Skip to content
WeType

About

A small app, made by one person, on purpose.

WeType is AI autocomplete for Mac. It is designed, built and supported by Aryaan Qaiser, an independent developer, which means the person who reads your email is the person who wrote the code.

Why it exists

Most AI writing tools ask you to leave what you are doing: open a chat window, write a prompt, copy the answer back. For the writing people actually do all day, replies, messages, notes, commit messages, that round trip costs more than it saves. WeType starts from the other end. It stays out of the way until you pause at the end of a line, then offers the likely rest of your sentence in grey, where you are already looking. One key takes a word.

It runs the model on the Mac because a tool that sees everything you type, in every app, should not be sending it anywhere. That constraint shaped everything else, including the choice to let you pick the model that suits your Mac rather than shipping one size for everyone.

What it is built on

How it works under the hood

Native macOS app
Written in Swift. It lives in the menu bar; there is no Dock icon and no main window.
llama.cpp
The open-source engine that runs the language model on the Mac's GPU through Metal.
Open models
GGUF builds of Qwen, Llama and Gemma models, downloaded from the Ollama registry inside the app.
macOS Accessibility API
Reads the focused text field and caret position, and inserts accepted text. The same interface screen readers use.
Sparkle
Signed, verified updates. Every build is Developer ID signed and notarized by Apple.
LemonSqueezy
Merchant of record: handles payment, tax, invoices and licence keys. WeType never sees card details.

For the details that matter to privacy, including every file WeType writes and every anonymous event it can send, see the privacy policy. For what changed in each version, the changelog.

Aryaan Qaiser

Developer of WeType

Aryaan Qaiser designs, builds and supports WeType, and writes the guides on this site. Support email goes straight to the developer rather than to a queue, and replies usually come within a day.

support@wetype.tech · Contact page · Press kit

At a glance

WeType, in ten facts.

The specifics, in one place, for anyone who would rather skip the argument.

What it is
AI autocomplete for macOS. It predicts the rest of the sentence you are typing and shows it as grey ghost text at the cursor, in any app.
How you use it
Tab takes the next word, the key above Tab takes the whole suggestion, and Esc dismisses it. Both accept keys can be rebound.
Where the AI runs
On your Mac, using llama.cpp. What you type is never sent to a server, and suggestions work offline.
Works in
Any app with a standard text field: Mail, Messages, Slack, Notes, Safari, Chrome, Notion, VS Code, Xcode, Terminal and more. No plugins.
Models
9 to choose from, 379 MB to 4.5 GB, including code models. The default is Qwen 2.5 3B (base).
Languages
Automatic, or pinned to one of 15, including English, Spanish, French, German and Chinese, Japanese, Korean.
Requirements
Apple Silicon (M1 or later), macOS 14 Sonoma or later. Intel Macs are not supported.
Price
$20 once, or $6 a month. Both unlock the same app. 30-day money-back guarantee.
Free trial
5 days, up to 100 accepted words a day. No account and no card.
Made by
Aryaan Qaiser, an independent developer. Current version 1.0.2.

New to the idea? What AI autocomplete is and how it works. Comparing apps? WeType next to the alternatives.