scrapi

Scrapi's MCP server converts web URLs into clean, LLM-friendly Markdown or plain text. It preserves structure such as headings, lists, and links, and can also provide metadata like title, author, date, and images. Two transport modes are supported: Stdio via npx for environments that can auto-run Node packages, and a Streamable HTTP transport for client applications that do not require a local Node.js installation. Once configured, you can send a URL to the MCP endpoint and receive a ready-to-use extract suitable for summarization or embedding in prompts. The server includes built-in features like JavaScript rendering for SPA pages, automatic retry on 429 responses, and basic billing/usage analytics tied to MCP keys.

Prerequisites:\n- Node.js 20+ (for building or running from source) or a runtime environment capable of executing npx commands.\n- An Scrapi MCP API key if you plan to use the HTTP/streamable transport or API-backed features.\n\nOption A: Using npx (Recommended)\n1. Ensure Node.js is installed.\n2. Use the MCP client configuration to point at the SCRAPI MCP server. A typical configuration is shown below.\n\njson\n{\n "mcpServers": {\n "scrapi": {\n "command": "npx",\n "args": ["-y", "@scrapi.ai/mcp-server"],\n "env": {\n "SCRAPI_API_KEY": "your-api-key"\n }\n }\n }\n}\n\n\nOption B: Install from Source\n1. Clone the repository and navigate into it.\n2. Install dependencies and build.\n\nbash\ngit clone https://github.com/bamchi/scrapi-mcp-server.git\ncd scrapi-mcp-server\nnpm install && npm run build\n\n\nAfter building, you can run the server directly (if you have a dist/index.js).\n\nbash\nnode dist/index.js\n

Tips and notes:\n- Obtain an SCRAPI_API_KEY from the Scrapi dashboard and keep it secure.\n- For the Streamable HTTP transport, you can connect clients without a Node.js runtime by pointing them to the hosted endpoint (https://scrapi.ai/mcp) or your self-hosted instance.\n- If you are self-hosting the HTTP server, you can start it with SCRAPI_API_KEY=your-api-key and the command line shown in the README.\n- Health endpoint for self-hosting is GET http://localhost:3000/health.\n- When configuring clients, you can pass the API key via environment variable SCRAPI_API_KEY or via CLI/HTTP headers as described in the documentation.\n- The MCP server supports both URL-to-Markdown and URL-to-Text outputs, with metadata extraction and clean output suitable for LLM prompts.