standard kimenet

in English: standard output / standard error

advancedwidespread

The program has two separate text channels: one outputs the result, the other outputs error messages and warnings.

In plain terms

Every program you launch from the command line can output text through two separate "pipes". The standard output (stdout) carries what the program was launched for, i.e., the result. The standard error output (stderr) carries everything else: error messages, warnings, and remarks made during execution.

It is like a restaurant kitchen. The finished dish arrives through the serving window, while the chef calls out through another door if the fish is out of stock or the order is delayed by ten minutes. If both came through the same window, the waiter would put the complaint on the plate.

On the screen, the two usually appear mixed together, so you do not initially see that there are two channels. The difference matters when you save the output to a file or another program reads it: only the clean result should reach there.

For AI agents, this is a daily question. The coding agent reads both channels of executed commands: from the output it knows the result, and from the error output it knows what went wrong and what needs fixing.

How it works

When a program starts, it receives three open channels, usually inherited from the program that launched it (for example, the command line): standard input (stdin, number 0), standard output (stdout, number 1), and standard error (stderr, number 2).
The program itself decides where to write: the result to channel 1, errors and log messages to channel 2.
By default, both channels are bound to the terminal window, so you see them together on the screen.
Redirecting separates the two: results can go to a file or another program's input, while errors remain on screen or are saved to a separate log file.
At the end, the program also returns an exit code. By convention, 0 means success, while any other value indicates an error or a result deviating from the norm. An agent examines this code along with the content of the two channels to decide the next step.

An everyday example

You ask the coding agent to run the tests, and in its response you see: 'the command stopped with an error, the stderr indicates a missing package'. The agent read the error output to identify the problem, so in the next step it installs the missing package and tries again.

Who is affected

developerIt is used daily: scripts, logging, and agent-linked tools work reliably only if the result and the error message go through separate channels. → Code

employeeIf you work with a coding agent or automation, you regularly encounter the terms stdout and stderr in runtime reports and error messages. → Work & HR

entrepreneur, executiveThe agents' runtime logs consist largely of these two channels, which is why it becomes clear afterward what the system did and where it got stuck.

A common misconception

MythIf a program writes to stderr, then an error definitely occurred.

In realityMany programs send progress indicators, warnings, and log messages to stderr so they do not mix with the output. Whether a real error occurred is mostly shown by the exit code, but this is also just convention: some programs do not use a non-zero code to indicate, for example, that nothing was found.

Limits and risks

The two channels are governed by convention, not compulsion: some programs write errors to standard output, or vice versa.
Passwords, access keys, or personal data may appear in the output, and once something reaches the log or the agent, it is difficult to retract.
An agent can only review a limited amount of text at once, so the system truncates very long outputs, potentially losing important details.
If a tool connected to an agent communicates over the standard channels, a single stray log line on standard output can corrupt the entire data exchange.

Similar, but not the same

standard bemenet – Standard input (stdin) is the opposite direction: the program receives data through this channel, not outputs it.
log file – The log file persists on the disk, while standard output is transient data that is lost if no one saves it.
exit code – The exit code is a single number indicating success or failure, while the two output channels provide the details in text.

Related concepts

Broader

standardized data streams, command line, operating system

Narrower

standard hibakimenet

On the edge

standard bemenet, exit code, pipeline, logging, terminal

A brief history

The concept of standardized data streams spread in the 1970s within the Unix operating system. The separate error output appeared in the mid-1970s, in the sixth edition of Unix. It was introduced so that error messages would not mix with the results when processed by another program, but would appear on the user's screen.

Where things stand

For half a century, it has been a basic element in every widely used operating system. At the end of 2026, it has come back to the forefront due to agents: coding agents read command results on these two channels, and many locally running, agent-linked tools also communicate via them.

Status: 2026 November

Also known as

stdout · stderr · standard output · standard error · standard output · standard error output · standard hibacsatorna

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.