Skip to content
WriteAmp
Now in 1.0 · macOS 15+ · Apple Silicon

Your typing, amplified.

WriteAmp reads what you've typed and predicts what comes next — as inline ghost text in any app. Press Tab to accept.

Notes — sample
Ghost text ready Press Tab to accept
100% on-device
Zero telemetry
No subscription
Workflow

Type. Predict. Accept.

Three keystrokes separate a thought from a sentence. WriteAmp finishes the thought — you decide what to keep.

1

Type naturally

Keep typing in any app — Notes, Mail, Slack, VS Code, anywhere. WriteAmp watches in the background, locally, with no lag.

2

See the suggestion

After a brief pause, the next few words appear in subtle gray ghost text right at your cursor. No popup, no interruption.

3

Tab to accept

Press Tab for the next word, ` for the whole thing. Esc to dismiss. Or just keep typing — it goes away.

Smart Context

Adapts to where you write.

WriteAmp detects which app you're in and adjusts the tone automatically. Mail gets formal sentences. Slack stays casual. Code stays verbatim. You can override per-app once and it remembers.

  • Auto-detects Mail, Slack, Notes, VS Code, Terminal, and 20+ apps
  • Per-app profiles for max words, stop sequences, insertion mode
  • You stay in flow — WriteAmp picks the right voice
Mail.app

Following up on our discussion, I'd like to...

Slack

yeah sounds good, lets do

VS Code

function handleRequest(req, res) {

Learned from your typing
@amit DataDab Swift 6 llama.cpp

Encrypted on-device, capped at 100 sessions, opt-in

Personalization

Learns the words you actually use.

Opt-in personalization extracts the names, code terms, and phrases you type often, then quietly injects them into suggestions. Encrypted with AES-GCM, capped at 100 sessions, and deletable with one click.

  • Bigrams, trigrams, named entities — all encrypted with a Keychain key
  • Off by default — you turn it on, you turn it off, no questions asked
  • 100 session cap prevents unbounded growth
Model Tiers

Fast by default. Sharp on demand.

The 270M Light model is bundled — instant suggestions, no download. Want better accuracy for code or long-form writing? Download Balanced, Sharp, or Pro on demand. All tiers run on your Mac, all with SHA-256 integrity.

  • Light · 292 MB bundled · every-day typing
  • Balanced · ~600 MB · better accuracy
  • Sharp · ~1.2 GB · code & technical writing
  • Pro · ~2.5 GB · maximum accuracy
L
Light
270M · 292 MB
Active
B
Balanced
~600 MB
Download
S
Sharp
~1.2 GB
Download
P
Pro
~2.5 GB
Download
Privacy

Your keystrokes never leave.

No cloud. No analytics. No model calls. No telemetry. The model runs on your Mac's GPU, your text stays in your apps, and WriteAmp doesn't even have a server to phone home to.

Encrypted history

Personalization data is AES-GCM encrypted with a Keychain-stored key, on disk and at rest.

No tracking

Zero analytics SDKs, zero crash reporting, zero "phone home". You can verify with a network sniffer.

No cloud sync

No account. No login. No sync. Your settings live in ~/Library/Application Support/WriteAmp/.

On-device inference

A 270M model running on Apple Silicon GPU completes 80-120 tokens/sec. No server round-trip.

"It's the first autocomplete that actually feels like autocomplete should — invisible until you need it, then obvious in hindsight."
A
Beta tester
Software engineer

Simple, honest pricing.

Free for the first 30 days, then 200 words a day forever, or unlock everything for a one-time payment.

Trial

Try everything for 30 days

Free
  • Unlimited completions
  • All model tiers
  • Personalization
Start trial

Free

After trial, 200 words/day

$0 forever
  • 200 words / day
  • Light model bundled
  • All features included
Download
Best value

Lifetime

Unlock everything, forever

$39 one-time
  • Unlimited words
  • All model tiers
  • Free updates for life
  • Priority support
Unlock lifetime

Questions & answers

Does my text ever leave my Mac?
No. The model runs locally on Apple Silicon using llama.cpp. Your text, predictions, and personalization data stay on your Mac. The only network calls WriteAmp makes are the initial model download, Sparkle update checks, and license validation — never your content.
Which apps does it work with?
Anywhere you can type. Notes, Mail, Slack, Messages, VS Code, Terminal, Safari, Chrome, TextEdit, Notion, Linear — the list goes on. Per-app profiles tune completion style for Mail (formal), Slack (casual), VS Code (verbatim), and more.
What's the difference between Light, Balanced, Sharp, and Pro?
Same architecture, different sizes. Light (270M, bundled) is fast and good enough for everyday writing. Balanced, Sharp, and Pro are progressively larger models with better accuracy for nuanced or technical text. All run on your Mac; downloads are SHA-256 verified.
Is there a subscription?
No. WriteAmp is either free (with daily limits) or a one-time $39 lifetime license. No monthly fees, no renewals, no surprise charges. Updates are free for the life of the app.
What permissions does WriteAmp need?
Accessibility (to read the focused text field and insert completions). That's it. No microphone, no screen recording, no Full Disk Access. The first run shows a system prompt to grant Accessibility.
What Mac do I need?
macOS 15 (Sequoia) or later on an Apple Silicon Mac (M1, M2, M3, M4 or newer). The Metal GPU is used to run the model; Intel Macs are not supported.

Start typing smarter.

Free 30-day trial. No account. No credit card. Just download and grant Accessibility — you'll be completing sentences in 30 seconds.

No telemetry. No account. MIT-licensed components, Apache 2.0 model.