What Are AI Agents?

What Are AI Agents?

An AI agent is a system where a large language model operates in a continuous loop — taking actions, observing results, and deciding what to do next until a goal is accomplished. This is a fundamental shift from the chatbot model.

The Core Loop

1. 1.Perceive — The agent receives input: a user goal, environment state, or output from a previous step
2. 2.Think — The LLM reasons about what action to take next
3. 3.Act — The agent calls a tool, writes a file, runs code, or queries an API
4. 4.Observe — The result feeds back into the model's context
5. 5.Repeat — Until the task is complete or a stopping condition is met

A chatbot answers. An agent does.

The ReAct Pattern

ReAct (Reasoning + Acting) is the foundational pattern. The model alternates between Thought (reasoning) and Action (tool call). Modern APIs encode this implicitly — the model reasons internally before emitting a structured tool call.

Agents vs. Chatbots

| Chatbot | Agent | |---|---| | Single request → single response | Multi-step loop until goal is met | | Stateless | Maintains working memory | | No external actions | Calls tools, APIs, runs code |

Real-World Examples

Claude Code — Reads your project, writes code, runs tests, observes failures, iterates — all from one natural language instruction.

Customer Service Agent — Looks up an order, checks shipping via API, drafts a resolution, logs the interaction.

Research Agent — Searches the web, reads pages, synthesizes information, produces a cited report.

Why Now?

Three things converged: models got good enough to reliably choose correct tool calls; context windows grew to 100K-200K tokens; and tool use became a first-class API feature across Anthropic, OpenAI, and Google.