Fetch any public web page (or PDF) and return clean, LLM-ready Markdown with title, metadata and links. Boilerplate, ads and scripts removed.