26
Lesson 26 · 2 min
AI agents: AI that takes action
An agent is AI that does not just answer but takes steps: it plans, uses tools, checks the result, and keeps going.
A chatbot answers. An agent acts. Give it a goal like "find three flats in Lisbon under 1,200 euros and put them in a table", and it will search, open pages, compare and write the table.
Agents work in a loop: plan the next step, act with a tool (search, browser, code, email), look at what happened, and repeat until done.
In 2026, many assistants have agent modes that browse the web, write and run code, or work inside your apps. They are useful for research and repetitive tasks, but still make mistakes and can go off track.
Keep a human in charge of anything that spends money, sends messages or deletes things. Try it below: step through an agent's loop.
Watch an agent work
An agent is an AI that loops: plan, use a tool, look at the result, decide. Step through it with Next.
Goal: Book a table for 4 near the office on Friday.
- Plan
- Act
- Observe
- Decide
Do it now
Give an agent a safe task
- 1.Pick a low-risk research task, like comparing three products or finding opening hours.
- 2.In an assistant with an agent or research mode, use the prompt below.
- 3.Watch its steps and check two of its results yourself.
Research [TOPIC]. Find [NUMBER] options that match [CRITERIA]. Put them in a table with name, price, key pros and cons, and the source link. Do not buy, book or send anything.
Words from this lesson
- Agent
- An AI system that takes several steps on its own, such as searching, clicking or using apps, to finish a goal you set.
- MCP
- An open standard that lets AI assistants plug into apps and data sources, like a universal adapter for tools.
- Prompt injection
- Hidden instructions planted in a web page, email or file that try to hijack an AI reading it.
- Guardrails
- Rules and filters around an AI system that block harmful, off-topic or unsafe requests and answers.