Glintype.
Inline autocomplete, on your Mac.

Glintype watches the field you're already typing in, finishes the sentence in ghost text, and accepts with Tab. Apple Intelligence or a local model. Your words never leave your computer.

One-time payment·macOS 15+·Apple Silicon & Intel

Attaches to every editable field on macOS

Apple MailMimestreamOutlookGmailMessagesSlackTeamsDiscordWhatsAppWeChatSafariChromeBraveEdgeOperaVivaldiArcFirefoxZen BrowserNotesTextEditWordGoogle DocsNotionObsidianVS CodeCursorWindsurfXcodeJetBrainsGitHubGitLabLinearJiraChatGPTClaudePerplexityLe ChatApple MailMimestreamOutlookGmailMessagesSlackTeamsDiscordWhatsAppWeChatSafariChromeBraveEdgeOperaVivaldiArcFirefoxZen BrowserNotesTextEditWordGoogle DocsNotionObsidianVS CodeCursorWindsurfXcodeJetBrainsGitHubGitLabLinearJiraChatGPTClaudePerplexityLe Chat
How it works

Four steps. Then it gets out of the way.

Glintype lives in your menu bar. It does its job, then disappears. The whole setup takes about a minute.

01

Grant three permissions.

Accessibility reads the focused field. Input Monitoring catches your Tab. Screen Recording is optional and only used for visual context.

Accessibility
Required
Input Monitoring
Required
Screen Recording
Optional
02

Pick an engine.

Apple Intelligence on supported Macs. Or download a local GGUF model in your tier of choice. Switch anytime.

Apple Intelligence
macOS 26+ · Apple Silicon
Open Source
Qwen · Gemma · Phi
03

Start typing.

Click into any field — Mail, Slack, Notes, Xcode, anywhere on macOS. Glintype is already listening. The ghost text appears beside the caret.

No mode switch
Always on
Caret-attached
No overlay
04

Press Tab.

Tab accepts the next chunk. Shift+Tab accepts the rest. Keep typing to ignore. The interaction stays out of the way.

Tab
Accept next chunk
Shift + Tab
Accept the rest
Keep typing
Diverge
Engines

Three engines. Your call.

Glintype picks the best path for your Mac automatically. You can override it any time from the menu bar app.

BETA

Apple Intelligence

On-device, Apple optimised

Routes through FoundationModels when your hardware, macOS, and language qualify. The fastest path on supported Macs. The model ships with macOS — no separate Glintype download — and stays on the Neural Engine.

Apple Silicon
macOS 26+
Supported language
On diskShipped with macOS
GGUF

Open Source

Local llama.cpp, your choice of model

Run any tier of local model in-process. Glintype ships with a curated set of Qwen, Gemma, and Phi weights, or drop in your own .gguf file. Available on Apple Silicon and Intel.

Apple Silicon or Intel
macOS 15+
Tier-appropriate RAM
On disk0.4 – 18.6 GB
BYOK

Cloud

Hosted models, opt-in only

Drop in your own OpenAI, Anthropic, or Mistral key and Glintype can route completions to the cloud. Use it as a fallback for older or slower Intel Macs, or in Hybrid mode where the local model handles the easy suggestions and the cloud only steps in when context gets long. Strictly opt-in. The key lives in Keychain and nothing is sent unless Cloud or Hybrid is selected.

Your own API key
Keychain permission
Active internet
On diskNo model on disk
Per-app profiles

Different apps, different voices.

Glintype detects the frontmost app and applies its profile. Length cap, tone, accept keys — all set per app, all overridable.

Slack

profile

Concise, friendly, uses emoji only when you do

Sounds good — I can push that by Thursday.

Apple Mail

profile

Formal, complete sentences, full sign-off

Thanks again for making the time. Talk soon.

Notes

profile

Matches your personal voice, streams longer

The next step is to ship the demo and watch what sticks.

Notion

profile

Structured, prefers lists and headings, expands shorthand

Q3 priorities: ship the Mac release, lock in pricing, hire a designer.

Accept modes

Tab to accept.
Or don't.

Glintype never blocks your cursor. Accept what you want, dismiss the rest, keep typing to diverge. The interaction is opt-in on every keystroke.

Every shortcut is fully customisable — rebind accept, dismiss, and chunk keys per app from Settings → Accept keys.

Tab

Accept the next word or phrase

Default accept. Works in any editable field.

ShiftTab

Accept the rest of the suggestion

One quick press takes the whole sentence.

Keep typing

Diverge naturally

No key required. The ghost text vanishes the moment you steer away.

If you can type in it, Glintype works.

Native apps, browsers, Electron editors, terminal-adjacent tools. Glintype attaches via macOS Accessibility and stays out of the way.

Apple Mail
GitHub
Outlook
Notion
Gmail
Word
Chrome
Linear
Slack
ChatGPT
WeChat
Discord
VS Code
Notes
GitHub
Safari
Notion
Apple Mail
Word
Outlook
Linear
Gmail
ChatGPT
Chrome
Discord
Slack
Notes
WeChat
Safari
VS Code
Local-first

Your writing never leaves the Mac.

Not encrypted in transit. Not anonymized. The autocomplete runs on local weights that never call home. There is nothing to intercept because nothing is transmitted.

Local models by default

GGUF weights are downloaded once, hashed, and stay on your disk. Apple Intelligence stays on the Neural Engine.

Cloud is opt-in, BYOK only

Want a hosted model? Drop in your own key. It lives in Keychain and only runs when you choose Cloud or Hybrid mode.

No telemetry, no tracking

No analytics SDK, no usage pings, no “anonymous improvement” opt-out. The only network calls are license activation and auto-update checks.

Encrypted dictionary and memory. Your learned phrases and per-app replacements sit in a Keychain-encrypted store. Even with disk access, the content is sealed.

Early access pricing

Buy once, own forever.

No subscriptions. No recurring fees. Pay once and Glintype is yours for life, including all future engines and model weights.

Free trial

$0for 7 days

Full feature parity. No card required. Cancel by uninstalling.

  • All engines and model tiers
  • Every per-app profile
  • Accept modes and key customization
  • Multilingual writing

Why one-time? Glintype runs on your Mac, on your hardware, with weights that never phone home. There is no usage to meter, no seat to count, no feature to drip out next quarter. You bought the Mac. You bought the model. We just built the bridge — once.

Frequently asked

Common questions.

What Mac do I need?

macOS 15 or later. Apple Silicon is strongly recommended for local model speed; Intel works with the Fast and Balanced tiers. The Apple Intelligence engine additionally needs an Apple Silicon Mac on macOS 26 or later.

Apple Intelligence vs Open Source — which should I use?

Apple Intelligence is the fastest path and is on-device with no Glintype-managed download, but only runs on supported Apple Silicon Macs and macOS versions. Open Source works everywhere, gives you a choice of model size, and is what you fall back to when Apple Intelligence is unavailable.

Can I bring my own .gguf model?

Yes. Drop any Q4_K_M GGUF into Glintype's models folder and refresh the model list. Glintype validates size and SHA-256 metadata before loading it.

Does it work offline?

Yes. All AI processing happens on your Mac. License activation checks the network on startup, but between checks the app works fully offline — including on a plane.

How do per-app profiles work?

Glintype detects the frontmost app and applies a profile with its own length cap, context mode, tone, and accept keys. You can override or add new profiles from the menu bar app.

What about my privacy?

Your writing never leaves your Mac. The only outbound traffic is license activation and auto-update checks. Cloud providers are strictly BYOK, stored in Keychain, and only used if you explicitly pick a cloud or Hybrid mode.

Is it a subscription?

One-time purchase. $18 launch price, $30 full price. Own it permanently. No renewal, no recurring fees.

Stop typing the endings
you have written a thousand times.

One-time price. No subscription. Yours forever.

macOS 15+ · Apple Silicon & Intel · Apple Intelligence & local GGUF