Skip to content

Unshell 2.5.1

Latest
Latest

Choose a tag to compare

View all tags
@dekrezz dekrezz released this 21 Oct 17:07
· 2 commits to main since this release
v2.5.1
4ca1511

Install

Firefox: download Unshell-2.5.1.xpi and open it in Firefox. It is signed by Mozilla and installs permanently. Firefox 109+.

The add-on ID is unshell@dekrezz.dev. If an earlier Unshell or ScrapLLM is installed, remove it in about:addons first: Firefox treats this as a separate add-on.

Chrome: download Unshell-Chrome-v2.5.1.zip and unzip it. Then open chrome://extensions, turn on Developer mode, choose Load unpacked and pick the unzipped folder.

Unshell-Source-v2.5.1.zip is the source package.

A popup built around the page you are on

The page comes first. A card at the top shows what is about to be copied: the site's own icon, the title and the address, or several selected tabs listed by name. When a copy finishes, its result and token count land on the card's foot.

One key copies. The variants sit in a row of equal keys under it: Save .md, the opposite scope for this one copy (Full page or Main content), Copy page on a conversation, and Selection when something is highlighted.

What the Markdown keeps is on the card. Links, Images, Tables and Metadata are switches, the same settings as in Settings. The Telegram limits sit on the card too.

A chat copy takes the last 3 days by default. The dates come from the messages themselves on ChatGPT and Claude. Settings → Chats switches to the last N exchanges or the whole chat. A chat without message dates is cut by the exchange count instead, and the copy and the popup both say so. A range with nothing in it is an error that names the latest message date. The range row on the card is hidden unless you turn it on.

Settings are rows, grouped into Copy, Chats, Sites, Interface and About. A site's own options open only while that site's mode is on.

The window stays put. It does not slide when you hold the mouse or select text, and settings fade in place.

Private by construction

Page content never leaves the browser, and the extension tells Firefox it collects no data. The only network requests are:

  • the optional token counter's tokenizer data;
  • the page card's icon;
  • the sites whose own API an extractor reads.

Removed

Research (searching the web and merging the results into one file) is gone. With it went the DuckDuckGo access, the optional access to every site, and the offscreen document.

Assets 5
Loading