> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xorafin.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Core concepts

Understanding the core concepts of Xorafin will help you quickly learn how the platform works and how to build AI agents effectively.

Xorafin is built around a few key ideas: **agents, inputs, outputs, models, and agent configuration files**.

***

## Agent

An **agent** is the core unit in Xorafin.

An agent is an AI-powered entity that receives an input, processes it using a language model (and optionally tools), and returns an output.

Agents can perform tasks such as:

* generating content
* summarizing information
* answering questions
* researching topics
* automating workflows

Each agent is defined by a configuration file that describes how the agent behaves.

***

## Agent File

Every agent in Xorafin is defined through an **agent configuration file**.

This file specifies:

* the agent name
* the model used
* the inputs the agent accepts
* the outputs the agent produces
* optional tools or settings

Xorafin supports multiple formats for defining agents:

* JSON
* YAML
* TOML

Example:

```json theme={null}
{
  "name": "research-agent",
  "description": "Research a topic and generate a summary",
  "model": "gpt-4",
  "input": "topic",
  "output": "summary"
}
```