Content
# Claude MCP Web Scraping — Give Claude Live Web Access
Connect Claude (Desktop, Code, or any MCP client) to a **web scraping MCP server** so it can fetch pages, run Google searches and extract structured data on its own — no copy-pasting HTML into the chat.
## Claude Desktop
Add to `claude_desktop_config.json` (Settings → Developer → Edit Config):
```json
{
"mcpServers": {
"quantumproxies": {
"command": "npx",
"args": ["-y", "quantumproxies-mcp"],
"env": {
"QUANTUMPROXIES_API_KEY": "qp_live_your_key_here"
}
}
}
}
```
Restart Claude Desktop — done.
## Claude Code (CLI)
```bash
claude mcp add quantumproxies \
-e QUANTUMPROXIES_API_KEY=qp_live_your_key_here \
-- npx -y quantumproxies-mcp
```
## What Claude can do once connected
| Tool | What it does |
| --- | --- |
| `scrape` | any URL → clean markdown (JS rendering + anti-bot handled) |
| `search` | live Google/Bing results as structured JSON |
| `crawl` / `map` | walk a whole site, or list its URLs |
| `batch` | scrape many URLs in one job |
Example prompts that just work afterwards:
> "Read https://competitor.com/pricing and summarize the tiers in a table."
>
> "Search for 'best CRM for startups', open the top 3 results, and compare their positioning."
## Why route it through a scraping backend
Claude's built-in fetch gets blocked by anything with bot protection and can't render SPAs. The MCP server proxies every request through a residential network with rendering and retries — so "read this page" works on real-world sites, not just blogs.
Server details & tool list: [quantumproxies.io/mcp-server](https://quantumproxies.io/mcp-server) · npm: [`quantumproxies-mcp`](https://www.npmjs.com/package/quantumproxies-mcp) · API key: [app.quantumproxies.io/api-keys](https://app.quantumproxies.io/api-keys).
