An MCP server that combines Puppeteer-based web scraping with Mozilla Readability and Turndown to deliver clean Markdown outputs from web pages. It leverages stealth browsing and AI-driven interaction to automatically handle overlays and interactive elements such as cookie banners, CAPTCHAs, newsletter prompts, paywalls/login walls, age verification prompts, interstitial ads, and other blockers. After performing interactions, the server uses Readability to extract the main content and converts the HTML into well-formatted Markdown, with special handling for code blocks, tables, and other structured content. The tool is designed to be consumed via NPX for easy usage without cloning, and supports three communication modes: stdio, SSE, and HTTP, enabling integration with MCP-compatible LLM orchestrators in various deployment scenarios. The MCP tool exposed is named scrape-webpage and accepts parameters such as url, autoInteract, maxInteractionAttempts, and waitForNetworkIdle, returning content (markdown) and metadata (including success and contentSize). Configurable environment variables include OPENAI_API_KEY, VISION_MODEL, API_BASE_URL, TRANSPORT_TYPE, and DISABLE_HEADLESS, with headless mode defaulting to off for real-time viewing when enabled.