FileConcat

Browser clipper

FileConcat Clipper is a Chrome extension that turns the page you are reading into Markdown and pushes it into an open fileconcat.com tab, where it joins your bundle like any dropped file. It covers articles on any site, plus YouTube transcripts, Reddit threads, Hacker News discussions, and ChatGPT, Claude and Gemini conversations.

There is a page about it at /clipper, and the extension itself is on the Chrome Web Store.

Install and open

  1. Install it from the Chrome Web Store.
  2. Reload any tab you already had open. Chrome does not inject content scripts into tabs that existed before an extension was installed, so a YouTube or fileconcat.com tab from before will not answer.
  3. Click the toolbar icon to open the side panel. It stays open while you browse.

The panel's Now block names the page it is looking at. It re-reports itself on every tab switch, every navigation, and whenever the page grows, so scrolling a feed to load more rows updates the list without a click and the panel never describes the page you came from.

What it reads

Any article. If a page reads as an article, the panel offers Clip this page. Mozilla's Readability decides what the body is and Turndown renders it, which covers Substack, Medium, documentation sites, news and blogs with no per-site code. Navigation, sidebars, cookie bars and footers are left behind.

YouTube. A watch page offers Clip this video and the clipping carries the video's description and its full transcript. A channel's Videos tab and a search page list every video the page has loaded, with a tap target on each row. A channel's Playlists tab lists playlists instead, and taking one row takes every video that playlist holds, filed under the playlist's name. Include comments adds the top 60 threads and the first page of replies under the 25 most-replied of them, and is off by default, because it costs about 30 more requests, five seconds, and three to four times the tokens.

Reddit. A thread offers Clip this post and takes the post plus the comments the page has rendered, keeping every author, score and level of nesting. Expand more comments clicks the thread's own "more replies" three rounds deep and is off by default because it is slower. Clipping from a subreddit listing instead fetches each post's full body but not its comments, which Reddit renders on the client; that clipping says so in place of its comment section and never overwrites a fuller clipping of the same post.

Hacker News. A thread returns its entire comment tree in one request, however deep it runs. The front page lists its stories with a tap target on each.

ChatGPT, Claude and Gemini. All three pages render a window of the conversation (4 of 82 messages on a measured ChatGPT page, 2 of 11 on Claude, 2 of 105 on Gemini), so the extension reads the conversation the way the site's own client does: one or two same-origin requests, using the page's session, nothing stored. The file holds every turn with the assistant's own Markdown. The opt-in adds reasoning and tool activity, measured at 3 to 6.5 times the clipping's size on ChatGPT, 1 to 3.7 times on Claude and 1 to 2.1 times on Gemini. Files a Claude conversation wrote, and HTML apps a Gemini conversation built, arrive as their own files under the conversation's folder.

Sending into a bundle

Clips collect in the panel's cart. Peek reads a clipping's Markdown back inside the panel before you send it, which is where a wrong page or a transcript that stopped early shows up while it is still cheap to fix.

Send pushes every finished clipping into an open fileconcat.com tab, opening one if there is none. A send extends the bundle rather than replacing it and de-duplicates by path, so a repository and the discussion about it can sit in one bundle in either order, re-sending a corrected clipping fixes it in place, and sending the same set twice does not double it. Start over in the app is the only thing that clears a bundle.

Closing the panel does not stop a batch. The clipping runs in the extension's service worker, so the panel reads the finished result whenever you open it again.

What it does not do

Nothing is read from a page until you ask for it. There is no background crawling, no page is touched because you happened to visit it, and no clipping is sent anywhere except the fileconcat.com tab you are looking at. There is no account and no server of ours in the path. Clippings are held in your browser's own storage between the clip and the send, and the sent list is a receipt rather than a library: rows leave after seven days. The privacy page covers the extension alongside the site.

Walkthroughs

Three site-specific guides cover what the manual route loses on each and what the panel returns instead: saving a Hacker News thread, saving a Reddit thread as Markdown, and getting a YouTube transcript into ChatGPT.