The calm, private writing app for clearer drafts.
A distraction-free desktop editor with local proofreading on demand, local-first Lex rewrites. No account or subscription is required.
Browsing on your phone? Grab Lexicon on a computer.
Click the preview above or watch the full product launch video on YouTube (with audio)
Write. Proofread. Polish. Export.
One private canvas for the whole writing process. The editor is the product. Proofreading, Lex, and export are built in, not bolted on.
A calm canvas for first drafts
Headings, lists, tasks, tables, and images, plus / slash commands and local auto-save. Templates for Academic, Novel, Blog, or Executive drafts. One shortcut clears the chrome when you need focus.
Grammar, style & readability when you check
Run inline checks for spelling, grammar, and style with 1-click Apply. Clarity, readability, and tone update with the check. Built-in checks run on your CPU with zero configuration; completely private and offline by default.
Local-first Lex rewrites & tone shifts
Highlight any passage and ask Lex to rewrite, tighten, summarize, or shift tone: Professional, Academic, Friendly, Casual, Formal, or Concise. Powered by local GGUF models or your own Ollama and LM Studio servers.
Deep Proofread: Deterministic Rules Meet Local AI Clarity
Deep Proofread pairs LanguageTool's deterministic grammar rules with Lex's local LLM clarity. Mechanical errors and clarity improvements appear side by side, each with a one-click action.
Deterministic grammatical and spelling checks running locally on CPU. Catches mechanical errors with zero false certainty.
Chunked local model analysis that tightens wordy phrasing and improves flow while strictly honoring semantic polarity.
From thought in any language to natural English
Write a thought in Spanish, German, French, Chinese, Hindi, or plain English. Lexicon translates and phrases it into natural English across six distinct voices. Auto preserves the exact register of your source text without forcing style shifts.
"Estoy muy contento con los resultados de Maria en 2024."
I am very pleased with Maria's results in 2024.
LaTeX math, code, tables & tasks
Inline and block LaTeX with live preview, syntax-highlighted code, task lists, and font controls.
A dictionary that learns from you
Save brand names, domain terminology, and unique jargon so proofreading never flags them again. The desktop app and extension share the list on your device.
Fully customizable keyboard shortcuts
Remap document proofreading, suggestion navigation, quick actions, and distraction-free focus mode with real-time conflict detection.
Write → Proofread → Polish → Export
Export what you just styled
Import from Word or Markdown. Ship PDF, EPUB, HTML, or DOCX with tracked grammar suggestions, or ask Lex to style the page.
Proofread across supported web fields.
Lexicon for Chrome & Firefox checks supported visible editable fields, including supported frames, on pages where you grant access. Checks run automatically by default until you pause or disable Lexicon for that site. It uses the same local-first desktop app and can forward text to a server you configure.
Beta build. Pairs with the desktop app (v0.11.0-beta.1+). It requests optional HTTP/HTTPS page access; when granted, it can inspect supported visible editable fields in frames. It does not work with stable v0.10.3, and the desktop app must be running.
Found a bug? Please report it on Discord.
Real-time in-page squiggles
Spelling and grammar issues are highlighted directly in text fields, textareas, and rich editors as you write, with zero cloud dependency.
Express in English, anywhere
Highlight text in any web field and phrase it as natural English across six tones, starting with Auto to keep your source register intact.
One-click inline fixes
Interactive suggestion chips float over flagged words with an Apply Fix button. No copying text back and forth.
Shared personal dictionary
Custom brand names, technical jargon, and domain vocabulary sync bidirectionally between your browser and desktop app via local loopback.
In-browser tone shifts & rewrites
Highlight any passage in an email, post, or forum reply to shift tone (Casual, Professional, Friendly, Formal, Concise) in place.
Deep Shadow DOM & rich editors
Accurately traverses complex web editors and nested shadow roots without trapping keyboard focus, with instant offline status alerts.
Install in Chrome
- Download the Chrome extension .zip from the button above (or GitHub Releases).
- Unzip it to a folder you will keep. Loaded unpacked extensions must stay on disk.
- Open chrome://extensions, enable Developer mode (top-right).
- Click Load unpacked and select the unzipped folder (open the inner zip folder if you see one).
- Open chrome://extensions/shortcuts and make sure Proofread the focused text field is set to Alt + Shift + L.
Install in Firefox
- Download the Firefox extension from the button above (or GitHub Releases).
- If you got a
.zip, unzip it and keep the folder on disk. - Open about:debugging → This Firefox.
- Click Load Temporary Add-on and pick the downloaded
.xpi, or themanifest.jsoninside the unzipped folder. - The add-on stays active until you close Firefox. The package is unsigned for now.
Known limits (beta)
- Google Docs is not supported yet.
- The Firefox package is unsigned and requires Firefox 142 or later. Temporary install only.
- The desktop app must be running (it can stay in the system tray).
- Packaged desktop builds enable operating-system startup by default and may check for updates in the background.
- The status badge turns gray when the desktop app is unavailable.
- Beta builds may be unstable. Report issues on Discord.
The same quiet desk, tuned to you
Four typeface pairings, five paper surfaces, and two reading modes. The same appearance controls as the desktop app, live on this page.
A quiet desk for your thoughts.
Lexicon proofreads when enabled and can rewrite with a downloaded local model on supported builds. Built-in proofreading and local models run on your machine; configured remote servers are an explicit exception.
Pick a typeface pairing, warm up the paper, or switch to Bionic or OpenDyslexic. The page follows you instantly.
A quiet desk for your thoughts.
Writing is personal. Cloud writing tools turned document editors into monthly subscriptions, background telemetry, and remote server queues. We built Lexicon because we wanted a quiet, distraction-free desk on our own machine, where prose is polished locally by default, data stays private by default, and software never asks for a login.
Built for people who prefer not to send drafts to someone else’s server; configured endpoints remain your choice.
Zero First-Party Telemetry
No tracking scripts or in-app analytics. Built-in writing processing stays local; configured processing servers are opt-in. Updates and model downloads may use the network.
Lexicon Code (MIT)
Lexicon-authored code is MIT-licensed. Bundled dependencies and model weights have separate licenses.
No Subscription
No monthly fees, hidden paywalls, or trial limits. Built out of passion for independent software.
Your prose belongs to you
No Lexicon cloud storage, no first-party tracking, and no subscription fees. Lexicon runs locally by default, with network exceptions for downloads, updates, and servers you configure.
Local Profile Storage
LOCAL BY DEFAULTYour drafts, personal dictionaries, custom settings, and local writing metrics are stored in your local user profile. No Lexicon cloud sync, mandatory account creation, or first-party background telemetry.
Embedded LanguageTool Engine
LOCAL BY DEFAULT
Grammar, spelling, and style rule passes run deterministically on your local CPU via the embedded runtime in packaged builds. A source or development user can configure a remote LanguageTool-compatible endpoint.
Local-First LLM Inference
LEX'S ENGINELex, Lexicon's writing assistant, can use a downloaded local GGUF model (Light ~1.15 GB, Standard ~3.0 GB, or Quality ~16.5 GB) on supported builds, or use your Ollama or LM Studio server. A network URL you configure may receive transform text and prompts; Windows x86 and ARM64 builds do not include the native GGUF engine. See the engine benchmark for how LanguageTool and local tiers score on grammar and fluency.
Complete Data Ownership
OPEN STANDARDS
Import `.docx`, `.md`, `.txt`, and `.html` documents. Export to PDF, EPUB, styled HTML, or DOCX. Remove downloaded models in Settings and manage drafts, history, and dictionary entries in the app. A full local-data wipe may still require browser or operating-system controls; locked history and the current document are not removed by the history controls.
Native GPU acceleration with dynamic VRAM layer offloading
Lexicon automatically detects NVIDIA CUDA, Apple Metal, Windows DirectML, AMD ROCm, and Vulkan on launch.
The Hardware tab calculates optimal layer offloading (n_gpu_layers) based on your free VRAM headroom, delivering 5-10x generation speedups while preventing out-of-memory crashes. If no GPU is available, Lexicon falls back smoothly to multi-threaded CPU.
Run bundled local models, Ollama, or LM Studio
Choose the setup that fits your machine. Switch between native bundled GGUF tiers or connect to your own existing local LLM server.
Why Writers Choose a Free Open Source Grammarly Alternative
Why local-by-default processing protects your privacy, budget, and intellectual property. For measured quality, see the GEC & fluency benchmark.
| Feature | Lexicon | Cloud Alternatives (Grammarly, ProWritingAid, Quillbot, etc.) |
|---|---|---|
| Privacy & Cloud Storage | Local by default / No Lexicon cloud storage | Drafts sent to third-party servers |
| Pricing | Free; Lexicon-authored code is MIT-licensed | $12-$30 / month subscription |
| Grammar Engine | Embedded LanguageTool | Proprietary cloud API |
| AI Processing | Downloaded local or configured server (llama.cpp / Ollama / LM Studio) | External cloud models |
| Internet & Account Needed | No account; network optional for updates, downloads, or configured servers | Requires active internet & account login |
Questions writers ask
Is Lexicon free to download and use?
Yes. Lexicon is MIT-licensed open source: no subscription, no account gate, and no trial that expires. Use it personally or commercially.
What operating systems can I install Lexicon on?
Windows (x64, ARM64, and x86), macOS (Apple Silicon and Intel), and Linux via .deb.
Pick the matching build from the download button on this page.
Does Lexicon upload my writing to the cloud?
Not to a Lexicon-operated cloud service. Built-in proofreading and local AI can run on your machine; a remote endpoint you configure may receive the text sent to it. Lexicon has no accounts or first-party writing analytics.
Do I need a Lex model just to proofread?
No. Grammar and style checks work without an AI model. Download a local Light (~1.15 GB), Standard (~3.0 GB), or Quality (~16.5 GB) model on supported builds when you want Lex rewrites, or connect your own Ollama or LM Studio server. A configured network server may receive the text and prompt sent to it. For how the local tiers compare on grammar and fluency tasks, see the benchmark results.
Can I proofread without an internet connection?
Yes. The packaged LanguageTool engine handles spelling, grammar, punctuation, and style locally when you run a check, so proofreading can work offline. A configured remote LanguageTool-compatible endpoint is an exception.
Is there a browser extension?
Yes. A beta browser extension for Chrome and Firefox proofreads multiple visible text fields, including supported frames, through the desktop app. It also offers Tone and rewrite actions from field badges. Google Docs is not supported yet, the Firefox package is unsigned (temporary install only), and the desktop app must be running. See the Extension section for downloads and install steps.
Why does Windows or macOS warn that Lexicon is unsigned or unsafe?
Desktop builds are not yet code-signed (Windows) or notarized (macOS). SmartScreen, Gatekeeper, and some antivirus tools warn on new or unsigned indie installers. That is a reputation / signing gap, not proof the file is malware. Only download from GitHub Releases, then follow the SmartScreen / Gatekeeper steps under Setup & Support.
Why use Lexicon instead of a cloud grammar checker?
Most cloud checkers bill monthly and process your documents on their servers. Lexicon is a free desktop writing app with offline grammar and local AI options. Updates, model downloads, and any server you configure are separate network paths.
Can Lexicon translate my writing into English?
Yes. Express in English takes a short thought in Spanish, German, French, Chinese, Hindi, or plain English and phrases it as natural English in six tones, starting with Auto, which keeps the source register. It runs on the Standard or Quality local model, or your Ollama or LM Studio server. Replace swaps it in with one undo.
Do I need a GPU to use Lexicon AI?
No. Lexicon runs on CPU alone; a graphics card simply speeds up local AI. On supported builds Lexicon detects CUDA, DirectML, ROCm, Vulkan, or Apple Metal hardware, offloads model layers to fit your video memory, and the Hardware tab in Settings shows your device and memory use. As a rough guide, Light fits about 2 GB of video memory, Standard about 6 GB, and Quality about 18 GB.
Can I change Lexicon's keyboard shortcuts?
Yes. The Shortcuts panel in Settings lets you remap proofread, suggestion, formatting, and focus commands, and warns about conflicts. In the browser extension, Alt+Shift+L proofreads the focused field.
Setup & Support
Step-by-step installation guide, OS launch verification, and issue tracking.
3-Step Setup Walkthrough
From download to your first draft in under 2 minutes.
Get the Installer
Click Download for Desktop at the top of this page to get the installer for your OS (Windows `.exe`, macOS `.dmg`, or Linux `.deb`).
First Launch Pass
Run the installer. Builds are not yet code-signed, so Windows SmartScreen or macOS Gatekeeper may warn. Follow the steps below (and allow antivirus false positives only if you downloaded from GitHub Releases).
Begin Drafting
Open Lexicon, choose your optional model, enable Proofread, and run a check to review grammar fixes and suggestion cards.
Windows (SmartScreen & antivirus)
Windows builds are not yet Authenticode code-signed. SmartScreen often shows "Windows protected your PC" for new or unsigned indie installers. That is a reputation warning, not proof the file is malware. Only install builds from GitHub Releases.
SmartScreen:
- Click More info under the warning prompt.
- Click Run anyway to complete installation.
Antivirus (Defender, AVG, etc.): some tools may quarantine the installer or flag “malicious code” as a false positive on unsigned NSIS installers. Confirm the file came from the official Releases page and that the asset name matches the selected build, then restore or allow it in your AV quarantine settings.
macOS (Gatekeeper)
macOS builds are currently unsigned / not notarized. Gatekeeper may show "Lexicon cannot be opened because it is from an unidentified developer", or block the background engine. That is an unsigned-developer warning. Only download from GitHub Releases.
- Option 1 (GUI): Control-click (right-click) Lexicon.app → select Open → click Open.
- Option 2 (System Settings): Open System Settings > Privacy & Security → scroll down and click Open Anyway.
- Option 3 (Terminal): Run xattr -cr /Applications/Lexicon.app in Terminal to clear the quarantine flag.
Linux (Debian / Ubuntu)
To install the .deb package on Ubuntu, Debian, Linux Mint, or Pop!_OS:
- Double-click the downloaded
.debfile to install via Software Center. - Or run sudo dpkg -i Lexicon_*.deb (or sudo apt install ./Lexicon_*.deb) in Terminal.
Found an issue, have a suggestion, or want to chat?
Join the Lexicon Discord for bug reports, feature requests, and community help.