Parent Guide — Kids OpenCode

A practical guide for parents whose kid is starting with Kids OpenCode, our AI coding mentor for kids 12 and older. Ten minutes of reading; written for parents who don't code.

Last updated: 2026-06-12 · Version: 0.1

1. What your kid is actually getting

Kids OpenCode is a program your family installs on your own computer. Your kid opens a terminal, types kids-opencode, and a patient AI coding mentor wakes up. It helps them build real things — a website about themselves, a small game — in guided two-hour Course Packs.

Three things make it different from handing your kid a chatbot:

1. It asks permission before every action. Before it writes a file, reads a file, or looks something up, it asks. Your kid approves each step. Nothing happens silently.
2. It teaches; it doesn't do the work. The mentor uses guided questions. Only after your kid has genuinely tried three times will it offer a small piece of code — with an explanation of every line.
3. It's locked down. It cannot run system commands. It cannot touch files outside the project folder. It can only look at four documentation websites (MDN, web.dev, the W3C specs, and our own docs). It never pretends to be human.

What it is not: a social network (no friends, no feed, no messaging), a homework machine, or an open-ended chatbot. If the conversation drifts off coding, the mentor steers it back.

2. Before your kid starts — setup checklist

• You create the Family Account (adults only) and your kid's profile. Use a nickname, not their real name — we designed the whole system so we never need your kid's real identity.
• Read and accept the Parental Consent — it's itemised, so you can see exactly what each consent covers.
• Install Kids OpenCode together (instructions in the open-source repo at github.com/kidsinai/kids-opencode).
• Sit with them for the first session. Not because the tool is risky — because watching the plan→approve→build loop once is the fastest way to understand what your kid is doing in there.

3. How a session works (and what it costs)

Each request to the AI costs Stars — prepaid credits you buy and control (packs from A$10; Stars never expire, and there is no auto-recharge). A typical request costs 1–30 Stars and the cost is shown before your kid confirms. When the Stars are gone, the session pauses until you decide to top up.

The mentor plans out loud: "Here's what I suggest we do, in three steps. OK to start?" Your kid approves or asks questions. Every tool action lands in the audit log you can read later.

4. Five rules worth teaching your kid

1. The AI is a tool, not a person. It will tell your kid this itself if asked — it never pretends to be human. Good to reinforce at home.
2. Never tell any AI your real name, school, address, or birthday. Kids OpenCode doesn't ask for these and doesn't need them. (We encourage keeping them out of project files too.)
3. AI answers can be wrong. Sometimes obviously, sometimes subtly. "Check it before you trust it" is the single most valuable AI habit a kid can learn.
4. Trying first is the point. The mentor holds back full answers on purpose. That's not the tool being unhelpful — that's the lesson.
5. If anything feels off, stop and tell a parent. The session can always wait.

5. Your dashboard: what you can see and control

From the Family Dashboard you can:

• Read the audit log — a plain-English record of what the AI did in each session ("wrote a file called index.html"), kept fully accessible for 90 days. The contents of your kid's files stay on your own computer; we log actions, not their work.
• Manage Stars — see spending per kid profile, reallocate between kids, top up (or not).
• Export everything — one click downloads all your family's data as a file.
• Delete everything — close the account; data is deleted within 30 days. Both controls are in Settings, no email required (though privacy@airbotix.ai works too).

6. Safety: what we built, and where you come in

There are three technical safety layers (the kid-safe system prompt — published publicly, the tool restrictions, and server-side content moderation on every AI call). They're described in plain English on our Compliance page.

One specific behaviour you should know: if your kid says something that sounds like self-harm or being in danger, the mentor is designed to stop the coding conversation and point them to Kids Helpline (1800 55 1800) — and the event appears in your audit log so you can follow up. Like any AI safety layer, detection is not guaranteed. You are the primary safeguard; the tool is built to support that, not replace it.

7. If something goes wrong

8. The paperwork, if you want it

• Privacy Policy — what we collect (very little) and what we never do (sell data, train AI on your kid's work, show ads)
• Terms of Service — the contract, including Stars and refunds
• Parental Consent — the itemised consent list
• Compliance Statement — the laws we operate under and how, in plain English

Questions this guide didn't answer: support@airbotix.ai. We read everything.