HTML Text to .md — Local Markdown Web Clipper

Select any part of a page and convert it to clean Markdown. Tables, code and math stay intact. Runs locally in Chrome.

Turn the part of a page you actually want into clean Markdown. The HTML is converted right where it sits, in your own browser — nothing is sent to a server. This is a web clipper you aim, not one that guesses. WHAT IT DOES You choose what to capture. Drag-select a passage, or switch on Highlighter mode and click the blocks you want: paragraphs, headings, lists, tables, code, quotes. Pinned blocks accumulate as you read, and one action turns the whole set into a single document, in the order they appeared. Alt+M copies your selection to the clipboard with an on-page toast and no panel at all. Alt+Shift+M opens the side panel and appends to what you have already collected. Hold Cmd or Ctrl to pick several separate ranges at once; they are stitched together in reading order. WHY POINTING BEATS GUESSING Tools that guess which part of an article is "the content" get it wrong quietly. A table disappears. Footnotes vanish. Half a section is missing and the saved file still looks perfectly fine. Point at what you want and there is nothing left to guess: what you picked is what you get. WHAT SURVIVES THE TRIP The HTML to Markdown step keeps headings, bold and italic, nested lists, links, images, blockquotes, horizontal rules, superscript and subscript. Tables keep their columns, and pipe characters inside cells are escaped instead of breaking the row. Fenced code keeps its language label. Formulas go through KaTeX, and MathML is rewritten as LaTeX first. Content inside Shadow DOM is flattened, so component-based apps and modern documentation sites come through intact. IN THE SIDE PANEL Preview the rendered result, or edit the raw text by hand. Undo and redo. Copy it, save a .md file, or take plain text with the syntax stripped out. One button hands the note straight to Obsidian, and on macOS another opens it in the EditMD editor. Title, source URL and capture date are attached as YAML front matter, and you can switch that off. PRIVATE BY ARCHITECTURE Everything happens inside the tab, in your own browser. There is no backend, no account, and no upload step. The extension sends exactly one anonymous usage event: a random install identifier, an event name, your interface locale and the version number. No addresses, no captured content, nothing else. That is the entire network footprint, and the privacy policy states it in the same words. FIFTY-TWO INTERFACE LANGUAGES Choose your interface language yourself instead of inheriting whatever the browser is set to, with full right-to-left layouts for Arabic, Hebrew and Persian. GOOD FOR - Building research notes in Obsidian, Notion, Logseq or a plain folder of text files. - Keeping a source link and a date attached to everything you clip. - Capturing formulas and code from technical documentation without repairing them afterwards. - Grabbing exactly the passage you need and pasting it into Claude, ChatGPT or any other model: nothing to strip out first, and nothing leaves your device. Open source under the MIT licence, built on Manifest V3: github.com/andryushkin/webtodotmd WHAT IT WILL NOT DO There is no one-click whole-page mode: you point, or you click elements. It cannot run on PDF files, local files or the browser's own internal screens. Images stay as links to their original host rather than being downloaded.