@perpage/mcp-server

Give your AI agent clean, LLM-ready web access. An MCP server that exposes PerPage as native tools — your agent can scrape any URL into clean markdown and extract structured JSON, no HTML parsing or scraping infra.

Tools

| Tool | What it does | |---|---| | scrape_url | Fetch a page → clean, LLM-ready markdown with token counts (optional RAG chunks). | | extract_structured_data | Page + your JSON Schema → validated structured JSON. |

Get a key

Subscribe to PerPage on RapidAPI (free plan available) and copy your RapidAPI key: https://rapidapi.com/infospectrum-inc-infospectrum-inc-default/api/perpage

Use with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "perpage": {
      "command": "npx",
      "args": ["-y", "@perpage/mcp-server"],
      "env": { "PERPAGE_API_KEY": "YOUR_RAPIDAPI_KEY" }
    }
  }
}

Works the same in any MCP client (Cursor, Cline, Claude Code, etc.) — it's a stdio server.

Configuration

| Env var | Required | Default | Notes | |---|---|---|---| | PERPAGE_API_KEY | ✅ | — | Your RapidAPI key (or a direct PerPage key). | | PERPAGE_API_URL | — | https://perpage.p.rapidapi.com | Set to https://api.perpage.dev for direct access. |

Auth is routed automatically: a RapidAPI host uses X-RapidAPI-Key; a direct origin uses X-API-Key. So the same PERPAGE_API_KEY works for both channels.

Example

"Scrape https://example.com and summarize it." "Extract the price and title from this product page as JSON."

The agent calls scrape_url / extract_structured_data behind the scenes and gets back clean content — one request = one page, no credit multipliers.

Links

• API & docs: https://perpage.dev
• RapidAPI listing: https://rapidapi.com/infospectrum-inc-infospectrum-inc-default/api/perpage
• Crawler policy / contact: https://perpage.dev/bot

© InfoSpectrum Inc.