AI activity is the last section in a class's sidebar. It answers three questions in three tabs: how much of the model your students used (Usage), what they said to it (Conversations), and what the safety checks caught (Flags). Each tab loads when you open it.
See how much students used the model#
The Usage tab opens on a window of 7 days, 30 days, 90 days, or All time, chosen with the buttons at the top right.
Four tiles sit above the rest: Credits spent, Runs, Active students, and, in a school class, Credits granted. Below them, a column chart shows credits per day over the window you picked, with a table view for the same numbers. The per-student table sorts on every column and carries Student, Credits spent, Granted, Balance, Runs, Model calls, Tokens, and Last run. Pressing a name opens that student's drawer.
The two main numbers count different things, which the page says underneath its heading. Credits count every AI request a student made with this class's credits, including work in the tutor. Runs count only projects that belong to this class. A student who leaves the class drops out of the totals.
Understand and grant credits#
A credit is what an AI request costs. Your students never buy or hold credits of their own, and in a school class your plan carries a balance that the class draws from.
When a student runs out, the roster in the Students section shows an Out badge beside their Credits figure, and the student's next AI request fails with "You have run out of AI credits. Ask your teacher or add more to keep going." Top them up from the roster by opening the menu at the end of their row, or by ticking several rows and using the toolbar, then choosing Grant credits. The dialog asks for Credits (or Credits per student when you picked more than one), says how many that is in total and that it comes out of your balance, and Grant sends it. Granting to a selection happens one student at a time, so if one fails the dialog names who was and was not topped up. This is a school-class feature, so a personal class shows neither the balance column nor the action.
Read what students said to the AI#
The Conversations tab is a read-only record of a student's AI chats on this class's work. Choose a student, then a conversation, and the transcript appears with each turn labeled Student or AI. You cannot reply from here, and a long conversation comes back trimmed to its most recent turns.
Every time you open a transcript, Trace writes an audit entry recording that you did. Students are told in the privacy policy that teachers and guardians can see a minor's AI interactions for supervision and that the access is audited.
Act on a safety flag#
The Flags tab lists what the safety checks caught in this class, newest first. A row names the student, where it happened (AI Tutor or Agent Builder), whether it was a Student message or an AI reply, the time, and a category badge. The categories are Sexual content, Graphic violence, Self-harm, Hate or harassment, Dangerous or illegal, Cheating, and Other.
A flag the check treated as a crisis also carries a Check on this student badge. Treat that as a reason to speak to the student in person and to follow your school's own safeguarding process.
The flagged text itself is not stored on the flag, so the row cannot quote what was said. It links into that student's conversations instead, which is where you read the context.
Answer a parent about what students can do with AI#
Two AI surfaces exist for students. The AI Tutor answers questions about the lesson they are reading, and the Agent Builder lets them wire up and run an agent against approved models. Both are covered by the same school policy on every request, which asks for answers appropriate to a K-12 classroom, refuses harmful and dangerous material, asks the model to respond supportively to distress, and blocks direct help with cheating on active tests and homework while still allowing explanations, hints, practice, and feedback.
Two separate checks run on each tutor turn. One reviews the student's message before it is stored or sent onward, and a blocked message is neither stored nor sent. The other reviews the proposed reply before the student sees it. If a check cannot return a clear verdict after one retry, the message or reply is withheld and the safety check reports itself as unavailable.
Explain the privacy position#
Everything below comes from the published privacy policy at /legal/privacy,
which is the document to send a parent who asks.
Trace does not use your data to train, fine-tune, or improve any AI model, and that covers student work, prompts, model outputs, quiz answers, and project content. Every model request carries instructions forbidding the provider from retaining the content or training on it, across the Agent Builder, chat, speech, the SDK gateway, and the safety checks themselves. Prompts go through the Trace backend rather than straight from the browser, which lets Trace strip the signed-in student's own name, email, account ID, and date of birth, along with anything shaped like an email address, phone number, street address, or identification number. The policy is plain about how far that reaches. It removes the identifiers of the student who is signed in and does not recognize a classmate's name, a teacher's name, or a school's name, which is why students are told not to type personal information about themselves or anyone else into an AI system.
Trace holds no integration with any student information system, so it never receives records from your school's own database. It does not make automated decisions about a person, and it does not compute a mastery score, a grade, or a judgment about a student's ability. A teacher sees the work their own students do in their own class, a school administrator sees the classes and teachers in their school, and nobody outside the class sees a student's work unless the student deliberately shares a project.
Voice input in the Agent Builder is still being settled for pilot classes. The pilot agreement says the speech-to-text block stays off for pilot student accounts unless the school agrees to it specifically, so ask us before a class relies on it.
What this page does not do#
- It shows this class only. There is no school-wide or cross-class AI report here.
- A flag does not carry the text that triggered it, so the row cannot show you what was said.
- You cannot reply to a student inside a transcript, and you cannot delete or edit one.
- Credits measure what AI requests cost. Nothing on this page is a score, and Trace does not assign a grade.
- Granting credits does not change a student's work, progress, or due dates.