CongoSky · Qapha Building · for families

Child-safe AI

A kind AI to talk to. A guardian who only hears when it matters.

Your kids are already talking to AI on their laptops and tablets. Qapha gives them an age-appropriate one that's actually safe — and gives you peace of mind without reading a single message.

Qapha · isiZulu / isiXhosa: “be watchful, be alert.” It watches the road, not the child.

Qapha is the watchful engine: the part that decides when a grown-up needs to be told. Ikhaya is the home it watches over — a child's diary-by-default companion. The home, and the watcher in it.

Try the live demo →

The problem, plainly

Children are pouring their hearts into free chatbots and open game chats that were never built for them. Two things go wrong there: the AI itself can be unsafe or inappropriate, and the humans who reach kids through these apps — the stranger posing as a friend, the predator on Roblox, sometimes someone the family knows — operate in the dark.

Parents are left with a bad choice: read everything and destroy your child's trust, or read nothing and hope. Qapha refuses that trade.

How Qapha works

  1. Your child gets a safe AI friendWarm, curious, age-tuned. It chats about games and homework, and it knows how to respond with care when a child is hurting — never cold, never clinical.
  2. Qapha reads every message — you don'tAn on-device first pass plus a safety model classify each message into safe, watch, or alert. The words stay with your child. You never see the chat.
  3. You hear only when it's realSelf-harm. Grooming. An abuse disclosure. Targeted cruelty. You get the category, what it means, and what to do next — with crisis resources — not a transcript.

See it in action →

What gets you an alert — and what doesn't

Safe — you hear nothing. Ordinary chat about games, friends, feelings. This is most of it, and it stays private. Privacy is the default.
Watch — a gentle nudge. A sad day, a mean exchange, a personal detail shared with a stranger. Tracked as a trend so you can notice a pattern, without an alarm for every bad afternoon.
Alert — you're told now. Signs of self-harm or suicidal thinking; grooming (secrecy, “let's move to Snap,” meet-ups, requests for photos); a child disclosing that someone is hurting them; someone telling your child to harm themselves. You get the category and a recommended next step.

Qapha alerts on the pattern, not a single scary word. “How old are you?” alone is nothing. “How old are you” + “don't tell your mum” + “send a pic” is a predator — and that's what crosses the line to red.

Why we don't show you the chats

A child who knows a parent reads everything stops telling the truth — to the parent and to the AI that could help them. Surveillance a child can feel is surveillance a child routes around, straight back to the unmoderated apps. Qapha keeps the child's trust so the safety net actually catches something.

  • You see category + severity + what to do — never the message text.
  • The AI never tells your child they were “reported.” It supports them; the alert reaches you quietly.
  • There is an audited break-glass path to reveal context when a life is at risk — a deliberate, logged action, not the everyday view.
  • We never sell chats, train ads on children, or build a profile of your kid.

Read the full design: Qapha safety model & privacy posture.

Technology & trust

Runs on CongoSky A sibling to Khuluma (speak) and Lalela (the kind games platform). Same sovereign, frugal, African stack — Auth0, Neon, Render, OpenRouter.
Two-layer moderation A fast on-device pass catches the obvious instantly and offline; a safety model on our API does the nuanced read. The model sees the message to classify it; you see only the verdict.
Data minimisation We keep alert categories, timestamps, and trend counts — the least we can hold and still warn you in time. Raw chat is not a guardian-visible store.
Not a replacement for you Automated moderation misses things and over-triggers. Qapha is a smoke alarm, not a fire brigade. The caring adult is still the plan; Qapha just makes sure you're in the room when it counts.

For your kids — and as a product

Qapha starts as something for our own families and grows into a product for every parent who wants the same deal: a safe place for a child to talk, and an alert they can trust. Try the working demo, then tell us what would make you hand it to your child.

Open the demo →

See the parent side too: the guardian dashboard — wellbeing, alerts, and what-to-do, never the chat. Running on CongoSky, the sovereign cloud for Africa. Back to congosky.cloud