Evidence-first authority page
What Is an AI Agent? Definition, Components and Examples
A plain-language definition of an AI agent: what separates it from a chatbot, its core components, and how to recognize one in practice.
Direct answer
An AI agent is a software system that uses a language model to perceive a task, decide on a sequence of actions, and execute those actions through tools — APIs, code execution, file access — with limited or no step-by-step human instruction, then observes the results and adjusts. The defining trait is the loop: perceive, decide, act, observe, repeat, rather than a single request-response exchange.
Last reviewed 2026-08-29
This page remains noindex while its page-level source receipts are completed and verified. It is available for review, not presented as publication-ready authority.
How an agent differs from a chatbot
A chatbot answers a message with a message — the interaction ends when the model finishes generating text. An agent is given a goal, not just a question, and it can take multiple actions across multiple turns without a human re-prompting it at every step: searching a codebase, calling an API, checking the result, and deciding whether to try again or move to the next step.
The line is blurry in practice. Many products marketed as 'AI agents' are closer to a chatbot with one or two tool calls bolted on, while some products not marketed as agents (an autonomous CI pipeline that retries failed builds using an LLM to diagnose the failure) fit the definition closely. The presence of a tool-use loop, not the marketing label, is the reliable signal.
The core components
- A model that can reason about a goal and decide what to do next.
- Tools — functions, APIs, or system access — the agent can actually invoke, not just describe.
- A way to observe the result of an action (a return value, an error, a file diff) and feed it back into the next decision.
- Some form of memory or state, so the agent doesn't re-derive the same context on every step.
- A stopping condition — a defined goal state, a step limit, or a human checkpoint — so the loop terminates.
Where to see verified examples
This site's agents directory only lists agent identities that resolve to a primary source — an official repository, product page, or vendor documentation — rather than accepting the 'agent' label at face value. Capability, pricing and benchmark claims are published only where their own evidence exists; a listing with no attached evidence is marked accordingly rather than assumed to work as described.