Your first quiz, with AI.

Do I need coding experience? Can I use a chat assistant?

No coding experience is required. An agent with file and command access can perform setup. A text-only chat assistant can explain steps and draft questions, but needs execution tools to set up your course.

What should my question-design specification say?

Have AI interview you about recall, concept application, theoretical reasoning, transfer across domains, or translating word problems into concrete models. Specify audience, difficulty, answer time and evidence of understanding. Exclude unsuitable tasks, such as untaught topics, lengthy calculations, trick questions or required writing.

How do I generate a large, course-aligned bank?

Give AI your materials and design specification. Request roughly 100 varied, source-grounded questions with reference answers and grading criteria. Have the agent validate them. Review samples across topics and difficulty, try several aloud, then approve and freeze the bank.

Bank-generation specification →

Do I need a website, paid hosting, or an AI grading account?

A free Cloudflare account provides hosting; no existing website or custom domain is needed. You need Gradescope Programming Assignment access. Default grading needs no external AI key. Your setup agent may require a paid plan.

What if setup breaks, or I want a change?

Describe what you expected, what happened and the error shown. Ask the agent to diagnose, fix and retest without bypassing safeguards. Describe changes in ordinary language; you handle sign-ins and approve consequential changes.

Where does student work go?

Audio goes only to Gradescope. The quiz server receives the key, recording fingerprints and timings, not names, emails or audio. Identity checks and grade assignment run in Gradescope. Assessment uses a local AI model or an institution-approved external model receiving de-identified, redacted text. Keep course banks and credentials private.

How do approved extra-time accommodations work?

Students select approved extra time. Check claims manually, or download Gradescope submissions and run the working local audit script yourself against a local list of approvals. The list stays off Gradescope and out of the autograder. The script flags unsupported claims only when excess time was used; review flags before acting. Grades are not changed automatically. Keep audit files out of AI tools. After review, delete local approval lists, downloads and audit outputs, following institutional retention rules.

What should I check before students use a real quiz?

Test the deployed quiz through to a Gradescope grade with its matching autograder. Start with ungraded practice, share keys privately and provide the Student guide. Explain deadlines and which submission counts. Review flagged work and sample ordinary results.