🔌

Model Context Protocol

Official

Use DreamScrape as a tool inside Claude Code, Cursor, Claude Desktop, and any MCP-compatible AI agent. Your agent can scrape any URL and the router picks the optimal engine automatically.

What this does

•Lets AI agents call DreamScrape as a tool — scrape any URL, extract data, take screenshots
•The smart router picks the cheapest engine automatically — your agent doesn't need to know about engine tiers
•Results come back as clean markdown ready for the agent's context window

Install

Add the DreamScrape MCP server to your tool's configuration file.

Claude Code

Add to .mcp.json in your project root:

{
  "mcpServers": {
    "dreamscrape": {
      "command": "npx",
      "args": ["tsx", "node_modules/@dreamscrape/mcp-server/src/index.ts"],
      "env": {
        "DREAMSCRAPE_API_KEY": "ds_pk_YOUR_KEY_HERE",
        "DREAMSCRAPE_API_URL": "https://api.dreamscrape.app"
      }
    }
  }
}

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "dreamscrape": {
      "command": "npx",
      "args": ["tsx", "@dreamscrape/mcp-server"],
      "env": {
        "DREAMSCRAPE_API_KEY": "ds_pk_YOUR_KEY_HERE",
        "DREAMSCRAPE_API_URL": "https://api.dreamscrape.app"
      }
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "mcpServers": {
    "dreamscrape": {
      "command": "npx",
      "args": ["tsx", "@dreamscrape/mcp-server"],
      "env": {
        "DREAMSCRAPE_API_KEY": "ds_pk_YOUR_KEY_HERE"
      }
    }
  }
}

Available tools

Your agent gets access to these tools once the MCP server is connected:

dreamscrape_scrape

Scrape a URL and return clean markdown + metadata

Input: { url, renderMode?, engine?, timeout? }

dreamscrape_extract

Extract structured data using CSS selectors or LLM

Input: { url, fields[] }

dreamscrape_auto_extract

AI-powered schema detection — extracts all structured data

Input: { url, hint? }

dreamscrape_screenshot

Take a screenshot of a web page

Input: { url }

dreamscrape_crawl

Multi-page crawl starting from one or more URLs

Input: { startUrls[], maxPages?, maxDepth? }

dreamscrape_crawl_status

Check crawl progress and get page results

Input: { crawlId }

dreamscrape_sports

Query live sports data (ATP, NBA, MLB, CS2, odds)

Input: { query }

dreamscrape_crypto

Query live crypto data (prices, DEX, DeFi, memecoins)

Input: { query }

dreamscrape_health

Check API status and engine availability

Input: {}

Example prompts

Try these with your MCP-connected agent:

“Scrape the top story on Hacker News and summarize it in 3 bullet points.”

“Get the current NBA standings from basketball-reference.com and tell me who's leading the Eastern Conference.”

“Scrape github.com/trending and list the top 5 repos with their star counts.”

“Check if www.zillow.com is scrapeable and tell me what engine tier it needs.”

“Get the current Bitcoin price from CoinGecko's API.”

Troubleshooting

Tools not showing up

Make sure DREAMSCRAPE_API_KEY is set correctly in the MCP config. Restart your editor after changing the config.

401 Unauthorized errors

Your API key may be expired or revoked. Get a new key at dreamscrape.app/docs.

Scrapes timing out

The default timeout is 30 seconds. For heavy pages, pass a longer timeout in the tool input or try engine: "camoufox".

Questions? Email us — we reply fast.