Downright — Copy Any Page as Markdown for AI

Description

Copy any page as clean Markdown in one keystroke — correct tables, code fences, and math. No tracking. No network. Ever. Press Alt+M and the page you're reading becomes clean, correct Markdown on your clipboard — ready to paste into any LLM or Note taker. Downright is a markdown clipper built around one idea: the Markdown must be RIGHT. WHAT COMES OUT CORRECT • Tables — real Markdown pipe tables, including merged cells (colspan/rowspan), alignment, and pipes inside cells. The thing most converters silently mangle. • Code — fenced blocks that keep their language (Prism, Shiki, highlight.js, GitHub markup all recognized), with line-number gutters stripped. • Math — KaTeX and MathJax equations come back as their original LaTeX, not garbled symbols. Wikipedia formulas extract perfectly. • Links & images — every URL resolved to an absolute address, including lazy-loaded images. Clips work outside the page they came from. • Modern sites — Downright reads the rendered page, so single-page apps, logged-in pages, and web components (Shadow DOM) all clip correctly. THREE WAYS TO CLIP • Alt+M (Option+M on Mac) — copy the article as Markdown. Select text first and the same keystroke clips just the selection, keeping list numbering and table headers. • Right-click — copy the selection or page, or save the page as a .md file. • Toolbar popup — preview the Markdown before you copy, switch between Article and Full-page capture, and see a token estimate for AI context windows. You always find out what happened: every clip raises a toast with the token count, or the reason it couldn't run. Pages extensions aren't allowed to touch — browser settings, the extension gallery, the PDF viewer, local files without file access — say so by name instead of failing silently. Optional YAML front matter (title, source URL, author, date) makes clips land in major note taking apps perfectly. PRIVATE BY CONSTRUCTION • No network requests, ever — verified by an automated audit on every release. • No host permissions — Downright cannot read any page until the moment you invoke it, and never in the background. • No accounts, no analytics, no tracking. Free, and open source on GitHub so you can check all of the above. • Text a page hides from you — off-screen, zero-opacity, or written in invisible Unicode — is dropped, not clipped. What you can read is what you get. Source code: https://github.com/jayasankarmr/Downright

Details & SEO