This website
The only information this site handles is what you type into the learn-more or contact form, delivered by email to a private inbox. Analytics are aggregate and cookie-free.
Data handling is a design decision.
Atlas is an open exploration rather than a hosted service. These are the positions the design takes on course and student data, and what you would own if you ran it yourself.
The application
Atlas the application is open source. If you deploy it, you choose the hosting, identity, AI providers, and retention rules, and you own the approvals your institution requires.
Positions built into the design.
Course and student data deserve the same care as grades. The design assumes that from the start.
Sample data by default
The demo course, people, and assignments are fictional. Nothing in the project depends on real student records.
Teachers see the evidence
Claims, revisions, sources, and the agent's part in them stay visible to the teacher. Nothing is scored in the dark.
Agents act within a role
A course agent sees what its role allows and records what it used and what it changed.
Evidence, not surveillance
The record exists so the teacher can assess the work, not to monitor the student.
What a real deployment would still need.
Running Atlas with real course data means answering the questions any institution asks before approving software.
Privacy and data
Decide what information is needed, how it is used, where it is stored, and when it is deleted.
Access and security
Define who can access the work and the identity, hosting, and security controls required.
AI providers
Review the providers and their data processing terms, including retention and model training.
Institutional approval
Expect a security questionnaire, agreements, and a compliance review before institutional use.
Thinking through the same problems?
I am happy to compare notes on how a course-grounded agent should handle the data it sees. If you find a problem with this site, report a security concern.
Want to learn more?