# fouradata/mcp [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/fouradata/mcp  
**GitHub Stars:** 4  
**npm Downloads (last month):** 798  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/fouradata-mcp

## Description
MCP server for the FourA web scraping API - four tools, six prompts, one API key.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "mcp": {
    "command": "npx",
    "args": ["-y","@fouradata/mcp"],
    "env": {
      "FOURA_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `FOURA_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What fouradata/mcp MCP server does

The fouradata/mcp MCP server gives MCP clients access to FourA's web scraping API. An agent supplies a public URL, and FourA retrieves the response using a suitable access method. The server is intended for reading web content rather than modifying remote systems: all four tools are marked read-only and open-world in their MCP annotations.

FourA supports ordinary HTTP requests, proxy rotation, and browser-based rendering. This makes the server suitable for static pages, JSON endpoints, server-rendered HTML, pages that apply bot checks, and pages whose content depends on browser execution. The README also describes ready-made prompts for product extraction, article cleanup, price monitoring, endpoint checks, and bulk URL retrieval.

## How it works

The default `foura_auto` tool accepts a URL and chooses the retrieval method. Optional validation rules can require specific content to appear, helping distinguish a target page from a challenge page. Its result includes status, headers, data, metadata about the selected method and credit usage, and—by default—a session containing proxy, cookie, and user-agent information for later requests.

The other tools provide more direct control. `foura_single` performs one HTTP request and supports request bodies, custom headers, redirects, response parsing, validation, and binary data. `foura_proxy` handles proxy-based retrieval, while `foura_browser` provides browser rendering. A proxy identifier returned from an earlier request can be passed to browser retrieval when a page needs JavaScript after proxy access.

Responses expose both text content and structured JSON validated against each tool's output schema. This gives clients a predictable contract instead of requiring agents to parse prose responses.

## Setup and configuration

The fouradata/mcp MCP server can run locally through npm as an MCP subprocess. Install it with `npx -y @fouradata/mcp` and provide the `FOURA_API_KEY` environment variable. The README shows the same stdio configuration pattern for Claude Desktop, Cursor, Windsurf, Claude Code, and other MCP clients. The key can be obtained from the FourA dashboard; examples use the `pk_live_...` format.

Clients that support Streamable HTTP can instead connect to `https://mcp.foura.ai/mcp` and send the key as a Bearer token in the `Authorization` header. The local and hosted options expose the same FourA API functionality, but the hosted setup requires the client to support that transport.

## Tools and capabilities

- `foura_auto` selects the access method for a URL.
- `foura_single` sends a configurable HTTP request.
- `foura_proxy` retrieves content through proxy rotation.
- `foura_browser` renders pages in a browser session.
- Browser and operating-system profiles can be selected for supported request tools.
- Timeout, validation, headers, request data, redirects, parsing, and session behavior can be configured where supported.

## Limitations and notes

A FourA API key is required for both local and hosted use. The server does not remove the need to choose an escalation path for every difficult target: a blocked direct request may require proxy retrieval, and JavaScript-dependent pages may require browser rendering. Proxy attempts, browser profiles, timeouts, and validation options affect how requests are completed and what is returned.

The README specifically documents Claude Desktop as requiring a full application quit before editing its configuration, because an active instance can overwrite changes. Hosted use also depends on MCP client support for Streamable HTTP.

_Full upstream README: https://allmcps.com/mcp/fouradata-mcp/readme_