Command Line Interface (CLI)

Most people operate a computer with a mouse and finger taps. They see icons, windows, and buttons and click on them. But there’s another way: you type a command as text and press Enter. The computer executes it and writes the result back as text. This kind of interaction is called a Command Line Interface, or in German Befehlszeile or Kommandozeile. The window where this happens is usually black and contains only text.

Why professionals type voluntarily

At first glance, this looks like a step backward into the 1980s. But the decisive advantage is repeatability. A sequence of clicks through five windows can’t be passed on to someone else exactly. A command is a line of text that can be copied, sent, and executed identically a million times over.

This leads to the second advantage: automation. You can write several commands into a file, called a script. The computer then works through them on its own, every night at three in the morning, without anyone sitting beside it. This is exactly how large server farms are operated. Nobody clicks their way through ten thousand machines.

A third reason is speed with large quantities. Anyone wanting to rename 4000 image files would need a full day using a mouse. With a command, it takes a second. The price for this: you have to know the commands. A command line does not, on its own, show you what’s possible.

What happens between Enter and the result

Behind the text window runs a program called a shell. It reads the typed line and breaks it into components. Almost always, the name of a program comes first. This is followed by additional specifications called arguments and options. The shell starts the program and passes it these specifications.

An example: in the line “copy report.txt backup.txt”, “copy” is the program for copying. The two file names tell it what to copy where. Options often begin with a hyphen and toggle a behavior on or off. The program writes its result back as text into the same window.

This becomes especially powerful through chaining. You can feed the text output of one program directly into a second program. This is called a pipe, written as a vertical bar. In this way, a few simple tools combine into complex workflows. Incidentally, a common misconception is that the command line and the operating system are the same thing. It is only one of several doors into the same system.

From developer terminal to AI tool

Anyone who writes code spends a lot of time in the command line. Software is compiled, tested, and deployed to servers there. Version control with the tool Git also traditionally runs via typed commands. Every operating system comes with a command line; on Windows it’s called PowerShell or Command Prompt, on macOS and Linux it’s called Terminal.

The term currently comes up especially often in the AI world. Many providers publish a CLI tool alongside their website. This allows a language model to be addressed directly from the terminal, without a browser. Developers use this to integrate it into their own scripts. That’s why announcements about new models almost always mention whether such an interface exists.

What’s new is that AI systems are now operating the command line themselves. So-called coding agents are given access to a terminal and type commands independently. They create files, run tests, and read the error messages. The command line has thus turned from a purely human interface into an interface for machines.