Features explained
Branch any web page into GPTree
Right-click any page and the Chrome extension captures what you are reading into GPTree as a document, with a conversation already open beside it.
The Chrome extension started with one job: move a ChatGPT conversation into GPTree. Branching a web page is the other direction. You are reading something, you want to think about it properly, and there is nowhere for it to go yet.
Right-click the page and choose "Branch this page in GPTree". Select some text first and the menu offers that selection instead.
What you get
Two panes, side by side. The page you captured is on the left, as a document. A conversation is open on the right, with the page already attached as context. Ask your first question and the model can see what you were reading.
The page is saved as a document in your library, so it stays available after you close the tab. The conversation behaves like any other GPTree conversation: branch it, compare two models on it, merge the answers back together.
What it captures
The main readable text of the page, up to about 16,000 characters, which is the same limit the model can see of any document. Pages longer than that are trimmed and marked as trimmed. If you had text selected, it captures only that.
The address is saved as provenance, with the query string and anything after a # removed before it leaves your browser. That is deliberate: query strings carry session tokens, reset links and search terms, and a tool that captures pages should not be the reason one of those ends up in a database.
What it can and cannot read
The extension reads a page only at the moment you ask it to. It has no standing permission to watch what you browse, and adding page capture did not widen what the extension can see: clicking the menu item is what grants it access, to that one tab, for that one read.
Some pages cannot be read at all. Chrome's own settings pages, other extensions' pages, the Chrome Web Store, and PDFs in Chrome's built-in viewer are all off limits to any extension, so the menu item either does not appear or tells you it cannot read that kind of page.
Signing in
Captures go to your account, so the extension has to be connected to it first. Connect it from the extension icon. If your browser is not signed in to GPTree when a capture opens, you land on the sign-in page and come back to the captured conversation straight after.
Captures in your knowledge library
A captured page is an ordinary document, which means that if you have the knowledge features turned on it becomes something GPTree can cite in later answers, the same as any document you upload yourself. Worth knowing before you capture something you would rather keep out of your own library.
Where it fits
Use it when the thing you want to think about lives on someone else's page. Use the ChatGPT import when the thing you want to continue is a conversation you already had. They are separate paths and neither replaces the other.