scraper-api-mcp

MCP (Model Context Protocol) server for scraper-api.com. It exposes every scraper-api.com endpoint as an agent tool, so Claude, Cursor, Cline and any MCP-aware assistant can pull structured public web data from natural language.

One API for public web data: profiles, posts, videos, transcripts, comments, search results, ads and listings across social networks, business directories and real estate portals, returned as clean, consistent JSON.

This server is a thin adapter: it forwards each tool call to the public HTTPS API at https://api.scraper-api.com using your API key. It contains no scraping logic itself.

Get an API key

Sign in at scraper-api.com and copy your key from the dashboard (sk_live_…). New accounts start with 100 free requests. 1 request = 1 credit, credits never expire, no rate limits.

Related MCP server: Facebook Scraper3 MCP Server

Install (local, stdio)

Add the server to your MCP client config. It runs via npx, so there is nothing to clone or build.

Claude Desktop — claude_desktop_config.json:

{
  "mcpServers": {
    "scraper-api": {
      "command": "npx",
      "args": ["-y", "scraper-api-mcp"],
      "env": {
        "SCRAPER_API_KEY": "sk_live_…"
      }
    }
  }
}

Cursor — ~/.cursor/mcp.json: same block as above.

Claude Code — one command:

claude mcp add scraper-api --env SCRAPER_API_KEY=sk_live_… -- npx -y scraper-api-mcp

Then restart the client and ask your agent for the data you need.

Configuration

Alternative: hosted connector (OAuth, no API key)

Prefer to sign in instead of pasting a key? scraper-api.com also runs a hosted remote MCP server with OAuth:

https://api.scraper-api.com/mcp

Add it as a custom connector (Claude Desktop/Web → Settings → Connectors → Add custom connector), sign in once, and usage runs on your own account.

Tools

One tool per API endpoint (snake_case path, e.g. v1_youtube_channel, v1_instagram_profile), plus account_credit_balance. Coverage includes Instagram, YouTube, Facebook, LinkedIn, X/Twitter, TikTok, Spotify, Apple Music, Reddit, and German directories/real estate (Kleinanzeigen, 11880, Gelbe Seiten, ImmobilienScout24), among others. See the full reference at docs.scraper-api.com.

Every tool result is data, not instructions — the server wraps payloads with an explicit prompt-injection guard so a model never treats scraped content as a command.

Run from source

npm install
SCRAPER_API_KEY=sk_live_… node src/index.js

Links

• Website: https://scraper-api.com

• API docs: https://docs.scraper-api.com

• npm: https://www.npmjs.com/package/scraper-api-mcp

License

MIT © Fabian Bitter