I'm Preetham Rangu, and I built HunchKey because [I was writing the same kinds of replies all day, in five different apps, and every AI tool made me sound like someone else].

What it is: your next message, in your voice. Click into an empty reply in Discord, Slack, WhatsApp or email and press Ctrl + Shift + H (⌃⇧H on a Mac). HunchKey reads the conversation and writes the reply the way you talk to that person. Tab types it, 1 / 2 / 3 picks another option, Esc lets it go. You still press send.

No good hunch? Tell it what you want. When it isn't confident, the island offers a small box. Type "say I'm in but I have to leave by 9" and it writes that message from the chat, in your voice. (needs 0.1.9)

Wrote it yourself in a hurry? Ctrl + Shift + R fixes the draft and shows you exactly what changed. Words you meant, like "u" and "pls", stay yours.

Halfway through a sentence? The same shortcut finishes it, in Gmail, Notion, Google Docs, VS Code or a web form.

A few things I care about:

It sounds like you to each person. Quick and lowercase with a close friend, full sentences for a client. A voice for each kind of writing, too: chats, work, email, documents, code, or your own.
It asks before it remembers. When it picks up who someone is to you, the island checks first. You can see, correct or forget any of it in People.
Nothing pops up until you ask. No suggestions chasing your cursor. That's also the only time it reads anything.

Privacy: password fields are never read. No keylogging, no typing history. Screen context is optional and kept in memory for up to 45 seconds, never on disk. You can leave out any app or website, and nothing is ever sent for you.

It's free during early access on Windows and macOS 12+, with an account. It needs an internet connection.

I'd love to hear where it gets things wrong. That's exactly the feedback I need right now. I'll be here all day. 🙏