Lyra rewrites your messages in every Mac app. On-device. Private. Learns how you write to each person in 50+ languages.
Mail, Slack, Cursor, Notion, WhatsApp — Lyra rewrites where you already write. No window to switch to.
If you can type in it, Lyra works in it.
Grammarly fixes your writing. Lyra understands how you communicate. It learns how you speak to different people, casual with your team, direct with your boss, warm with your family, and adapts every message while staying true to your voice. The more you use Lyra, the more it sounds like you.
Download for MacLyra runs a quantized language model locally on Apple Silicon and picks the right one for your Mac's memory. Nothing leaves your device, and rewrites feel instant.
Every rewrite runs on your Mac. No servers, no accounts, nothing to leak.
Q4_K_M quantized models keep latency low so your writing flow never stalls.
Compact models sit quietly in memory without slowing the rest of your Mac.
Lyra detects your memory at install and loads the fastest model it can run well.
Model selection happens automatically. Larger models produce richer rewrites; smaller ones keep things snappy on lighter machines.
Free forever for everyday use. Pro for power users who write all day.
Must-have features for everyday use.
Apple Silicon Mac, macOS 14+
For people who write all day.
No credit card required. Cancel anytime.
Everything you might want to know before you download. Still curious? Message me on X.
Lyra fixes and sharpens your writing in every app on your Mac — Mail, Messages, Slack, WhatsApp, your browser, anywhere you type. It works quietly in the background: when you pause typing, Lyra offers a cleaner version of what you wrote. Accept it or ignore it.
No. Rewrites happen entirely on your Mac, using a small language model that runs locally. Your messages, emails, and documents are never sent to a server to be rewritten. Lyra works even when you're offline.
Lyra requires an Apple Silicon Mac (M1 or later) running macOS 14 or newer. Intel Macs aren't supported.
On first launch, Lyra downloads the language model it uses to rewrite text (about 1GB, a few minutes on a typical connection). This is a one-time download — after that, everything runs locally. Lyra stays quiet until the download completes.
Lyra picks the right model for your Mac automatically. Macs with less than 20GB of RAM get Qwen3-1.7B (Q4_K_M, ~1GB) — small, fast, and light on memory. Macs with 20GB or more get Gemma 3 (Q4_K_M, ~3GB) for a quality bump where there's headroom. Both run entirely on your Mac.
Lyra uses macOS Accessibility to see the text field you're typing in and to place the rewrite back in the same spot. That's what lets it work in every app instead of just one. The permission is used only for this — nothing is logged to a server.
Effectively all of them — native Mac apps, Electron apps like Slack and Teams, and anything in your browser. If you can type in it, Lyra can rewrite in it.
Free includes 500 rewrites a month. Pro is $6.99/month, or just $39.99/year — that's 52% off — for unlimited rewrites.
Your account tracks your plan and rewrite count across reinstalls. Sign-in is only used for this — your writing is never associated with your account or uploaded anywhere.
Lyra is designed to stay light. The model uses roughly 1.4GB of memory while active and only runs when you pause typing. If you're on a heavy workload, you can quit Lyra from the menu bar and relaunch it any time.
Two things. Lyra works in every app — not just the browser, not just Apple's apps. And it runs on your Mac rather than in the cloud, so it's private by default and works offline.
Lyra checks for updates automatically and asks before installing anything. You're always in control.
Quit Lyra from the menu bar, then drag it from Applications to the Trash. To remove the downloaded model
as well, delete the folder at ~/Library/Application Support/Lyra.
Message me directly — @buildbypratik on X, or email pratik.garg@codestackconsulting.co.uk. Lyra is built by one person and every report gets read.