Sample content · easy to replace
explainer
What is an AI agent? A plain-language map
Sample explainer of what people mean by AI agents, how they differ from a chatbot, and which questions to ask next.
- Byline:
- AI Agents Info Editorial
- Published:
- Last updated:
Sample article. This page exists to demonstrate the blog template. Replace the file in
content/blog/with a reported piece. Claims below are generic and point to public sources.
The short version
An agent is a system that is given a goal and can take actions in some environment to pursue it. In today's products that environment is often a computer: a browser, a terminal, a shopping cart, a classroom tool, or an API.
A chat box that only writes text is usually not an agent in this sense. It can still be useful. The extra pieces — tools, a loop, memory, and stopping rules — are what people now call a harness.
A picture in four parts
Most public write-ups of LLM agents share the same sketch:
- A language model that proposes the next step in language.
- Tools the model is allowed to call (search, code, a browser, a store).
- A loop that feeds results back until the goal is met or a limit is hit.
- Guardrails: permissions, budgets, human approval, logs.
If you take away the loop and the tools, you have a chatbot. If you add them carelessly, you have a system that can take actions you did not intend. That is why harnesses and evaluation matter as much as the model.
What “autonomy” actually means here
Autonomy on this site is a sliding scale, not a personality trait. A coding assistant that waits for you to accept a diff is barely autonomous. A shopping agent that can place an order is more so. Neither is “unsupervised” in the science-fiction sense: someone chose the tools, the limits, and the logs.
Questions worth asking
When a product says it is an agent, ask:
- What goal is it optimizing, in one sentence?
- What actions can it take without asking?
- How do you inspect what it did?
- What benchmark, if any, is being cited — and does it match the job?
Those questions travel from a classroom to a research lab. The glossary and the research desk go deeper on each one.
Where to go next
- How agent harnesses work — the loop around the model.
- Curious Kids — the same idea in smaller words.
- Directory — products that call themselves agents, with sample listings only for now.