Turn any GitHub CLI tool into a point-and-click interface — automatically.
CLUI lets non-technical users run command-line tools without touching a terminal. Search for any CLI project on GitHub, install it, and the app analyzes its interface using AI to generate a custom GUI. No configuration, no coding.
- Search — find any CLI tool on GitHub (or describe what you want and let AI suggest options)
- Install — CLUI clones the repo and builds a Docker container automatically
- Use — AI analyzes the CLI and generates a point-and-click interface tailored to that tool
Every tool runs in an isolated Docker container. No setup, no dependency conflicts, no system pollution.
- AI-generated UI — Claude reads the tool's
--helpoutput and README to generate form-based workflows with the right input types (file pickers, dropdowns, sliders, toggles) - Finder mode — choose a folder or a single file and CLUI explains what it sees, offers safe actions, and suggests matching tools in plain language
- New Use Case — describe a new task in plain language and AI adds a new workflow tab on the fly
- Improve UI — give feedback and Claude regenerates the interface to match your needs
- AI repo recommendation — describe what you want to do ("merge PDFs", "compress videos") and get curated GitHub suggestions
- Auto-fix — if a command fails, AI diagnoses the error and suggests a corrected command
- Dark / light mode — clean neutral design, no distractions
- Output panel — live stdout/stderr stream with progress detection and copy-to-clipboard
- macOS (Linux/Windows support planned)
- Docker Desktop — must be installed and running
- Anthropic API key — for UI generation and AI features (get one here)
- Node.js ≥ 20 (for development only)
# 1. Clone
git clone https://github.com/florianmodel/clui.git
cd clui
# 2. Install dependencies
npm install
# 3. Start
npm run devOn first launch, open Settings and enter your Anthropic API key.
- Click Add New Tool in the sidebar
- Search for any CLI project (e.g. "ffmpeg", "imagemagick", "yt-dlp")
- Click Install — the app clones the repo and builds a Docker image
- Once installed, click Generate UI — AI inspects the CLI and creates a workflow interface
- Click Finder in the sidebar
- Choose either Choose folder… or Choose file…
- Let CLUI explain what it found in plain language
- For folders, use the main suggested action or open a matching tool
- For files, edit safe Finder-style details like tags, locked state, and file extension visibility, then open a matching tool if needed
| Tool | What it does |
|---|---|
| FFmpeg | Video/audio conversion |
| ImageMagick | Image processing |
| yt-dlp | Video downloader |
| Pandoc | Document conversion |
| ExifTool | Metadata editor |
Any tool that has a --help flag or argparse/click-based CLI is a good candidate.
clui/
├── packages/
│ ├── main/ # Electron main process (Node.js)
│ │ ├── analyzer/ # CLI introspection + LLM UI generation
│ │ ├── docker/ # Container lifecycle (dockerode)
│ │ ├── executor/ # Command builder + file I/O
│ │ └── ipc/ # IPC handlers
│ ├── renderer/ # React UI (Vite)
│ │ └── components/ # DynamicGUI, ProjectBrowser, LogPanel…
│ └── shared/ # TypeScript types (IPC channels, UISchema)
└── docker/ # Test Dockerfiles
- Main process — manages Docker, runs analysis, handles all file system and network access
- Renderer — pure React UI, communicates with main via typed IPC channels
- Containers — run with
NetworkMode: none, auto-removed after exit, isolated from host
- Electron — desktop shell
- React + Vite — UI
- dockerode — Docker API client
- Anthropic SDK — AI features (claude-haiku-4-5)
- TypeScript throughout, npm workspaces monorepo
npm run dev # Start Electron + Vite dev server
npm run build # Production build
# After editing packages/main:
npm run build -w packages/main # Rebuild main
# Then restart Electron
# Tests
npm test --workspace=packages/main # 105 unit tests (vitest)MIT