
It sits in your menu bar and does one job: make the words already in front of you better. Select anything, press ⌘E, keep working. Slack, Mail, Notion, Cursor, your terminal — it reads from your operating system, so there is no list of supported apps to check.
Get accessWorks on the free plan · no card
Why a shortcut
Everyone already has a good AI. What nobody has is a cheap way to reach it. The round trip — leave, paste, prompt, read, copy, come back, fix the formatting — costs more than the thinking does, so twenty times a day you skip it and send the rough version instead. A shortcut is the whole difference between using AI and meaning to.
You never leave the app. The improvement streams into the box you were already typing in, and your cursor stays exactly where you left it.
Three power moves
There is no thread to scroll, no conversation to manage, nothing to catch up on. It lives in the menu bar and stays out of the way until one of these three keys asks for it.
Rough sentence, half-formed thought, a prompt you can't be bothered to write properly. Select it, press it, and the better version replaces it in place — without your cursor ever leaving the box.
38 built-in packs plus your own templates. Type a //keyword, hit Enter, and it lands in whatever app you were typing in. Your prompt library stops being a folder you forget about.
Improve, Code, Grammar or Concise over the same selection — for the times the default rewrite isn't the one you were after.
Takes a second — we'll email you when you're in.
Next
Right now ⌘E does one job everywhere. The next thing we want it to do is notice where you are — because in Cursor you are writing a prompt, and in Slack you are writing to a person, and those two should never come back looking the same.
The same sentence, answered four ways. Nothing picked, nothing configured — the idea is that the app reads the window you're in.
What you typed, in all four cases: “hey so i looked at the numbers from last week and i think we should maybe push the launch back a bit, theres still 2 bugs open and the copy isnt final. lmk what u think”
Hey — I looked at last week's numbers and I think we should push the launch back. Two bugs are still open and the copy isn't final. Let me know what you think.
Same length. Still casual. Nothing added.
Hi — after reviewing last week's numbers, I'd suggest we push the launch date back. We still have two open bugs and the copy isn't final. Happy to talk through the details — let me know your thoughts.
One notch more formal, because it's email.
Review our launch readiness and recommend a go/no-go date. Context: 2 bugs still open, final copy not approved, last week's metrics reviewed. Return: a recommended date, the risk of shipping on time, and what has to close first.
Talking to a model, not a person — scaffolding is right here.
### Launch recommendation: push the date - Two bugs still open - Copy not final - Last week's numbers reviewed Proposal: move the launch, confirm the new date once the bugs close.
A doc can carry headings. A chat message can't.
Every tool like this eventually grows a menu — formal, casual, concise, professional — and then you spend your day in the menu instead of in your work. So it would work it out itself, from three things it can already see: the app you're in, what you actually wrote, and whether you selected one sentence or left the whole box highlighted.
When it guessed wrong, fixing it would cost one keystroke rather than a settings screen:
To be straight with you: this part isn't built yet. ⌘E works everywhere today, but it doesn't yet change its mind based on the app you're in. Whether that gets built next — and which apps get tuned first — is decided by what people tell us below.
The promise
This is the part that would be easy to get wrong. Turning your quick Telegram message into a LinkedIn post is worse than doing nothing — you'd stop trusting the shortcut and never press it again. So when you're writing to a person, the rule is the smallest edit that works: fix the grammar, tighten the rambling, stop.
Expanding, structuring and adding context are genuinely useful — but only when you are writing to a model. Knowing the difference is the whole job.
Nothing is read, watched or sent in the background — your text leaves your machine only when you press the shortcut. How we handle your data.
Two quick questions with it — they decide what gets built next, and which apps get tuned first.