AI tutor app development
A tutor that has read the course, and only the course.
An AI tutor app answers a student's questions from their own course material, then turns that material into flashcards and quizzes. Acasa Labs builds them on retrieval, so answers come from the lectures and notes, not the open web. NoteTube, our own learning platform, does this today for PDFs, video lectures, audio, and notes.
The problem
A general chatbot does not know what is on the exam.
Ask a general chatbot about a lecture and it answers from the internet: close enough to sound right, different enough to cost marks. It cannot see the professor's slides, the textbook chapter, or Tuesday's recording. A tutor worth using has read exactly what the student is studying, keeps each student's material separate from everyone else's, and checks understanding with practice instead of only answering questions.
What we build
- 01
Chat grounded in the student's sources
Answers retrieved from the PDFs, lectures, recordings, and notes the student uploaded, streamed back as they are written.
- 02
Practice from the same index
Quizzes and SM-2 spaced-repetition flashcards built from the material, so revision follows the course rather than a generic question bank.
- 03
Ingest for every format
Video, PDF, DOCX, images, articles, and audio parsed and chunked into one searchable index per student.
- 04
Your tutor's rules
Subject scope, tone, language, and what the tutor should decline to answer, set per course or per institution in a custom build.
From production
How NoteTube's tutor works.
Per-student, per-session scope
NoteTubeChunks and embeddings are stored in Qdrant scoped to the user and the session, so retrieval only searches the material that conversation is about.
Streaming that cleans up after itself
NoteTubeChat streams over server-sent events with inactivity timeouts, so a student who closes the tab does not leave a dead connection holding resources.
Pick the model per conversation
NoteTubeStudents choose across OpenAI, Gemini, Claude, Llama, and Mistral for each conversation. Retrieval stays the same; only the model that writes the answer changes.
Notes reachable from other AI tools
NoteTubeA remote MCP server on Cloudflare Workers, with OAuth and audience-checked tokens, lets ChatGPT, Claude, and Cursor query a student's notes once the student signs in.
Honest trade-offs
When an AI tutor is the wrong tool
An AI tutor is good at explaining, quizzing, and revising material the student already has. It does not replace a teacher marking long-form work, exam rules that need a human decision, or supervision for young children who should not chat with a model alone. If those are your users, we would build the tutor as an assistant to the teacher, with the teacher in charge of what it sees and says.
What we reach for
- FastAPI
- Qdrant
- Server-sent events
- OpenAI, Gemini, Claude, Llama, Mistral
- Next.js
- Expo, React Native
- Cloudflare Workers
Proof
Shipped, not pitched.
01Our product
NoteTube
A lecture becomes a study system: chat, flashcards, quizzes, all grounded in the source.
An AI learning platform that turns PDFs, YouTube lectures, recordings, and notes into a structured notebook with grounded chat, spaced-repetition flashcards, and graded quizzes.
Web · iOS · Android · Chrome extension · MCP server
Questions
AI tutor apps questions we get asked.
How do you stop the tutor from making things up?
We retrieve from the student's own sources first and give the model only those passages to answer from. No method removes errors completely, so we also test the tutor on real questions from your course before launch.
Can it handle video lectures?
Yes. NoteTube takes YouTube lectures and recordings, and its Chrome extension transcribes Google Meet tab audio, so a live class can become study material.
Does each student get their own tutor?
Each student's material is indexed separately, scoped by user and by session, so answers come from what that student uploaded.
Can the tutor quiz students, not just answer?
Yes. NoteTube builds graded quizzes and SM-2 spaced-repetition flashcards from the same sources the chat answers from.
Can we use our own syllabus and question bank?
In a custom build, yes. We ingest your syllabus, notes, and question bank into the index and set the tutor's scope to match.
Where do the AI keys and student data live?
In NoteTube, model keys live only in a separate FastAPI service that the apps never see. In a custom build, we can deploy to your own cloud account if the data needs to stay there.
AI tutor app development
See the tutor on your own course.
Send us a lecture or a chapter. We will show you what NoteTube does with it, and scope the tutor you would want to own.