Students get a capable study partner: explanations, practice questions, feedback, research help. You get the controls: a spend cap per classroom, an isolated key you can rotate on demand, and guardrails you configure for your deployment. Frontier-grade help, locked to your rules.
Banning it just pushes students toward unmonitored consumer tools with no oversight and no guardrails.
No spend control, no per-student isolation, no clean way to scope what it will and will not do.
A single shared account with an open tab is a budget incident waiting to happen.
A tool that goes off-script in front of a minor is a headline, not a feature.
Build the student-facing tools yourself. Qai gives you the model and the controls to keep them in bounds.
A subject tutor that explains concepts, walks through problems step by step, and adapts to the student, scoped by your system prompt to teach rather than to hand over answers.
qai-pro for most subjects; qai-think for math and multi-step reasoning.const teachingPrompt = 'You are a patient tutor. Guide the student with hints and questions. Never give the final answer outright.'; const stream = await qai.chat.completions.create({ model: 'qai-pro', messages: [ { role: 'system', content: teachingPrompt }, { role: 'user', content: studentQuestion }, ], stream: true, });
Generate practice questions at the right difficulty and give students immediate, constructive feedback on their answers, freeing instructors from rote marking.
qai-pro.Let students ask questions of a curated, instructor-approved reading set. Answers stay grounded in the provided material, with sources cited.
qai-pro; qai-think for synthesis across sources.Reading-level adjustment, translation, and plain-language rewrites so the same material meets every student where they are.
qai-pro; qai-flash for quick rewrites at volume.Qai is an API, not a black box. These are the real primitives you build a safe deployment on, plus the guardrails you define in your own prompt and pipeline.
Issue a key per classroom or cohort with its own daily budget. A runaway script or a curious student cannot blow the term budget - the key stops at its cap and nothing else is touched.
Each deployment gets its own key. Rotate or revoke it per environment without touching anything else - no shared logins floating around the staff room.
You own the system prompt. Scope the assistant to teach rather than solve, keep it on-subject, and set the tone - then pair it with your own moderation pass on input and output.
Every call is metered per key in the dashboard, so you know exactly what is being used, by which class, and what it costs. No surprise invoice.
The free sandbox tier - pilot the whole flow with a class before you spend a cent.
The everyday tutor and grader across most subjects.
Math, logic, and synthesis where the model needs to reason before it answers.
Quick rewrites, translations, and reading-level adjustments at volume.
Start on the free tier, scope it to your classroom, and add a budget when you roll it out. The controls come standard.