Offprint

Offprint takes a page out of the browser in whatever form you actually need: clean Markdown, a YouTube transcript, a full-page screenshot, or a single offline HTML file with the images baked in. It is built for people who feed pages to AI agents, so the clipboard is a first-class destination and every capture carries a token estimate. CAPTURE - Readable (article extraction), Full page, or Selection - YouTube transcripts, timestamped, human captions preferred over auto ones - Full-page screenshots in one shot, including the part below the fold - Offline reader HTML: one file, images inlined, opens anywhere, forever THEN - Copy to the clipboard or save to a folder - click saves, shift-click copies - Edit the Markdown before you take it - Save every highlighted tab at once, as a zip or one merged file - YAML front matter and sortable filenames: {date}-{n}-{title} - Clip history, so the page you took an hour ago is still there - Per-site rules: a site remembers the mode, folder and summary you want ON-DEVICE AI, STRICTLY OPTIONAL Summaries use Chrome's built-in model. Nothing is sent anywhere, there are no API keys, and the model only loads if you ask for a summary. ALSO DRIVABLE FROM A TERMINAL - A PAID ADD-ON An optional native-messaging bridge lets "offprint capture" return a page as Markdown on stdout, without touching the browser UI or stealing focus. The extension is free and stays free; the command-line tool is unlocked by a one-time EUR 15 licence, with no subscription and no account. PRIVACY No account, no telemetry, no analytics, and no server of ours - we never receive anything you capture. A few sites need one request to the site itself to get what the page does not contain: YouTube and Bilibili for transcripts, and X posts via fxtwitter.com. They are listed by name in the privacy policy at https://marcusvanesveld.com/offprint/privacy/ Extraction is Defuddle + Turndown; the Markdown tidy and token estimate are Rust/WASM.