Back to AI Builder Studio

Single Agent Systems

Learn to build your first AI agent - from basic concepts to working code

What is an AI Agent?

The building block of modern AI applications

|

Imagine you had a very clever assistant who could read anything, write anything, and use any tool you give them - but they can only work when you ask them something. That is essentially what an AI agent is.

Unlike a simple chatbot that just answers questions, an agent can take actions. It can search the web, send emails, update spreadsheets, book appointments, and more. This transforms AI from a "question answering machine" into a genuine assistant.

Examples
  • •A customer service agent that can look up orders and process refunds
  • •A research agent that searches multiple sources and summarises findings
  • •A scheduling agent that checks calendars and books meetings
  • •A coding agent that writes, tests, and fixes code

An AI agent is an LLM plus tools. That is the entire concept.

Agent = LLM + Tools + Instructions

The LLM (like GPT-4 or Claude) provides the "brain" - the ability to understand language and make decisions. The tools provide the "hands" - the ability to actually do things in the world. The instructions tell the agent what its job is and how to behave.

LLMToolSystem Prompt

Every agent follows the same basic loop:

  1. Receive - User sends a message or task
  2. Think - LLM decides what to do next
  3. Act - If needed, use a tool to take action
  4. Observe - See the result of the action
  5. Respond - Either repeat (more actions needed) or reply to user

This is sometimes called the "ReAct" pattern - Reason, then Act.