HTML to Markdown MCP Server converts web pages into a concise Markdown representation suitable for AI context. It preserves essential content such as tables, images, and links while removing noise like scripts, styles, navigation, and headers/footers, achieving substantial size reductions (approximately 90-95%) compared to the original HTML. The server uses trafilatura for robust content extraction and BeautifulSoup4 for reliable HTML parsing, with an architecture designed for streaming processing to efficiently handle large pages. It offers configurable options to include or exclude images, tables, and links, and enforces content-size limits (1 MB to 50 MB) to prevent excessive downloads; caching is optional to speed repeated conversions. For JavaScript-heavy sites and authenticated pages, the browser mode leverages Playwright to render pages, execute JS, and interact with cookies via a user profile. It supports Chromium, Firefox, and WebKit, and provides configurable wait strategies for dynamic content. Deployment options include uv-based usage, pip install, and Docker images preloaded with Playwright. The server is intended to run as an MCP endpoint and can be integrated with Claude Desktop or other MCP clients for automated conversions.