AI Basics
AI agent
An AI system that can take actions to finish a task, such as searching, clicking, or sending a message, instead of only writing an answer in the chat.
- Parents
- Businesses
What parents should know
An AI agent does more than reply. It takes steps: it might browse, fill a form, or send a message. That is useful for an adult with a credit card and a bad idea for a child. Ask what the tool is allowed to do when nobody is watching the screen.
On this page
What is an AI agent?
A chatbot answers. An agent is allowed to act. The actions can be small, such as running a web search, or large, such as emailing someone or buying something. The product decides the list. If the list includes sending, paying, or posting, a child should not hold that login.
Marketing uses agent for ordinary chat. Read the settings. If the only action is a reply inside the window, you have a chatbot. If it can leave the window and change something in the world, you have an agent, and you need a parent gate on those actions.
Why an AI agent matters
Kids will click allow because the bot asked nicely. An agent that can message a stranger, spend money, or post a project publicly is a different safety problem from a wrong math hint. The mistake leaves the house.
A calm rule is action by action. Search may be on. Purchases stay off. Public sharing waits for a parent. You should not have to discover the permission after the message is sent.
How it shows up in practice
- A tool asks to connect email, and a parent says no for the child's profile.
- Web search runs inside the chat and shows sources, without sending a text to anyone.
- A public link to a kid-built game waits for parent approval.
- A teen product that can book or pay is kept on the parent's own login.
How HeyOtto helps
Otto is the assistant. It does not book travel, pay, or message people outside the account on its own. Inside HeyOtto, a prompt can use kid-safe search that cites sources, and a child can build a project. A public share of that project stays off until a parent approves it. Tool permissions are per child, so search or image generation can be off for one kid and on for another.
- Parents can read the chat that led to the action.
- Account locking and per-child tool controls live in the parent dashboard.
- HeyOtto does not auto-contact emergency services.
For families
Try freeFAQs
Is Otto an AI agent?
Otto is the assistant. It can help inside the account, including search with citations and projects a child builds in chat. It does not independently spend money or contact people outside the product. Public sharing needs a parent tap. If a product can act in the world, ask for the list of actions before you turn it on.
What actions should stay off for kids?
Payments, posting in public, emailing strangers, and any connection to a parent's bank or social account. Search and homework can stay on if you can read the transcript. Turn one tool off without deleting the whole account when you are not ready.
How is an agent different from a chatbot?
A chatbot returns a message. An agent is allowed to take a next step outside that message. The risk shows up when the step is hard to undo. Read the permission screen. A friendly name does not tell you the permissions.
Can my child approve those actions?
On HeyOtto, public sharing is a parent approval, and tool permissions are parent settings. The child does not hold the parent dashboard. Say that when you create the account so the rule is not a surprise.
Sources
Last reviewed September 26, 2026. This entry is reviewed twice a year.