MCP Scorecard

Mission StatementGitHub
← All posts

Scrapling: A 19,000-Star Scraper Becomes the Registry's Most Trusted Server

A mature Python web scraping framework with 19,000 stars adds MCP support and immediately claims the top trust score in the registry. This is what it looks like when established projects decide MCP is worth supporting.
io.github.D4Vinci

Most servers enter the MCP registry with zero stars and a template description. io.github.D4Vinci/Scrapling arrived with 19,392 stars, 1,300 forks, and years of active development. It scored 92 — the highest trust score of any new entry we have recorded — and became the first server to enter the top 25 at position #1.

What Scrapling Actually Is

Scrapling is not an MCP-native project. It is a full-featured Python web scraping framework that handles everything from single HTTP requests to large-scale concurrent crawls. Built by D4Vinci, a developer with a security and scraping background, it has been under active development well before MCP existed.

The framework offers three fetcher tiers, each designed for different levels of target resistance:

  • Fetcher — plain HTTP with TLS fingerprint spoofing and HTTP/3 support. Fast, lightweight, works against most sites.
  • DynamicFetcher — Playwright/Chromium for JavaScript-rendered pages. Handles SPAs and dynamic content.
  • StealthyFetcher — full anti-bot bypass with automatic Cloudflare Turnstile solving. The heavy artillery.

Beyond the fetchers, Scrapling includes adaptive element relocation — a system that tracks DOM elements even when website structures change between visits, using similarity algorithms rather than brittle CSS selectors. A Scrapy-like spider framework supports async parse callbacks, per-domain throttling, pause/resume checkpoints, and streaming mode. The project claims 10x faster JSON serialization than stdlib and 92% test coverage. BSD-3-Clause license.

The MCP Integration

The MCP layer sits on top of the scraping engine. Install with pip install "scrapling[ai]" and the server pre-extracts and structures web content before passing it to the AI assistant. Instead of dumping raw HTML into Claude's context window, Scrapling parses, filters, and formats the data first — reducing token consumption significantly.

This is the interesting pattern: the MCP server is not the product. It is a distribution channel for an already-proven tool. Scrapling's value was established long before MCP — what MCP adds is discoverability by AI agents and a standardized interface for tool invocation.

Why This Matters

The MCP registry has 2,822 servers. Most are small, single-developer projects built specifically for the protocol. Scrapling represents a different category: established tools adding MCP as an interface. When a project with 19,000 stars decides the protocol is worth supporting, it sends a signal to every other library maintainer watching from the sidelines.

This is the pattern that drives protocol adoption. Not greenfield projects — those will always appear. The inflection point comes when existing tools with existing users decide the integration is worth the effort. Scrapling is the most visible example yet.

Score: 92. No flags. BSD-3-Clause.

Sources: D4Vinci — GitHub · Scrapling — repo · Scorecard: io.github.D4Vinci (score 92)

← Game Engines Get API Servers: Unity and Unreal in Sub-Millisecond LookupsThe EU AI Act and MCP: What Tool Developers Need to Know Before August →