TuxAideAsk your terminal in plain English.

Type a Linux question as if it were a command. A model on your own computer answers right there — and it never runs anything on its own.

Install and first question

How does it know it's a question?

GitHub

No trigger word

Type the question itself. Real commands, aliases and functions run exactly as before.

The command, on your prompt

Commands in an answer are numbered. Press 1 and it's on your prompt, ready to edit — you press Enter. Destructive ones are flagged first.

? explains the last error

A command failed? Type ? and TuxAide reads the error (asking before re-running anything that isn't read-only) and tells you the fix.

Typos fixed instantly

gti status → git status, cd.. → cd .., without asking the model. Your distro's own "install it with apt" hint still works.

Answers from your man pages

With Smart RAG, questions about a command are answered from the man pages installed on your system, with the source cited.

Local by default

A model on your machine through Ollama. No account, no API key, no telemetry. A remote backend is optional — and marked ☁ on every answer.

See it in 90 seconds

A real terminal and a real model (qwen2.5-coder:7b on Ollama): a question and its command put on the prompt, a follow-up with a warning before deleting, a question in Portuguese, ? explaining a failed command, a typo fixed, and tuxaide doctor. No sound; a caption opens each scene. How the video is made.

Who is it for?

People learning Linux, and anyone who doesn't remember every flag. Instead of leaving the terminal to search, you ask where you already are:

bash

$ how do I find files bigger than 100MB in this folder
$ and how do I delete them?
$ como vejo o espaço livre no disco
$ ?            # after a command that failed

It answers in the language you wrote in, with commands for your system (dnf on Fedora, brew on macOS).

Install in one command