YAML
YAML is a simple, human-readable text format in which program settings and data are described, organized by indentation.
In plain terms
YAML is like a carefully filled form or shopping list: on the left is what it is about, on the right is the value. For example, "language: Hungarian" or "max_steps: 10". It contains no program code, only data and settings.
It is useful for telling a program how to behave without modifying the program itself. It is similar to a washing machine's program selector dial: the machine is the same, only the setting changes.
What distinguishes it from other similar formats is that it contains almost no parentheses or quotation marks. The structure is given by line indentation, just as you indent sub-points in an outline. This makes it easy to read, but even a misplaced space can cause an error.
In the AI world, you encounter it frequently because agents, automated workflows, and cloud system settings are very often stored in YAML files.
How it works
- The basic unit is the key–value pair: after the key name comes a colon and a space, then the value, for example "model: fast".
- You group related data by indentation: anything starting further in belongs to the line above. You cannot use tabs, only spaces.
- List items begin with a hyphen, and text after the hash (#) is a comment that the program ignores.
- When the program starts, it reads the file and builds its own internal data structure from the text: lists, numbers, yes/no values.
- If the indentation or the syntax is wrong, at best parsing stops with an error message. At worst, the program understands something different from the file than what you intended.
An everyday example
You download an open-source AI tool, and the instructions ask you to enter your own API key (the access code received from the provider) and the name of the chosen model into the 'config.yaml' file. You open it with a simple text editor, change two lines, save it, and the tool starts with your settings.
Who is affected
A common misconception
Limits and risks
- Indentation matters: a single missing or extra space can change the meaning or render the file unusable.
- Some values the program may interpret differently than you would expect: for example, according to the older standard (YAML 1.1) still used in many programs today, the word 'no' represents a false value, and the version number can become a decimal fraction (from 1.10 to 1.1) if you do not put it in quotes.
- YAML files often contain passwords and access keys, so never share or upload them to public places without review.
- Loading a YAML file from an unknown source can be a security risk if the program does not use a secure reading mode.
Similar, but not the same
- JSON – JSON describes similar data using curly braces and quotation marks; it is easier for machines but harder for humans to read, and does not allow comments.
- XML – XML places data between opening and closing tags, so it is much more verbose but can be validated more strictly.
- Markdown – Markdown formats text for humans (headings, emphasis, lists), while YAML describes machine-readable, structured data.
Related concepts
Data format, Configuration file, Data serialization
Front matter, Docker Compose file, Kubernetes-manifeszt
A brief history
YAML was introduced in 2001 by Clark Evans, who developed it together with Ingy döt Net and Oren Ben-Kiki. The name originally stood for 'Yet Another Markup Language', but later received the playful, self-referential expansion 'YAML Ain't Markup Language'. The currently valid 1.2 standard was released in 2009, and its latest revised edition (1.2.2) is from 2021.
Where things stand
As of late 2026, YAML is one of the basic formats in software development and cloud operations: deployment pipelines and containerized systems typically describe their settings in it. It is also common in AI tools, both in the descriptions of agents, skills and workflows and in the headers of Markdown files.
Also known as
YAML · yaml · .yaml · .yml · YAML file · YAML file · YAML Ain't Markup Language · Yet Another Markup Language
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.