Frequently Asked Questions

A magic link is a secure, one-time login link that Kaze sends to your email address. On the Parent sign in page you type your email, choose Email me a sign-in link instead, and click the link we send you. Each magic link is unique to you and expires after 30 minutes. Magic links remain available even if you also set a password.

Check your spam or junk folder first. The email comes from noreply@mail.kazeapp.com. If you still do not see it, return to the login page and choose Email me a sign-in link instead again to request a new link. Each new request invalidates any previous link.

Yes. You can create an account with an email and password, add a password later from Dashboard > Settings > Security, or sign in with Google. Passwords must be at least 8 characters and mix letters, numbers, and a symbol. Magic links keep working too, and if you forget your password you can reset it with the Forgot password? link.

No. Children do not have separate accounts, passwords or email addresses. Your child accesses Kaze through a device session that you link from your dashboard. You control when and how long your child can use Kaze by managing sessions and setting time limits.

On your child's device, open the Kaze website and tap I'm a kid. An 8-character code appears on screen. In your dashboard, go to Sessions > Link Device, enter the code, and select your child's profile. The device connects automatically. If you'd rather hand your child your own browser, use Quick Start from the dashboard. Without a parent PIN it signs you out of that browser and you sign in again; with one, your account stays behind the PIN and opening the dashboard there asks for it.

Kaze is in early access: you can create an account now. There will be different plans in the future, but there will always be a basic free tier with limited time and features. Regardless of the plan, Kaze will never show ads or share your data with third parties.

No. The AI is included in Kaze. You don't need an account with OpenAI or any other AI company, and there are no usage bills to watch. Kaze chooses and vets the model your child talks to.

Yes. You can link multiple devices to your child's profile. Each device session is independent, and you can see and disconnect any of them from Dashboard > Sessions at any time.

To protect your children's information. You can sign up, set a password, manage your account, and redeem invites right away, but before you add a child, view any family's information, or enter any personal details, we confirm that the email address really belongs to you. This makes sure a mistyped or wrong email can never receive or be linked to your kids' information, including for someone invited as a co-custodian. To confirm, click the link we email you; the dashboard shows a reminder with a Confirm email button until you do.

Nothing about your family. "Sign in with Google" only uses Google to confirm your own email address and name so you can log in, the same way any website's Google sign-in works. Google never receives your children's profiles, their chats, activity summaries, safety logs, or any other information you keep in Kaze. Signing in with Google also does not give Kaze access to your Gmail, Drive, contacts, or anything else in your Google account.

As many as you have. Each child gets their own profile, with its own learning settings, accessibility options, chats, projects and badges. Brothers and sisters never see each other's chats.

A learning profile describes how your child learns best: their age, reading level, learning style, interests and any needs such as dyslexia, ADHD or autism. Kaze reads it before every chat and uses it to shape how the AI talks: sentence length, vocabulary, whether to use a diagram, and which examples to reach for.

Yes. You can update your child's learning profile at any time from Dashboard > Children. Select the child, click Edit, and adjust any field. Changes take effect in the next chat your child starts.

A first name or nickname, an age, and optionally a learning style, reading level, interests, accessibility needs and the language for chats. You fill in as much or as little as you want. The more detail you provide, the better Kaze adapts. No surname, no photo, no school.

Kaze is designed for children from about 6 to 16. The AI adapts its language, depth and format to each child's age and reading level: short, playful answers with simple words for younger children, and deeper, more challenging conversations for teens. Children of any age are meant to use Kaze with a grown-up nearby.

When your child starts a chat, Kaze reads their learning profile and builds instructions that tell the AI how to communicate. Depending on the profile, the AI adjusts sentence length, vocabulary, response format, diagrams and the examples it uses. A child with dyslexia gets shorter sentences and simpler formatting. A gifted child gets more depth and intellectual challenge.

No, and that's on purpose. Kaze works like a tutor: it gives a first step or a hint, asks a question back, and explains another way when your child says they don't get it. It won't write the essay or fill in the worksheet. You can see how it went in the chat's summary, and open the full messages whenever you want.

Anything a curious child would ask a tutor: school subjects, how things work, stories, ideas, what they're reading. Kaze kindly declines requests outside that, such as adult tasks or long pieces of writing to hand in as their own, and steers back to learning. If your child is sad or worried, Kaze listens first and encourages them to talk to a grown-up.

Yes. Each child has a daily message limit and each session has a time limit. When your child reaches one, they see a friendly note that they can come back later, and you can see it in your dashboard. These limits are deliberate: enough for Kaze to be useful, without becoming something to sink an evening into.

English and Spanish. You set each child's default chat language, and which languages they may choose, in Dashboard > Children > Edit. Your child can also pick a language when starting a new chat. Dictation and read-aloud follow the chat's language.

Private by default, though never from you. You see a summary of every chat. If you choose to reveal the full messages, your child sees a note that a grown-up reviewed the conversation. The design protects your child's sense that reading their chat is a deliberate act, not a default, while keeping everything within your reach.

Yes. Every chat has a microphone. Your child taps it, talks for up to 30 seconds, and the words appear in the message box to check before sending. The recording is turned into text on Kaze's own server and deleted right away; it is never sent to another company and never kept. Dictation is on for every child, and you can turn it off in the child's profile.

Yes. A speaker button next to each reply reads it out loud, highlighting each sentence as it's spoken, so a child can follow along. It uses the voices built into the device, so nothing is sent anywhere. You turn it on in the child's profile.

A project groups chats about one topic, such as a school assignment or a book, so your child can come back to "my volcanoes project" and find everything in one place. The logbook is a short learning journal: after a chat, your child writes a few lines about what they learned, starting from a recap written in their own voice, which they can keep or change. Badges are small personal rewards for keeping at it, visible only to your child; there are no leaderboards. You can see all three from your dashboard, and print a project as a PDF.

Next to any message there is a small hand button. When your child is stuck, or finds something they want to show you, they tap it, add a short note, and send you that message and the one before it. You answer from your dashboard, and your reply appears inside their chat. Only those two messages are shared, even from a private chat, and the AI never sees your answer.

Every message your child sends, and every reply the AI writes, goes through several checks before it's shown. Personal details such as phone numbers, addresses and full names are caught before anything reaches the AI. Sensitive topics and attempts to trick the AI are detected, and a content moderation service scores every message and reply. Anything doubtful is held for you to decide on, with an entry in your safety log and an email to every grown-up who looks after your child.

Flagged content is held before your child sees it or the tutor gets it. Your child sees a simple, friendly note that a grown-up will check it, and can keep chatting. The event is recorded in the safety log with its severity level. From Dashboard > Safety Logs you can say whether the flag was right, and allow the content, which then reaches your child's chat, or reject it.

Chat pauses. Your child sees "Kaze is taking a short break to keep you safe" and cannot start a conversation or send a message until the safety checks are back, usually within a minute. Kaze does not carry on with weaker checks. A message sent while a check is delayed is kept encrypted until it can be checked, for at most a few hours, and is deleted if it can't be.

Not yet. Kaze currently applies the same safety rules to every child. A future update will let you configure thresholds per child. In the meantime, every flag is recorded in your safety log with a severity, so you can see patterns, and you decide on anything that was held.

Safety events are classified as low, medium, high, or critical. Low events include messages that are too long. Medium events include attempts to share personal information. High and critical events involve harmful content detected by the moderation service. You can filter safety logs by severity in Dashboard > Safety Logs.

Three companies see your child's messages, and no other: Anthropic's Claude model writes the replies, and OpenAI's moderation and TypeSafe's Jev check the messages and the replies. The model that writes the replies receives the message with personal details removed, the conversation so far, and instructions built from the learning profile (an age, a reading level, needs and interests, never a name). The safety checks, OpenAI's moderation and TypeSafe's Jev, read the message as written, because they must judge exactly what was typed; TypeSafe returns scores and never writes anything. Nothing from Kaze is used to train anyone's models. Anthropic deletes what it receives within 30 days, except what its safety systems flag, which it can keep for up to 2 years.

Each child at a glance: recent chats with their summaries, anything waiting for your decision, help requests, projects, logbook entries and badges. Open any conversation to read its summary, reveal the full messages, move it into a project, or print it.

Yes. A weekly recap is on from day one: number of chats, topics, any safety events, and a short narrative in plain language. You can switch it to daily or monthly, choose the hour, and have it by email instead of only on the dashboard. You can also generate one on demand.

Go to Dashboard > Sessions and click Disconnect next to the device. Your child's session ends immediately, and they need a new code from you to continue. Session timers and the daily message limit also end sessions on their own.

Yes, in two ways. Print a conversation or a whole project as a clean PDF from your dashboard; the PDF is made in your browser and nothing is stored on Kaze's servers. Or invite the teacher to a group with the reviewer or teacher role, so they see the child's activity directly, with the access their role allows.

Yes, both, from Settings > Data & Privacy. Download everything as a JSON file at any time. Delete a child's profile and all its data, or your entire account, and it is permanently removed. There is no need to contact us, though you can.

It never shows ads to children, never sells or shares your data with data brokers, never runs third-party analytics on children's pages, never lets a child share personal details in a chat, and never sends notifications to pull a child back in.

A group is a shared space where more than one adult supports a child's learning: two parents in different homes, a family and a tutor, a teacher and a class. When you enroll a child in a group, every member can see and interact with that child according to their role.

Kaze offers seven roles, from most to least privileged: Owner (full control), Admin (manage members), Custodian (a parent's level of trust: can review and enroll the child), Tutor (educational role, can enroll), Teacher (can view enrolled children), Reviewer (can review conversations and safety logs), and Member (basic view-only access).

Owners and admins invite new members by email from the group page. Enter the person's email address and choose a role. Kaze sends an invitation email with a link. The recipient clicks the link to accept and join the group; if they don't have a Kaze account yet, they create one on the way. Invitations expire after 7 days.

Yes. You can be a member of as many groups as you need. You might be the owner of your family group, a teacher in a classroom group, and a reviewer in a tutoring group. Each group has its own role for you, and a group can have sub-groups, such as a tutoring group inside a family group.