Stop memorizing flags. Just say what you want.
Type a task in plain English and see the exact shell command clai generates — with its risk level — then run it right here. You review every command; nothing executes until you say so.
Real cliai output — you review every command, then run.
why it sticks
Why people keep it in their shell
Never google a flag again
Describe the result you want — clai writes the exact command, flags and pipes included. No more digging through man pages or old Stack Overflow answers.
Stay in your terminal
No browser tab, no chat window, no copy-paste. Your request, the command, and its output all stay in the same session.
See and control every command
Every suggestion comes with a plain-English description and a risk level. You read it, refine it, and run it only when it matches your intent.
One install, everywhere
A single client for Linux, macOS, and Windows. Installs into your home directory without sudo and updates itself with clai update.
get started
Up and running in minutes
Ask in plain words
clai show the 10 largest files in this folder
Check the command
# read the command + risk level, press r to refine
Run it
# Enter runs it here — output stays in your terminal
what you type
Ask for the thing. Get the command.
A few real prompts and what clai hands back.
find . -type f -size +100M
SAFE
lsof -ti :8080 | xargs kill
CAUTION
git log --oneline @{u}..HEAD
SAFE
tail -f /var/log/nginx/error.log
SAFE
0 9 * * 1-5
SAFE
df -h
SAFE
Every command is shown with its risk level before anything runs.
safety model
You see and approve every command
Each suggestion is risk-classified before it touches your shell. Nothing runs until you confirm — and the most dangerous operations take more than a keystroke.
Checks & diagnostics
Read-only commands — listing, searching, inspecting. Review and press Enter to run.
Changes you confirm
Commands that modify files or state. You see exactly what will run and confirm it before it does.
Destructive — gated
Irreversible operations require you to type "yes" in full, and reckless patterns are blocked outright.
questions
Straight answers
What is CLI AI for?
You describe what you need in plain language and clai gives you the exact shell command to run — handy for quick checks, troubleshooting, and the CLI tasks where you would normally stop to look up syntax.
How is this different from asking AI in another tab?
There is no second window. Your wording, the suggested command, and its output all stay in one terminal session — you never copy commands out of a browser chat.
Do I need to sign up or create an account?
No. The client registers itself on first run and stores a local recovery secret, so you can start asking right away — no forms, no logins.
What is sent to the service?
Your task text, any file context you attach (for example paths passed with @), and refinement messages — only what is needed to generate or adjust the command. Requests travel encrypted end to end.
Can a risky command run without me?
No. You always see the command and its risk level first. Destructive operations require you to type "yes" in full, and reckless patterns are blocked outright.
Which systems are supported?
Linux, macOS, and Windows. Install with one command from the home page, or grab a checksum-verified binary from Download.
How do I update the client or check usage?
clai update fetches the latest client, and clai stats shows your usage. Every release is published with checksums and matching documentation.
Spend less time looking up commands
Say what you need in plain English, read the command clai suggests, and run it in the same session — no copy-paste from a side chat.
curl -LsSf https://cliai.tech/install.sh | shirm https://cliai.tech/install.ps1 | iexYou confirm before anything runs • Checksum-verified downloads • Docs tied to every release