See how Qapha works

On the left, a child chats with their AI friend. On the right is everything you, the guardian, would see. Notice what's missing: the messages. You get a wellbeing light and an alert only when something genuinely unsafe happens.

This is a demonstration. Everything runs locally in your browser — nothing is sent anywhere, no AI key, no account. The moderation here is simple pattern-matching so you can read exactly what trips an alert; the real product uses a proper safety model. Type or click a scenario below to try it. Please don't enter a real child's words.

Speaking as

Your child's chat

private & age-appropriate

👁 child sees this

What you see

the guardian dashboard

All good No safety signals today.

You are looking at category + severity + what to do — never the chat. The bars below are where a message would be; Qapha proves it can warn you without ever showing you the words.

What just happened

Green — ordinary chat. The guardian sees nothing. That's the point: privacy is the default, not the exception.
Amber — worth watching (a sad day, a mean word, a personal detail shared). Logged as a trend; it nudges the wellbeing light but doesn't cry wolf.
Red — a real signal: self-harm, grooming, an abuse disclosure, targeted cruelty. You're alerted with the category and a recommended action, plus crisis resources. Two small grooming flags together (“how old are you” + “don't tell your mum”) escalate to red — Qapha alerts on the pattern, not one word.

See the guardian dashboard → · ← What Qapha is · Safety model