures the open page as an image without reloading it. You can take the visible area, the full page, or a single element, as PNG or JPEG. Useful for checking filled form fields, open menus, charts or result cards.

  Try: “Show me a screenshot of the checkout form with the filled fields visible.”

  Read · browser_screenshot

• Loads one page in a browser you opened earlier, reusing that browser's settings

  Loads a single URL in a browser session you already opened, running its JavaScript first. The options you set for this request apply only to this request. It can also return a snapshot of the page's controls and their positions.

  Try: “Open the account page in my saved browser and tell me the plan name.”

  Runs code · browser_fetch

• Fetches one page without a browser, reusing the cookies of an open HTTP session

  Sends a request through an HTTP session you opened earlier, so cookies, connections and the browser fingerprint carry over between requests. It runs no JavaScript. Options you set apply only to this one request.

  Try: “Fetch the next results page with the same cookies I used on the first one.”

  Network · session_make_request

• Runs a list of steps in the open browser: clicks, typing, key presses and waits

  Runs a list of steps in order inside the open browser, such as clicks, filling fields, pressing keys, waiting, and answering alerts or prompts. It stops at the first error and names the step that failed. Steps that ran before the error are not undone.

  Try: “Search for 'invoice' on the site, press Enter, and open the first result.”

  Runs code · browser_actions

• Runs your own JavaScript on the open page and returns the result as JSON

  Runs a JavaScript snippet on the page currently open in a browser, waits for any promises, and returns the result as JSON. Useful for counting elements, reading page state, or collecting values from many elements at once. Script effects are not undone if the script fails.

  Try: “Count how many product cards are on this page and list their titles.”

  Runs code · browser_evaluate

• Lists the background requests a page made, to find the data calls behind it

  Lists requests the browser completed since the session opened, across tabs and page loads. Use it to find the API calls behind a page's data or to spot HTTP errors. Static files such as images are hidden unless you ask for them, and long lists are paged.

  Try: “Show the API calls this dashboard page makes when it loads.”

  Read · browser_network_requests

• Shows the headers, body or details of one recorded request from that list

  Reads one request from the recorded list and returns the part you choose: summary, headers, sent data or the text response body. It does not send the request again. Very large or non-text bodies come back with a note saying why they are missing.

  Try: “Show the response body of the third API call on this page.”

  Read · browser_network_request

• Opens a browser window that stays open for several page loads

  Starts a stealth browser that stays open across several page loads, so you do not launch a new browser each time. The browser-level settings are set here, and each page is requested separately.

  Try: “Open a browser I can reuse for several pages on this site.”

  Runs code · browser_open

• Opens a reusable HTTP connection that keeps cookies between page requests

  Starts a plain HTTP session with no browser. It keeps cookies, connections and the chosen browser fingerprint across requests sent through it.

  Try: “Start an HTTP session that keeps my cookies while I fetch several pages.”

  open_request_session

• Saves the browser's logins and cookies to a file so you can reuse them later

  Writes the open browser's cookies, local storage and IndexedDB to a JSON file, so a login can be restored later. Session storage is not included. If the file already exists, it is replaced.

  Try: “Save my logged-in browser state to a file so I can reuse it tomorrow.”

  Files · browser_save_state

• Loads a saved login file into an open browser so it can use that login

  Reads a saved state file into an open browser session, replacing its current cookies and storage. Open tabs stay as they are. Reload or fetch the page afterward to use the restored login.

  Try: “Load my saved login file into the open browser before I fetch the dashboard.”

  Files · browser_load_state

• Used by the assistant behind the scenes

  Closes a browser or HTTP session you no longer need and frees its resources

  Closes a browser or HTTP session that is still open and releases the resources it holds.

  Try: “Close the browser session I opened for this scrape.”

  close_session

• Lists the sessions currently open and the settings each one uses

  Shows every open browser or HTTP session with its status and effective settings. The assistant uses it to check which sessions exist before it fetches or closes one.

  Try: “Which browser sessions are still open, and what settings do they use?”

  list_sessions

Watch out for

• Reaches any URL your machine can reach

  Fetch tools take any URL, including local network addresses. Redirects to private IPs are blocked by default, direct requests are not. github.com

• The browser can click, fill forms and run JavaScript

  Browser tools act on live pages, so a page you open can have effects. Only point it at sites you trust with your logins. github.com

• Reported: file:// links can read local files

  An open pull request says the fetchers return local files for file:// addresses. The maintainer has not responded yet, so keep sensitive files out of reach. github.com

• Network mode lets anyone with the token use it

  In network mode every client shares one token, and it travels unencrypted without TLS. Keep it on localhost or put TLS in front. scrapling.readthedocs.io

• Reported: stealth browser accepts invalid certificates

  An open discussion says the stealth browser ignores HTTPS certificate errors by default. The plain HTTP tools do check them. github.com

• Writes login state files to disk

  Saving browser state writes cookies and storage to a path on your machine and overwrites an existing file. Treat those files like passwords. github.com

• Hidden page content is stripped by default

  Hidden elements, comments and zero-width characters are removed from scraped text to reduce prompt injection. Turning off main-content mode keeps them. scrapling.readthedocs.io

Security advisories

In their words