Four for the institution, four for teachers, three for learners, three for parents, and one that answers questions about the platform itself. Each page below states exactly what is sent to the provider — and requests asking for work about a named child are refused before any network call is made.
Across four consoles.
Enforced in code, not policy.
Entirely, with the reason stated.
Every learner question, and they are told.
Which sources actually convert, and where the pipeline stalls.
Open ↗ §Where the recoverable money actually is.
Open ↗ ◫What the generator could not solve, explained.
Open ↗ ▤Circulars, policies and letters to parents, drafted for you to edit.
Open ↗Objective, sequence, checks for understanding, likely misconceptions.
Open ↗ ◉Photograph a textbook page. Get a planned lesson.
Open ↗ ▥Practice questions from straightforward to demanding, with answers.
Open ↗ ◇A second explanation for the learner who did not follow the first.
Open ↗Ask when you are stuck. Your teachers can see what you asked.
Open ↗ ▤A similar question to try, once you have understood the method.
Open ↗ ✓Explain your reasoning and find out where it goes wrong.
Open ↗What the school sent you this week, in plain language.
Open ↗ ◐A school message in the language you read most comfortably.
Open ↗ ?Plain answers about grades, attendance thresholds and school terms.
Open ↗No learner name, no learner work, no marks and no attendance data are ever sent to an AI provider. What goes out is a teacher's own prompt, a photographed textbook page, or a learner's own question. Nothing else.
Requests that ask the model to mark named work, grade submitted answers or write a report card comment about a child are refused before any network call happens. Policies get forgotten on a busy Friday. A refusal in the code does not.