·onboard.wiki

AI agent

An AI agent is an LLM connected to tools—it can search the web, write files, run code, or send emails. Unlike a chatbot that only talks back, an agent takes actions in the world, usually in a loop: think, act, observe, repeat.

In one sentence

An AI agent is an LLM that can do things, not just say things—it writes code, runs searches, and makes decisions in a loop.

What it's for

Agents automate multi-step tasks: build entire apps from a description (Claude Code, OpenClaw), book flights and hotels (early travel agents), or manage your calendar. They are where the industry is heading in 2026.

How to think about it

Think of an agent as a junior employee who can use a computer. You describe the goal; the agent breaks it into steps, tries things, fails, retries, and reports back. You still review its work—it’s not autonomous.

Common misunderstandings

People think agents are self-driving AI that needs no supervision. They are not. Every major agent framework (OpenClaw, Hermes, Claude Code) requires human approval for consequential actions. Agents amplify your output, they don’t replace your judgment.

If you want to try it

Install Claude Code or OpenClaw. Give it one simple coding task—“Build a page that counts button clicks”—and watch it write, test, and iterate. Stop it when it goes off track. That loop is the whole experience.

Sources