Skip to content
Meta Ad Library scraper is live, spy on any brand's adsExplore
MCP Server is live

MCP Integration

Wire Claude Desktop, Cursor, Continue.dev, OpenAI, or any MCP-aware agent to ArrowCrawl in one command.

One-line install

Configures Claude Desktop, Cursor, and Continue.dev in one shot. Backs up any existing config first.

npx arrowcrawl-mcp install

Non-interactive: npx arrowcrawl-mcp install --key sg_live_... --clients claude,cursor

1. Get an API key

Sign up at /signup and create a live API key in your dashboard. Free tier is 1,000 credits/month, no card.

2. Manual configuration (if you prefer)

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "arrowcrawl": {
      "url": "https://scrapegod-backend-production.up.railway.app/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer sg_live_YOUR_KEY"
      }
    }
  }
}

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "arrowcrawl": {
      "url": "https://scrapegod-backend-production.up.railway.app/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer sg_live_YOUR_KEY"
      }
    }
  }
}

Continue.dev

Add to ~/.continue/config.json (same JSON shape as above).

3. Available tools

arrowcrawl_scrape1 credit

Scrape any URL → markdown, html, json, or links.

Arguments: url, formats?

arrowcrawl_extract5 credits

AI-powered structured extraction. Provide a JSON schema or free-text prompt.

Arguments: url, schema?, prompt?

arrowcrawl_marketplace1 credit per ID

Batch-scrape Amazon (IN/US), Flipkart, Myntra, or Nykaa. Rich mode returns every image, full description, specifications, variants with stock, and top reviews. same as a console-paste script.

Arguments: marketplace, product_ids[], mode?, concurrency?

arrowcrawl_marketplaces_listFree

List supported marketplaces and their ID types.

Arguments: (none)

arrowcrawl_wikidata_enrichFree

Enrich a marketplace product with Wikidata structured data: GTIN/UPC/ISBN, brand, manufacturer, country of origin. Uses CC0-licensed data.

Arguments: marketplace, product_id

arrowcrawl_heal2 credits

Self-healing CSS selectors. Pass a URL + the fields you want; cached selectors are tried first, then the LLM rediscovers any field that returned empty.

Arguments: url, fields[], cached_selectors?

4. Try it

Once installed, ask Claude / Cursor:

Use arrowcrawl_scrape on https://example.com and return the markdown.

Use arrowcrawl_marketplace to fetch product data for ASIN
B0BGT6Q343 on amazon_in. Mode: rich.

Use arrowcrawl_extract on https://news.ycombinator.com with 
schema {"top_stories": [{"title": "string", "points": "number"}]}.

Use arrowcrawl_heal on https://shop.example.com/p/123 with 
fields ["title", "price", "image"].

5. Pricing

1 credit = 1 request. No hidden multipliers like Firecrawl. AI extraction costs 5 credits because the LLM call is real work. not because we want to inflate your bill. See the head-to-head with Firecrawl.

Direct API

The MCP endpoint speaks JSON-RPC 2.0. List tools:

curl -X POST https://scrapegod-backend-production.up.railway.app/api/v1/mcp \
  -H "Authorization: Bearer sg_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Legacy scrapegod_* tool names are aliased to arrowcrawl_* for backward compatibility.