PullMD
A self-hosted service that converts a web URL, or as of v3 a PDF, Office document, EPUB, image, audio file or YouTube video, into clean Markdown for humans and AI agents to read.
PullMD strips navigation, ads and boilerplate from a page, auto-detecting Reddit and Hacker News threads with full comment trees, using Cloudflare's native Markdown extraction when available, running Mozilla Readability and Trafilatura on static HTML, and rendering JavaScript-heavy pages through a headless Chromium sidecar as a last resort.
Every conversion gets a stable share id that re-fetches the source if the cached copy is older than an hour, useful for a subreddit feed or similar recurring source. It ships as a PWA frontend with raw and rendered views, sharing, history and local file drag-and-drop; a REST API; an MCP server over Streamable HTTP; and a downloadable Claude Code skill.
Version 3 moved metadata such as the source URL and fetch date into YAML frontmatter, leaving the Markdown body as just a title and content by default, for a leaner, more token-efficient output; the old inline header can be restored with an environment variable.