Educators Who Build

Tutorial

Getting Started With PrairieLearn

Your own AI agent takes you from nothing installed to PrairieLearn running on your computer, with a course of your own and a question that generates a new version of itself for every student, designed with you around something you actually teach. An hour or two, and you can stop and pick up again with the same link.

Be clear about where this ends. PrairieLearn on your computer is for building and trying things: only you can reach it, and students can't. Putting questions in front of students takes a course on an official PrairieLearn, either your institution's or PrairieLearn's hosted service, which is free for instructors and for courses of up to twenty students a term, and charges per student above that. A last, optional step moves your work there when you have one. You don't need one to start.

It needs an agent that runs on your computer, such as the Claude desktop app (with Claude Code), Codex, or Cursor, and permission to install software, since PrairieLearn runs in Docker. Your agent explains all of this before anything starts, and nothing happens until you say yes.

Why Do This

PrairieLearn is an assessment platform that does far more than the quizzes in Canvas or Moodle. A question is a small program, so every student gets their own version, answers are computed rather than typed in by hand, and students can enter numbers with units, formulas, code, and more, with feedback on the mistakes they actually make. Students can practice a question until they've mastered it, and exams draw fresh versions for everyone.

It's also much easier to build with AI. A PrairieLearn question is a few plain files, so your agent can write one with you, run it, and check hundreds of its versions before a student sees any, which a quiz you click together in an LMS can't offer.

What We Collect

As you go, your agent tells this site which steps you reached, where it stalled and why, whether you're heading for an official PrairieLearn and its address if so, your broad field (such as chemistry), and what it is: the app it runs in, its model, and your operating system. That is tied to your email, so we can see where this tutorial works, where it gets stuck, and follow up if you're stuck. If you join the room at one of our sessions, its organizers can see your progress too, so someone can come help. Your agent explains all of this and asks before it starts. If you tell it you don't want reports, it sends one last message saying only that, which we keep with your email, and nothing after it.

Never your questions, your course's name, or anything from your conversation. Anything your agent writes in words, it shows you first. If you're stuck, it can send a question to the people who run this, and only when you say so.

Curious what your agent reads? The instructions, as plain text.