LiveQuill
Edit your local website by asking.
A floating prompt bar for any local site: type a change, the model of your choice edits your real files, and the page reloads.
Website · Install · Models · FAQ
A real first run with Claude Code (Opus): a portfolio from a template, then “turn the Selected work section into a 3D gallery” with five project images, a quick color change and Undo. Waiting for the model is sped up. MP4 with music.
Why LiveQuill
- Ask on the page. Describe the change on the page you are looking at. The model finds the code, edits it, and the page reloads.
- Any model. Cloud APIs, local models, or the coding agent you already use, even if you only have the Claude Desktop or ChatGPT app.
- A new site in a minute. A short wizard starts from a template or a blank page. No technical skills needed.
- Nothing to lose. Every request is a version you can preview and go back to, with one-click Undo. Zero dependencies, nothing added to your project.
Install
Linux and macOS, with Python 3.10 or newer.
curl -fsSL https://ybucaille.github.io/livequill/install.sh | sh
No root, no git, no pip: the script checks and installs a single file, then creates the livequill command. If Python is missing, it tells you how to install it on your system.
| |
|---|
| pipx / uv | pipx install livequill or uv tool install livequill |
| Single file | Download livequill.pyz from a release and run python3 livequill.pyz |
| From source | git clone https://github.com/Ybucaille/livequill.git && ./livequill/bin/livequill |
| Uninstall | curl -fsSL https://ybucaille.github.io/livequill/install.sh | sh -s -- --uninstall (or pipx uninstall livequill) |
Update
Run the install command again: it replaces LiveQuill with the latest release and keeps your settings and sites.
curl -fsSL https://ybucaille.github.io/livequill/install.sh | sh
With pipx or uv, run pipx upgrade livequill or uv tool upgrade livequill. From source, run git pull. Check your version with livequill --version, and restart LiveQuill after updating.
Get started
livequill
- Choose a model. The first launch opens the setup page. Click a model detected on your computer, or paste an API key.
- Create a site. On My sites, click New site, pick a starting point (blank page, business website or portfolio), give it a name. Or add a site you already have.
- Ask. The site opens with the floating bar. Try “make the hero title red”, watch it change, then click Undo.
| Option | Description |
|---|
FOLDER | Add this folder to your sites (if needed) and open it. Without it, the dashboard opens. |
--proxy URL | Serve FOLDER behind your dev server (Vite, Next, Astro…) instead of serving its files. |
--port N | Dashboard port on 127.0.0.1 (default 8777). Each site gets its own port after it. |
--config FILE | Alternative config file (default ~/.config/livequill/config.json). |
--no-open | Do not open the browser. |
Screenshots
Models
| Providers | What you need |
|---|
| On your computer | Ollama, LM Studio, llama.cpp, any OpenAI-compatible server | The app or server running. Free and private. |
| In the cloud | Anthropic, OpenAI, Google Gemini, Mistral, OpenRouter | An API key, pasted on the setup page or read from your environment. |
| Coding agents | Claude Code, Codex, Gemini CLI | The CLI, or just the Claude Desktop or ChatGPT app, signed in to your account. |
Switch models at any time from the bar. Reasoning and Fast appear when the model supports them.
| Provider | Notes |
|---|
| Anthropic | Key from console.anthropic.com. Effort levels on Claude 4.6+, Fast on Opus 5.5 / 5 / 4.8. |
| OpenAI | Key from platform.openai.com. reasoning_effort on o-series and gpt-5. |
| Google Gemini | Key from aistudio.google.com. Thinking budget (2.5) or level (3). |
| Mistral | Key from console.mistral.ai. |
| OpenRouter | Key from openrouter.ai/keys. Capabilities are read for each model. |
| Ollama | ollama serve, then ollama pull qwen2.5:7b-instruct (any model that supports tools). |
| LM Studio | Developer tab → Start server. |
| llama.cpp | llama-server -m model.gguf --jinja. |
| Other | Any OpenAI-compatible server (vLLM, LocalAI, Jan, Groq, DeepSeek…): base URL, optional key, model name. |
| Claude Code | claude signed in, or the Claude Desktop app. Runs confined to the project, see SECURITY.md. |
| Codex | codex login, or the ChatGPT or Codex app. Runs in its sandbox, but can read your whole disk: see SECURITY.md. |
Privacy and safety
- Local only. LiveQuill listens on
127.0.0.1 and checks a session token on every call.
- Your keys stay on your computer. They never reach the browser, and are masked in logs.
- The model stays in your project. No path outside it, no
.git or .env files, no shell commands.
- Every change is backed up before it happens, so Undo always works, CLI agents included.
How each protection works is described in SECURITY.md, where you can also report a vulnerability privately.
FAQ
Roadmap
- Dynamic sites with a disposable demo database, so the model can work on data-driven pages safely.
- More templates and languages in the new-site wizard and the language menu.
- Point and ask: click an element on the page to scope the request to it.
- Review before writing: see the proposed changes and accept or reject them.
Contributing
Contributions are welcome: providers, themes, templates, translations and fixes. Start with CONTRIBUTING.md, which also explains how LiveQuill works inside.
License
MIT © 2026 Ybucaille
