When you have Claude open in several terminals, each one only knows about itself. lookover lets you just mention another one, and it brings over what's happening there, so you don't have to go find it yourself. Everything stays on your machine.
You don't run any command. You just mention the other terminal in a normal prompt.
One terminal wrote a story; from another, we ask for the hero's color, and lookover brings that context over automatically.
Three simple steps, all on your machine.
Each terminal quietly keeps short notes on what it's up to, what you asked and what it's working on. Nothing heavy, and no full text is kept.
When you mention another terminal, it picks the one you mean, by its name or by what it's working on. If it's not sure, it asks you.
It hands that to Claude as background info, so the answer already takes into account what the other terminal was doing.
Name a terminal and you get its real context. Ask for more and it digs deeper.
Point at a terminal and lookover brings its recent context, what you asked there and what came back, so the answer actually knows what happened.
Ask for "the full context", "what was written", "the whole thing" and it reaches further back and shows more of it, word for word.
One line. It downloads the latest version, sets it up, and shows you exactly what it changes before going ahead.
curl -fsSL https://espigah.github.io/lookover/install.sh | bash
Prefer a package? Grab a .deb, .rpm or .tar.gz
from the releases page.
Then turn it on with lookover init.
Changed your mind? lookover uninstall undoes everything. To pause it
on the spot: touch ~/.claude/lookover/disabled.
Beyond the automatic in-prompt reference, there's a CLI to inspect.
lookover list # live sessions + topics/skill lookover show <name|id> # a session's digest lookover doctor # health: sessions, digests, version lookover status # adoption: live vs covered
It only speaks up when you ask about another terminal.