Scrapfly MCP README

Scrapfly MCP Server

Give your AI real-time access to any website

🌐 Landing Page • 📖 Documentation • 🎮 Live Demo • 🔑 Get API Key

At a glance

Add it to any MCP client:

{
  "mcpServers": {
    "scrapfly": { "type": "http", "url": "https://mcp.scrapfly.io/mcp" }
  }
}

Tools that fetch pages need no setup beyond the key: web_get_page (quick fetch), web_scrape (full control: JS rendering, proxy pool, country, extraction) and screenshot.

What is Scrapfly MCP?

The Scrapfly MCP Server connects your AI assistants to live web data through the Model Context Protocol. Transform your AI from being limited by training data to having real-time access to any website.

✨ What Your AI Can Do

🏆 Why Scrapfly?

Built on battle-tested infrastructure used by thousands of developers:

• 99.9% Uptime — Enterprise-grade reliability
• 100+ Countries — Global proxy network with datacenter & residential IPs, in three proxy pools (Basic, Unblocker, Unblocker Ultra) or Auto, where the API picks the pool
• Anti-Bot Bypass — Advanced Unblocker technology defeats modern protections
• OAuth2 Security — Enterprise authentication for production deployments

📖 Learn more: Why Scrapfly MCP?

🚀 Quick Install

Click one of the buttons below to install the MCP server in your preferred IDE:

📦 Manual Installation

Standard Configuration

Works with most MCP-compatible tools:

{
  "servers": {
    "scrapfly-cloud-mcp": {
      "type": "http",
      "url": "https://mcp.scrapfly.io/mcp"
    }
  }
}

Cloud Configuration (NPX)

For tools that require a local process:

{
  "mcpServers": {
    "scrapfly": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.scrapfly.io/mcp"
      ]
    }
  }
}

🔧 IDE-Specific Setup

code --add-mcp '{"name":"scrapfly-cloud-mcp","type":"http","url":"https://mcp.scrapfly.io/mcp"}'

code-insiders --add-mcp '{"name":"scrapfly-cloud-mcp","type":"http","url":"https://mcp.scrapfly.io/mcp"}'

claude mcp add scrapfly-cloud-mcp --url https://mcp.scrapfly.io/mcp

{
  "mcpServers": {
    "scrapfly": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.scrapfly.io/mcp"]
    }
  }
}

{
  "scrapfly-cloud-mcp": {
    "type": "http",
    "url": "https://mcp.scrapfly.io/mcp"
  }
}

{
  "context_servers": {
    "scrapfly-cloud-mcp": {
      "type": "http",
      "url": "https://mcp.scrapfly.io/mcp"
    }
  }
}

[mcp_servers.scrapfly-cloud-mcp]
url = "https://mcp.scrapfly.io/mcp"

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "scrapfly-cloud-mcp": {
      "type": "http",
      "url": "https://mcp.scrapfly.io/mcp",
      "enabled": true
    }
  }
}

🛠️ Available Tools

The Scrapfly MCP Server exposes 62 tools in seven groups. Most tasks need only the first group.

Web scraping (5)

Stateless, one call per page.

Account (2)

Cloud Browser sessions (9)

A stateful real browser, kept between calls.

Page interaction (17)

Act on the open Cloud Browser session. Elements are addressed by the uid that take_snapshot returns.

take_snapshot, take_screenshot, find_elements, get_page_url, evaluate_script, click, hover, drag, fill, type_text, press_key, select_option, dropdown_options, scroll, scroll_to_text, go_back, wait

WebMCP (2)

For pages that register their own tools with navigator.modelContext.

list_webmcp_tools, call_webmcp_tool

Credential vaults (15)

Store logins for Cloud Browser sessions, optionally mirrored from 1Password.

cloud_browser_vault_create, cloud_browser_vault_list, cloud_browser_vault_get, cloud_browser_vault_update, cloud_browser_vault_delete, cloud_browser_vault_rotate, cloud_browser_vault_item_create, cloud_browser_vault_item_list, cloud_browser_vault_item_update, cloud_browser_vault_item_delete, cloud_browser_vault_service_link, cloud_browser_vault_service_update, cloud_browser_vault_service_sync, cloud_browser_vault_service_test, cloud_browser_vault_service_unlink

Alerts (12)

Create, test and snooze alert rules on your account metrics. Writes take a confirm step.

alert_metric_families, alert_preview, alert_create, alert_update, alert_delete, alert_get, alert_list, alert_count_active, alert_series, alert_test, alert_snooze, alert_unsnooze

📖 Full reference: Tools & API Specification

Proxy pools

web_scrape and web_get_page take a proxy_pool, which sets both the proxies and the Unblocker (anti-bot bypass):

Pick the pool a site needs: basic where datacenter proxies get through, unblocker or unblocker_ultra where it needs residential proxies and the Unblocker, auto to let the API pick. A target or an anti-bot protection with a price of its own refuses basic and unblocker with ERR::PROXY::UNBLOCKER_ULTRA_REQUIRED (HTTP 422, not billed): resend with proxy_pool=unblocker_ultra, or with proxy_pool=auto, which is accepted on every target, the ones that require unblocker_ultra included, where their price is added. public_tor_pool runs on the Tor network (.onion).

Deprecated: unblocker and asp

The unblocker input of web_scrape and its former name asp are deprecated: use proxy_pool (unblocker_ultra for the Unblocker). The migration guide covers requests written before the proxy pools, the pool names public_datacenter_pool and public_residential_pool included. Without a proxy_pool, the server sends the API no pool and, by default, asp=true, a request the guide covers as well.

Example: Scrape a Page

User: "What are the top posts on Hacker News right now?"

AI: Uses web_get_page to fetch https://news.ycombinator.com and returns current top stories

Example: Extract Structured Data

User: "Get all product prices from this Amazon page"

AI: Uses web_scrape with extraction_model="product_listing" to return structured JSON

📖 More examples: Real-World Examples

🔐 Authentication

Scrapfly MCP supports multiple authentication methods:

🔑 Get your API key: Scrapfly Dashboard

📊 Configuration Reference

🖥️ Self-Hosted / Local Deployment

You can run the Scrapfly MCP server locally or self-host it.

CLI Arguments

Environment Variables

Per-request key: start the server without a baked-in key (-http only) and each caller supplies their own key per request — as Authorization: Bearer <key>, or ?key=<key> / ?apiKey=<key>. This path is authenticated; the server-key path above is not, which is why it stays on loopback by default.

Examples

# Start HTTP server on port 8080
./scrapfly-mcp -http :8080

# Start HTTP server using PORT env var
PORT=8080 ./scrapfly-mcp

# Start with an API key. Prefer the environment variable: a key passed as a
# flag lands in your shell history and in the process list.
# Server-key mode has no per-caller auth, so it binds 127.0.0.1 here.
SCRAPFLY_API_KEY=... ./scrapfly-mcp -http :8080

# Same, but expose it on all interfaces — only behind your own auth proxy.
SCRAPFLY_API_KEY=... ./scrapfly-mcp -http 0.0.0.0:8080 -allow-remote

# Start in stdio mode (for local MCP clients)
./scrapfly-mcp

Docker

# Build
docker build -t scrapfly-mcp .

# Run (Smithery compatible - uses PORT env var)
docker run -p 8080:8080 scrapfly-mcp

# Run with custom port
docker run -e PORT=9000 -p 9000:9000 scrapfly-mcp

🤝 Framework Integrations

Scrapfly MCP also works with AI frameworks and automation tools:

📖 All integrations: Integration Index

📚 Resources

• 🌐 MCP Cloud Landing Page — Product overview & features
• 🎮 Live n8n Demo — Try it in your browser
• 📖 Full Documentation
• 🛠️ Tools Reference
• 💡 Examples & Use Cases
• ❓ FAQ
• 🔐 Authentication Guide

💬 Need Help?

• 📖 Scrapfly Documentation
• 📧 Contact Support
• 🌐 Model Context Protocol

Made with ❤️ by Scrapfly
_{The Web Scraping API for Developers}