Desktop & terminal
Loxley runs in three places with the same engine: the web, a desktop app, and your terminal.
Desktop app
Download the installer for your system from the releases page:
| System | File |
|---|---|
| Windows | Loxley-x.y.z-win-x64.exe (installer) or Loxley-x.y.z-portable.exe |
| macOS | Loxley-x.y.z-mac-arm64.dmg (Apple silicon) or -x64.dmg (Intel) |
| Linux | .AppImage or .deb |
The desktop app is the full studio — editor, live preview, diagrams, docs — running privately on your computer. There is no sign-in; your projects stay on your disk. Open Settings to choose an AI provider.
Terminal
npm install -g loxley-cli
loxley setup
cd my-project
loxley
loxley starts a full-screen agent that works on the files in the current folder. It can read, search, create and edit files and run commands — and it asks first before changing anything.
| Command | What it does |
|---|---|
loxley | Interactive agent in this folder |
loxley run "prompt" | One request, answer streamed to stdout |
loxley studio | Open the visual studio locally in your browser |
loxley setup | Choose provider, key and model |
loxley models | List the provider's models |
Useful flags: --model, --dir, --yes (approve without asking). Inside the agent: /model, /models, /effort, /yolo, /clear, /cost, /help.
Put an AGENTS.md in your project root to give the agent project-specific instructions.
AI providers
Use any of these in the desktop app, the terminal, and self-hosted web servers:
- Anthropic — Claude Opus, Sonnet, Haiku and Fable.
- OpenAI and OpenRouter.
- Ollama and LM Studio — models running on your own machine, no key and no data leaving it. For example
ollama pull qwen3-coder. - Custom — any server that speaks the OpenAI
/v1/chat/completionsAPI (vLLM, llama.cpp, Together, Groq and more).
Choose models that support tool calling: the agent works by calling tools.