You caught it.
Close the tab.

The chunk, not the whole page.

Grab the part you cared about; where it came from comes along on its own. Ever tried to share one piece of an email without forwarding the whole thread? Ever tried to make sense of six screenshots in a folder? That's the job. Draw a region, pick one element, take the visible window or a transcript panel. And when the piece that matters is the whole page, scroll it yourself and catchroll captures the full column, top to bottom. Move a results list to its next page and catchroll follows the page you move to; it never clicks Next for you. Paged Scroll handles pages that swap their content in place.

Bring in your own.

Not everything you need lives on a web page. Press ADD FILE, drag and drop, or paste with Ctrl+V: images, PDFs (each page becomes its own tile), Word .docx, Excel .xlsx, and text files. Draw a text box and type, or paste copied text onto the board and it becomes a text box. A rich copy from a document, a spreadsheet or a web page keeps its headings, lists, tables and links.

A board, not a folder.

Everything lands on one canvas: a big vision board for the thing you're working on, whether it's a purchase, a trip, a project or real research. Arrange it, mark it up, keep building. No feed, no followers, no algorithm: the board is for you and your AI. You don't remember filenames; you remember what things looked like. Connect your library folder and finding a board is just load, load, load, until you recognize the one you wanted.

Black out, pixelate or blur any part of a tile. Redaction destroys the pixels and removes the saved text and the source pointer from every file you save or export. Its limits are on the security page.

Your chats swallowed your research.

Use AI seriously for a few months and you'll have hundreds of chats. Somewhere in there is the analysis, the source, the thing it explained perfectly once, and you'll never find it again. So you end up trusting the model to remember, and that memory is one you can't see, can't edit, and that's sometimes flat wrong. catchroll pulls what matters out of the chat stream: files you keep, visible, ready for the next conversation. Chats end. Your library doesn't.

Hand your AI the frame.

Models say they read your PDFs and links. Press them, and… they didn't. catchroll hands the model what you saw: the words it can read and the pictures, in a file you can open yourself. No fetch to fake. You see what you're handing over, and you decide what's on the table: “just look at this.” Either file can go to an AI. The .md from EXPORT FOR AI is text only and the lightest to share; the .catchroll carries the same reading plus every picture.