Media
6 items
Description
Copy the page you are reading as clean Markdown for ChatGPT, Claude or Gemini. Tables and code blocks survive. Copy the page you are reading as clean Markdown for ChatGPT, Claude or Gemini. Tables and code blocks survive the trip. You know the routine: select the article, copy it, paste it into the chat — then spend two minutes rebuilding a table that arrived as a wall of text. This extension removes that repair work. One toolbar button, one job: Copy Page as Markdown. HOW IT WORKS 1. Open the page you want to hand to the AI. 2. Click the toolbar button, or press Alt+Shift+M, to copy the page as Markdown. 3. It is already on your clipboard, and the result is shown in the popup so you can check it before pasting. If the browser blocks the automatic copy, the text box is already selected the moment a result is shown — Ctrl+C (Cmd+C on Mac) picks it up from there. WHAT YOU GET - Two modes. "Article only" converts the article to Markdown and drops the navigation, sidebars and footer, so you keep what you were actually reading. "Whole page" converts the whole page to Markdown, header and menus included. - An honest fallback: if the article cannot be detected, it says so on screen and converts the whole page instead, rather than handing you a broken result in silence. - Structure that survives: headings, lists, links, tables, fenced code blocks and strikethrough. - The title and the address on top, so the model knows what it is reading and you can find the source again: # The title of the page https://example.com/the-page ...the converted text... - A character count in the popup, for when you are working close to a context limit. TWO PERMISSIONS, AND NO INSTALL WARNING It requests exactly two permissions — activeTab and scripting — and neither of them shows a warning screen when you install it. That is the point of the design, not a lucky side effect. - activeTab: at the moment you press the button, and only then, it may read the one tab you are looking at. - scripting: needed to run the conversion inside that tab. It has no standing access to any website, and nothing of ours runs in the background while you are not using it. It cannot see your other tabs or your browsing history, because it never asks for the permission that would allow that. Nothing leaves your computer either: no server, no account, no sign-up, no analytics, no ads. The page is converted in your own browser, and nothing is stored — no settings, no history. Close the popup and the result is gone. WHAT IT DELIBERATELY DOES NOT DO - It copies the result to your clipboard. It does not save a .md file, and it has no download button. - It does not convert all your open tabs in one go. - It converts either the whole page or the article it detects. It does not convert just the part you have highlighted on the page — there is no "selection only" mode. - It does not crawl sites, follow links, or fetch anything on a schedule. - It keeps no history of what you converted, here or anywhere else. - No options page and no output templates in this version; the output format is fixed. - Chrome does not let any extension run on certain pages. There, nothing can be converted — and it tells you so instead of handing you an empty box. - Article detection is a good guess, not magic. When no article can be found at all, it says so and converts the whole page instead. But on an unusual layout it can settle on the wrong part of the page without being able to tell — which is why the result is shown to you before you paste it. - Formulas rendered by KaTeX or MathJax do not come across reliably. - Some sites build their code samples with scripts in a way that no extension is allowed to read. On those pages the samples may arrive as plain text, or not at all. PRICE Everything above is free. No account, and no limit on how often you use it. UNDER THE HOOD Three bundled open-source libraries do the work, rather than anything fetched from the internet: Mozilla Readability (Apache-2.0) finds the article, Turndown (MIT) turns HTML into Markdown, and turndown-plugin-gfm (MIT) handles the tables.