Reply Better AI

Improve and reply to messages with AI. Free and private, no key: on-device or a local model (Ollama/LM Studio). Or your own key.

Reply Better AI helps you write better anywhere on the web. It rewrites your drafts and writes replies in context, using the AI engine of your choice, and it is free to use with no setup. ENGINES (pick one, or leave it on Auto) • On-device, free and private. Runs locally in Chrome using the built-in Gemini Nano model. No key, no account, and your text never leaves your computer. Needs a supported Chrome desktop, and the model downloads once on first use. • Local (Ollama / LM Studio), free and private. Run your own model on your machine with Ollama, LM Studio, or any OpenAI-compatible server. No key, works offline, and nothing leaves your computer. Set a preset or your own server URL in settings. • Groq, free and fast. Add your own free Groq key for quick cloud generation. • OpenRouter, 500+ models. Use Claude, GPT, Gemini, DeepSeek, Llama and more, free or paid, with a searchable picker, live per-token pricing, and context-window badges. Auto uses on-device when your device supports it, then your Groq key, then OpenRouter, and it falls back automatically if one is busy. (Local is opt-in: pick it in settings when you want to run your own server.) The active engine is always shown, inline and in settings, so you know exactly what is running. KEY FEATURES • One button that morphs to fit what you are doing. Select the messages you are replying to and it becomes a Reply button; start a draft and it becomes an Improve button. It only shows on real message composers, so it stays out of search bars and form fields. • Reply in context. Pick a tone (Match thread, Professional, Friendly, Concise, or Summarize) or type your own instruction in any language, and the reply comes back in that language. • Live streaming results. Watch the result type itself out as it is generated, then flip to the Changes view to see exactly what was edited, word by word. Want another take? Hit Regenerate and cycle through versions, then Insert with one-tap Undo. • Auto, Fastest free. One pick that routes to the fastest available free model and fails over automatically when one is busy. • Dynamic model picker. Search the live OpenRouter catalog inside the extension, with three tabs (Popular, Free, All), a provider filter, live per-token pricing, and context-window badges. The list refreshes hourly so you never use a model that has been deprecated. • Multiple writing styles. Improve, Professional, Friendly, Concise, and Persuasive built in, or write your own custom prompts and set one as the default. • TextBlaze-style snippets. Define triggers like /addr or /sig and they expand to longer text wherever you type, no AI call required. • Refreshed design with dark mode, plus a standalone full-tab settings page in addition to the toolbar popup, so you can manage your engine, key, and default model without juggling popups. PRIVACY-FIRST The extension is local and runs no servers. Your keys, prompts, and snippets are stored in your browser's local storage, never synced to a cloud account. With the on-device or Local engine, nothing ever leaves your computer. With a cloud engine, only the text you select is sent, and only to the provider you chose. The only outbound network hosts are openrouter.ai and api.groq.com, plus your own local server if you use the Local engine. Zero telemetry, zero analytics, no third-party SDKs. Read the full policy at https://github.com/dantnan/reply-better-ai/blob/main/docs/privacy.md GETTING STARTED 1. Install the extension and open it from the toolbar. 2. Leave the engine on Auto. On a supported Chrome desktop it uses free on-device AI with no key. Otherwise, add a free Groq key at https://console.groq.com/keys or a free OpenRouter key at https://openrouter.ai/keys (no credit card needed for free models), or run your own local model with Ollama or LM Studio. 3. Paste your key into the popup, and pick a default model if you are using OpenRouter. 4. Either type into the popup textarea and click Improve, or focus any text composer on a website and click the button. OPEN SOURCE Source code, issue tracker, and full changelog at https://github.com/dantnan/reply-better-ai (MIT licensed). Built and bundled with esbuild, no obfuscation, every transformation reproducible from the public source.