You know the loop. A task pops up — find files over 100 MB, kill the process on port 8080, rsync a folder excluding node_modules — and your hands hesitate. Was it find -size +100M or +100MB? Is it lsof -i :8080 or ss -ltn? You alt-tab to a browser, paste the question into a chat, copy the answer back, and squint at it before pressing Enter.
CliAI removes the round trip.
What it is
CliAI is a single binary — clai — that lives in your $PATH. You describe a task in plain English (or in your system language). It returns a shell command, shows you what it does, and waits. You press Enter to run it, R to refine, or Esc to cancel. The command runs in your shell, in your terminal, with your environment.
Why it isn't a chat
A web chat breaks the terminal flow. It can't see your shell, your aliases, or your working directory. It can't pipe its answer into your clipboard reliably. Most importantly, it puts a generated command in front of you with no context for what's about to happen.
CliAI is the opposite: stay in the shell, see the command before it runs, decide whether you trust it. We call it inspect-then-run. It's the only mode the tool supports.
Install in one line
$ curl -LsSf https://gate.cliai.tech/install.sh | shDetected: linux x86_64Installed: /usr/local/bin/claiRun `clai --help` to get started.
The same one-liner works on macOS. Windows users get a native PowerShell installer — see the download page.
Where to go next
- Read the docs for the full subcommand reference.
- The next post explains why we picked a CLI over a chat box — and what we lose and gain by that.
- If you're worried about what gets sent over the wire, every request is encrypted; the docs walk through the model.
You'll spend less time typing the same five find flags from memory and more time thinking about what you actually want to do.