Skip to main content
Scrapling provides a Model Context Protocol (MCP) server that enables AI assistants to perform web scraping operations. The server exposes Scrapling’s fetching capabilities as MCP tools.

​
Overview

The Scrapling MCP server allows AI models to:
  • Fetch web pages with stealth HTTP requests
  • Extract content using CSS selectors
  • Handle JavaScript-heavy sites with browser automation
  • Bypass Cloudflare and other protections
  • Process single or multiple URLs concurrently
  • Return structured data in Markdown, HTML, or text format

​
Quick Start

​
Start the Server

scrapling mcp
This starts the server in stdio transport mode (default).

​
HTTP Transport

For network access, use streamable-http transport: 
scrapling mcp --http --host 0.0.0.0 --port 8000

​
Server Configuration

The MCP server configuration is defined in server.json: 
{
  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
  "name": "io.github.D4Vinci/Scrapling",
  "title": "Scrapling MCP Server",
  "description": "Web scraping with stealth HTTP, real browsers, and Cloudflare bypass. CSS selectors supported.",
  "version": "0.4.1"
}

​
Available Tools

The MCP server exposes six main tools:

get

Make HTTP GET requests with stealth features

bulk_get

Fetch multiple URLs concurrently with GET requests

fetch

Use Playwright for JavaScript-heavy sites

bulk_fetch

Fetch multiple URLs with browser automation

stealthy_fetch

Advanced stealth with Cloudflare bypass

bulk_stealthy_fetch

Stealthy fetching for multiple URLs

​
Transport Modes

​
Stdio Transport (Default)

Used for local AI assistants: 
scrapling mcp
  • Standard input/output communication
  • Best for Claude Desktop, Cline, and other local clients
  • No network configuration needed

​
HTTP Transport

Used for network-accessible services: 
scrapling mcp --http --host 0.0.0.0 --port 8000
  • Streamable HTTP transport
  • Accessible over network
  • Configurable host and port

​
Response Format

All tools return a structured ResponseModel: 
{
  "status": 200,
  "content": ["Extracted content..."],
  "url": "https://example.com"
}
​
status
integer
HTTP status code from the website
​
content
array[string]
Extracted content as Markdown, HTML, or text
​
url
string
The URL that was requested

​
Installation Methods

uvx scrapling mcp
Or with pip: 
pip install scrapling
scrapling mcp

​
Docker (OCI)

docker pull ghcr.io/d4vinci/scrapling
docker run ghcr.io/d4vinci/scrapling mcp

​
Use Cases

Enable AI assistants to gather information from websites, extract articles, and process web content.
Let AI models collect data from multiple sources concurrently with bulk operations.
Extract and analyze web content using CSS selectors for targeted data extraction.
Use stealthy fetching to access Cloudflare-protected sites through AI workflows.

​
Next Steps

Setup Guide

Configure MCP server for your AI client

Capabilities

Explore available MCP tools and features

Build docs developers (and LLMs) love

Get started for freeTalk to us