Markdown Scraper API (MCP Enabled)

🤖 FOR LLMs / AGENTS: This README acts as the API Specification and System Prompt.

This is a serverless Web Scraper API built with Node.js and Hono, deployed to Cloudflare Workers. It uses the r.jina.ai engine to bypass captchas and extract clean Markdown from any given URL.

🧠 System Architecture & Context Engineering

• File Map: Always read CODEBASE.md to understand file dependencies and system routing before modifying code.

• Session Memory: Always read and update STATE.md at the beginning and end of each session to maintain context across chats.

📡 API Specification

1. GET /mcp/manifest (Discovery)

Returns the Model Context Protocol (MCP) JSON manifest. Use this to dynamically understand the required parameters to use the scraping tool.

2. POST /scrape (Action)

Requires Authentication: Authorization: Bearer <token>

Este endpoint utiliza o protocolo HTTP 402 Payment Required.

• Se você não fornecer um token ou o token não tiver saldo, a API retornará um erro 402.

• A resposta do erro 402 conterá uma paymentUrl (Dodo Payments) onde você pode adquirir créditos.

• Após o pagamento, você receberá um token que deve ser enviado no header Authorization.

Request Body:

{
  "url": "https://example.com"
}

Success Response (200 OK):

{
  "success": true,
  "data": {
    "title": "Page Title",
    "url": "https://example.com",
    "content": "# Markdown extracted..."
  }
}

🚀 How to Run Locally

# Start the local Cloudflare dev server
npm run dev

For generating/synchronizing types based on your Worker configuration run:

npm run cf-typegen

Pass the CloudflareBindings as generics when instantiation Hono:

// src/index.ts
const app = new Hono<{ Bindings: CloudflareBindings }>()