Use with Claude

1. Locate the config file

Claude Desktop reads MCP servers from a single JSON file. Open it in your editor:

If the file doesn't exist yet, create it with an empty {} object.

2. Add the Crawlbase server

{
  "mcpServers": {
    "crawlbase": {
      "type": "stdio",
      "command": "npx",
      "args": ["@crawlbase/mcp@latest"],
      "env": {
        "CRAWLBASE_TOKEN": "YOUR_TOKEN",
        "CRAWLBASE_JS_TOKEN": "YOUR_JS_TOKEN"
      }
    }
  }
}

// Claude Code reads the same shape from claude.json
{
  "mcpServers": {
    "crawlbase": {
      "type": "stdio",
      "command": "npx",
      "args": ["@crawlbase/mcp@latest"],
      "env": {
        "CRAWLBASE_TOKEN": "YOUR_TOKEN",
        "CRAWLBASE_JS_TOKEN": "YOUR_JS_TOKEN"
      }
    }
  }
}

3. Restart Claude Desktop

Quit fully (not just close the window — use Cmd+Q on macOS or right-click the system tray on Windows) and relaunch. The next conversation will have access to the Crawlbase tools.

4. Verify it's working

Click the slider/tools icon in the chat input. You should see crawlbase listed with 9 tools available (3 crawl + 6 storage). Try a quick test:

Crawl https://anthropic.com and tell me what's in the navigation.

You'll see Claude call crawl (or crawl_markdown), then summarize the result. If you don't see the tool call, check the logs:

Example prompts

Things that just work once Crawlbase is connected:

• "Compare the pricing pages of Stripe, Adyen, and Square. Make a table of plans and headline rates."
• "Take screenshots of these 5 landing pages and tell me which has the strongest above-the-fold."
• "What are the top 3 results for 'open source LLM observability' right now? Summarize each."
• "Watch news.ycombinator.com and tell me when a post about my company hits the front page."

Troubleshooting