Skip to content

FAQ

General

What is Smart-Shell?

A terminal assistant that turns plain English into Bash or Zsh commands, checks them for risk, and runs them after you confirm.

Which AI providers does it support?

Any OpenAI-compatible chat API. Presets exist for Gemini, OpenAI, DeepSeek, OpenRouter, Groq, Ollama and LM Studio; custom accepts any base URL. See Installation.

Can I use it without an API key or offline?

Yes. Use Ollama or LM Studio; see Using local models (Ollama).

Which shells and systems are supported?

Bash and Zsh on Linux, macOS and WSL.

Is it safe?

Every command is classified as blocked, high, medium, info_leak or safe. Blocked commands never run; high and medium ask for confirmation. The rules are pattern-based, so always read the preview. See Safety.

Setup

How do I configure a provider or API key?

Run smart-shell setup. You can also set SMART_SHELL_API_KEY and SMART_SHELL_BASE_URL, which override the saved config.

Where is the configuration stored?

~/.config/smart-shell/config.json (mode 0600), with the keys provider, base_url, api_key and default_model. History and the error log are in ~/.local/share/smart-shell/.

Does Smart-Shell store my sudo password?

No. sudo prompts in the terminal. Passwords stored by older versions are removed from the config automatically.

smart-shell: command not found after installing

Restart the terminal or run source ~/.bashrc (or ~/.zshrc). Make sure ~/.local/bin is on your PATH (uv tool update-shell or pipx ensurepath).

pipx or PEP 668 errors during install

Newer distros block pip install into the system Python, which is why the installer uses uv or pipx. Install either one (uv, or sudo apt install pipx), open a new terminal and rerun the installer.

Setup says it cannot connect

Check the API key and base URL. For Ollama, make sure ollama serve is running and you have pulled at least one model.

Usage

How do I switch models?

smart-shell models or /models lists what the provider offers. Use --model <name> for one run, /model <name> in interactive mode, or smart-shell setup to change the default.

Can I use it in scripts?

Yes, with smart-shell run "<prompt>". Add -y to skip confirmations (blocked commands still never run). Review generated commands carefully before automating them.

How do I update?

Use /update in interactive mode, or uv tool upgrade smart-shell-cli / pipx upgrade smart-shell-cli.

What is sent to the AI provider?

Your prompt plus basic OS and shell information so the command fits your system. In interactive mode, your last five requests from the same session (prompt, generated commands, and whether they succeeded) are included so follow-ups work. Command output and file contents are not sent. With a local model, nothing leaves your machine.

What does it cost?

Smart-Shell is free and open source (Apache 2.0). Cloud providers bill according to their own pricing; local models cost nothing.

Contributing

See Contributing and the Development Guide. Bugs and feature requests go to GitHub Issues.