# AI Tutor on Codex: Adaptive Learning, Session by Session > AI tutor on Codex that teaches any topic adaptively, then writes an interactive HTML lesson you keep forever. Worked example, quiz, one file. BYOK. Fork on… - Published: 2026-07-29 · Updated: 2026-07-29 · jurniti - Canonical: https://www.jurniti.com/blog/teacher-adaptive-learn-tutor-codex-template If explanations from your AI keep vanishing into scrollback, this Codex agent teaches any topic adaptively and then writes a self-contained interactive HTML lesson you can reopen tomorrow, fork, and share. **AI Tutor: Adaptive Learning Agent (Codex)** is a public jurniti template (`/templates/teacher-adaptive-learn-tutor-codex`). You fork it into a paid **codex** microVM — not a Notion checklist and not a free shared sandbox. If you searched for **AI tutor agent**, this page maps that intent to a concrete fork: product truth, gates, BYOK, and the demo you should run before real customers see anything. ## What searchers mean by AI tutor agent People typing **AI tutor agent** usually want one of three things: a hire-me agency, a course, or a runnable operator. Hire-me pages sell retainers. Courses sell modules. jurniti sells a **running agent seat** you can fork, key, and operate — with Firecracker isolation and a human still on the hook for dangerous sends. This post is written for the third group: operators who will pay for a real box and own the demo. ## Why AI Tutor: Adaptive Learning Agent (Codex) runs on a Firecracker microVM Laptop agents die when the lid closes. Docker density is not the same as multi-tenant isolation. jurniti provisions a Firecracker microVM per fork. The **codex** harness install lives on the platform golden path; your template persist (skills, instructions, config shapes) restores on top. Boot is fast; full provision is still about **three minutes**, not a marketing 124ms hand-wave for the whole journey. More on the harness: [codex guide](/blog/codex-cli-hosting). ## What you get when you fork AI Tutor: Adaptive Learning Agent (Codex) When you fork **AI Tutor: Adaptive Learning Agent (Codex)**, you are not buying a PDF. - The adaptive /learn tutor flow - The HTML lesson generator: a keepable file, not chat vapor - The worked-example-first teaching pattern You still configure secrets after fork. Nothing valuable should be baked as a live credential in a public capture. ## The undeniable demo for teacher-adaptive-learn-tutor-codex After keys are set, run the demo the gallery About body describes for **AI Tutor: Adaptive Learning Agent (Codex)**. Do it against a dry-run or your own inbox first. If the demo cannot complete, do not point the agent at real customers yet. ## BYOK, human gates, and what you still own **BYOK:** model keys and third-party API keys are yours. jurniti does not mark up tokens in v1. Your own OpenAI / Codex credentials, and nothing else. BYOK: keys live only inside your VM, and no third-party LMS account is required. Where outbound mail, social publish, or paid fulfillment exists, treat **human approval** as product. Auto-spam is the wrong fit for this gallery. No free trial. No free tier. First purchase: **30-day money-back guarantee**. Model and SaaS fees stay on your accounts. ## AI Tutor: Adaptive Learning Agent (Codex) vs building the same stack yourself | Path | What you get | Cost of time | | --- | --- | --- | | DIY scripts + laptop | Fragile glue | Nights and weekends | | Generic course | Concepts, no seat | Weeks of setup | | **Fork AI Tutor: Adaptive Learning Agent (Codex)** | Pre-wired agent on microVM | Minutes to first demo | Courses do not remove wiring. Shared multi-tenant toys do not give you isolation. The template is the middle path: paid, isolated, opinionated. ## Fit and not-fit for teacher operators Not for you if you need an LMS with gradebooks and SSO, a pure flashcard app, or a harness other than Codex. - Learners who want durable lesson files instead of chat scrollback - Codex users studying technical topics - Teachers prototyping interactive explainers ## How to fork AI Tutor: Adaptive Learning Agent (Codex) on jurniti 1. Open [AI Tutor: Adaptive Learning Agent (Codex)](/templates/teacher-adaptive-learn-tutor-codex). 2. Fork on a paid plan. 3. Wait for the microVM (about three minutes). 4. Export BYOK keys listed in the About body. 5. Run the undeniable demo before real traffic. That is how **AI tutor agent** becomes an operated system instead of another bookmark. ## Frequently asked questions ### What is AI Tutor: Adaptive Learning Agent (Codex)? AI Tutor: Adaptive Learning Agent (Codex) is a public jurniti template. You fork it into a paid codex microVM with the agent setup already on disk. ### Who should search for AI tutor agent? Operators who want a runnable agent seat for this job — not only a course or a hire-me agency landing page. ### Is there a free trial? No free trial and no free tier. First purchase includes a 30-day money-back guarantee. Model and third-party fees are yours. ### Whose API keys? Yours (BYOK). Keys stay inside your microVM. jurniti does not proxy or mark up model tokens in v1. ### Which harness does AI Tutor: Adaptive Learning Agent (Codex) use? codex. Choose another gallery template if you need a different agent runtime. ### How is this different from Zapier or a course? Zapier wires apps. Courses teach ideas. This template forks an always-on agent microVM with the workflow already installed. ### How long does provision take? About three minutes to a running microVM. Fast boot metrics are not the whole provision story.