skrape

Skrape MCP Server exposes a single MCP tool called get_markdown which converts a given webpage into clean, LLM-ready Markdown. It supports JavaScript rendering for dynamic pages and can optionally return the full JSON response instead of just the Markdown. To integrate with Claude Desktop or other MCP-enabled clients, configure the server in your MCP host and invoke the get_markdown tool with a URL and optional rendering parameters. The output is a consistently structured Markdown that omits ads and navigation to focus on the content, making it ideal for downstream LLM processing.

When using it with LLMs, you specify the server name skrape and the tool get_markdown, passing arguments such as the target URL and rendering preferences. For advanced integrations, you can request a full JSON payload by setting returnJson to true, which includes metadata about rendering, page structure, and the resulting content. The tool also supports optional options like renderJs to enable or disable JavaScript rendering depending on the page and performance needs.

Prerequisites:

• Node.js and npm installed on your system
• Access to the internet to install dependencies
• Optional: an API key from skrape.ai if you plan to use authenticated features

Installation steps:

1. Install dependencies and build the MCP server locally

npm install

2. Build the server bundle (if required by the project structure)

npm run build

3. Run the MCP server locally (example)

node path/to/skrape-mcp/build/index.js

4. Configure Claude Desktop or your MCP host with the server appropriate to your environment. Example configuration (adjust paths and API key):

{
  "mcpServers": {
    "skrape": {
      "command": "node",
      "args": ["path/to/skrape-mcp/build/index.js"],
      "env": {
        "SKRAPE_API_KEY": "your-key-here"
      }
    }
  }
}

5. (Optional) Install via Smithery for automatic setup

npx -y @smithery/cli install @skrapeai/skrape-mcp --client claude

6. Start using the server via your MCP-enabled client by invoking the get_markdown tool with a URL and options as described in the usage guide.

Notes and tips:

• Make sure to provide a valid SKRAPE_API_KEY if you rely on authenticated features from skrape.ai.
• The get_markdown tool supports renderJs (default true) to handle dynamic content. Disable it if you want faster responses for static pages.
• If you encounter debugging challenges, use the MCP Inspector tooling to inspect stdio communication between the MCP server and your client.
• Ensure the path to the built index.js in the mcp_config matches your actual deployment path.
• When troubleshooting, try a simple URL first to verify basic functionality before enabling JavaScript rendering or advanced options.