Scrapfly MCP

What scrapfly/scrapfly-mcp does

scrapfly/scrapfly-mcp gives AI agents access to live websites through the Model Context Protocol and Scrapfly’s web data infrastructure. It is suited to tasks such as collecting current listings, prices, news, or other page content; turning web pages into structured JSON; and taking visual snapshots for analysis.

The server exposes 62 tools across several areas: web scraping, account operations, Cloud Browser sessions, page interaction, WebMCP, credential vaults, and alerts. The primary page-fetching tools are web_get_page, web_scrape, and screenshot.

How it works

The hosted endpoint is available at https://mcp.scrapfly.io/mcp using streamable HTTP, so clients that support HTTP-based MCP connections can connect without installing a local server. Clients limited to local processes can use npx mcp-remote to bridge to the same endpoint.

For teams that want to run the service themselves, the repository supports a Go binary over stdio or HTTP, as well as a Docker image. Requests that use the web-fetching tools are handled by Scrapfly and can use JavaScript rendering, proxy pools, country selection, and extraction settings. Scrapfly also provides CAPTCHA, proxy, rate-limit, and anti-bot handling as part of its web access infrastructure.

Setup and configuration

A hosted configuration can point an MCP client directly at the endpoint:

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

Authentication can use OAuth2 sign-in on the hosted server or a Scrapfly API key. For API-key authentication, set SCRAPFLY_API_KEY, or send the key as an Authorization: Bearer <key> header. With clients that require stdio, configure npx mcp-remote https://mcp.scrapfly.io/mcp as the command and argument pair.

The server has documented setup paths for VS Code, Cursor, Claude Code, Claude Desktop, Cline, and Visual Studio. VS Code and Cursor also provide installation links, while Claude Code can register the endpoint with its CLI.

Tools and capabilities

• Fetch a page quickly with web_get_page.
• Scrape with controls for JavaScript rendering, proxy pools, country selection, and extraction through web_scrape.
• Capture a page or page element with screenshot.
• Extract website content into structured JSON using AI-powered extraction.
• Interact with pages through the page-interaction and Cloud Browser tool groups.
• Manage account functions, WebMCP operations, credential vaults, and alerts.

Limitations and notes

The hosted option depends on the Scrapfly service and requires authentication. The server provides access to websites through Scrapfly rather than directly exposing a local browser session. Use the remote HTTP configuration where the MCP client supports it; use mcp-remote for stdio-only clients. Self-hosting is available through the Go binary or Docker image, with deployment mode chosen as stdio or HTTP.