← Blog Open Source Nov 21, 2026 · 6 min read

DeepTutor: The Open-Source AI Tutor Trending on GitHub

DeepTutor is an open-source AI tutor with quizzes, cited research, memory and spaced review. What it does, how to run it from the README, and its real limits.

DeepTutor: The Open-Source AI Tutor Trending on GitHub

DeepTutor is an open-source AI tutor from the HKUDS lab that is one of the fastest-growing projects on GitHub this month. This guide explains what DeepTutor does as a long-term personal tutor, how students and self-learners can run it using the official README, and where it falls short compared with the study modes built into paid chat apps.

On GitHub's monthly trending page on November 21, 2026, HKUDS/DeepTutor showed 61,082 stars, with 19,312 added in the month, one of the largest monthly gains on the list. The repo is licensed under Apache-2.0, and its latest release at the time was v1.7.4, dated November 20, 2026, described as a desktop security update. Always check the repo yourself before installing, because the project moves quickly.

The README describes DeepTutor as "Lifelong Personalized Tutoring" and as an agent-native learning workspace that connects tutoring, research, creation and practice in one system. There is also an arXiv preprint behind the project.

What DeepTutor can do

DeepTutor is not one chatbot. It is a workspace with many learning modes, according to the README:

Learning modes

  • Chat: the default tutor conversation.
  • Ask Questions: focused Q&A.
  • Quiz and Practice: generated quizzes, a Question Bank and review scheduling with recall ratings (a form of spaced repetition).
  • Solve: worked, step-by-step reasoning.
  • Research: cited reports on a topic.
  • Visualize and Video Lesson: visual explanations.
  • Mastery Path: a structured path through a subject.
  • Immersive Reading and Book: interactive "living books".

Knowledge bases

You can load your own material and let the tutor answer from it. DeepTutor supports several retrieval engines, including LlamaIndex (the default), GraphRAG, LightRAG, Kiwix and Obsidian.

Memory

A three-layer, file-backed memory (traces, facts and synthesis) that you can read and edit. This is what makes it a long-term tutor: it can remember what you studied and where you struggled.

Workspace tools

A Task Board, a Calendar with .ics export, a Learning notebook with Anki export, and Favorites. Optional multi-user mode with separate workspaces is off by default.

How to run DeepTutor (from the official README)

You need to connect an AI model provider. After starting, the README says to configure a provider and model under Settings → Providers and Language models. That means you either use an API key from a provider or a local model setup; the README mentions LM Studio and llama.cpp support in its release notes.

Option 1: Docker (easiest if you have Docker)

docker run --rm --name deeptutor \
  -p 127.0.0.1:3782:3782 \
  -v deeptutor-data:/app/data \
  ghcr.io/hkuds/deeptutor:latest

Then open http://127.0.0.1:3782 in your browser.

Option 2: PyPI (Python 3.11 to 3.14 and Node.js 20.9+)

mkdir -p my-deeptutor && cd my-deeptutor
python -m pip install -U "deeptutor[full]"
deeptutor init
deeptutor start

Option 3: Command line only

mkdir -p my-deeptutor && cd my-deeptutor
python -m pip install -U deeptutor
deeptutor init --cli
deeptutor chat

The README notes the core CLI package includes only Chat, Ask Questions, the CLI and the SDK; other modes need extra modules. Preview desktop installers are also listed for macOS and Windows; the README warns Windows preview builds are unsigned and may trigger a SmartScreen warning.

A study workflow that uses DeepTutor well

  1. Create a knowledge base from your own notes, lecture slides or a textbook you legally own.
  2. Start with Ask Questions on the chapter you are studying. Ask it to explain in simpler words, then with an example.
  3. Use Solve for exercises, but try the problem yourself first and only check the steps afterwards.
  4. Generate a Quiz at the end of each session.
  5. Rate your recall honestly so the review schedule adapts.
  6. Export cards to Anki if you already use it on your phone.
  7. Check the memory weekly to see what it thinks you struggle with, and correct it.

Prompt ideas for any tutor mode

  • "Explain this concept as if I am 15, then again at university level."
  • "Give me three questions that test whether I really understand this, not just memorised it."
  • "I got this wrong. Do not give me the answer. Give me a hint first."
  • "Make a 7-day review plan for this chapter, 20 minutes a day."

DeepTutor vs paid study modes

Most big chat apps now offer some kind of study or tutor mode. How DeepTutor compares, in general terms:

Where DeepTutor is stronger

  • You own the setup and data. Notes, memory and knowledge bases live in your workspace.
  • More learning tools in one place: quizzes, spaced review, Anki export, calendars and research reports.
  • Model choice: you pick the provider and model.
  • Open source: you can inspect it, and the Apache-2.0 license allows modification.

Where paid study modes are easier

  • No installation. Open the app and start.
  • No API setup. DeepTutor still needs a model provider, and API usage may cost money.
  • Polished mobile apps. DeepTutor's desktop apps are still in preview.
  • Support. Open-source projects rely on community help through GitHub.

So "free" is relative: the software costs nothing, but the model behind it may not, unless you run a local model on capable hardware.

Limits to know before you rely on it

  • AI tutors can be confidently wrong. Cross-check facts with your textbook or teacher. Our AI research and fact-checking workflow helps.
  • Setup takes some technical comfort. Docker or Python is required.
  • Some features need extra configuration, such as image and video generation, per the README.
  • School rules still apply. Using AI to do graded work for you can break academic rules. See our AI study rules for students.

Who should try DeepTutor

  • University students who want quizzes and review built from their own notes.
  • Self-learners following a long learning plan, such as the one in our AI learning roadmap for non-programmers.
  • Teachers and tutors exploring how AI tutoring tools work.
  • Developers who want to build on an open learning platform.

Key takeaways

  • DeepTutor had 61,082 GitHub stars on November 21, 2026, with 19,312 added in the month, under the Apache-2.0 license.
  • It combines tutoring, quizzes with spaced review, cited research, editable memory and knowledge bases.
  • Run it with Docker or pip as shown in the README, then connect a model provider in Settings.
  • It is more flexible than paid study modes but needs setup, and the model may still cost money.

FAQ

Is DeepTutor free?

The software is open source under Apache-2.0. You still need a model provider, which may charge for API use, or a local model setup.

Can DeepTutor use my own notes?

Yes. You can build knowledge bases from your own documents and let the tutor answer from them.

Do I need to know programming?

Not deeply, but you need to be comfortable running Docker or Python commands.

Sources

Read next