What it does

DeepTutor is a self-hosted web app and command-line tool for personalized tutoring. Chat, quizzes, research, visualizations, problem solving, course study, mastery practice, immersive reading and audio overviews share one runtime and session context. Learners can build knowledge bases from their own material using several retrieval engines, including LlamaIndex, PageIndex, GraphRAG and LightRAG, or link an Obsidian vault. Memory is kept in three layers that can be viewed and edited, and a task board tracks study tasks across To do, In progress and Done. From chat, DeepTutor can consult agent tools such as Claude Code or Codex, or persistent partner agents that also run in messaging channels. It supports MCP servers, CLI apps and installable community skills, and the page describes learning from YouTube or Bilibili links with transcript search and timestamp-grounded tutoring. Install with pip (Python 3.11 to 3.14 and Node.js 20 or newer) or try it with a single Docker command, then configure your model providers in Settings. The project is Apache 2.0 licensed and ships frequent releases.

Capabilities

Chat, quiz, research, solve and course modes on one runtime
Knowledge bases with several retrieval engines
Three-layer memory you can inspect and edit
Immersive reading and YouTube or Bilibili video learning
Task board for study tasks
MCP servers, CLI apps and community skills

When to use it

You want a self-hosted AI tutor that works from your own documents and notes
You are preparing for a subject and want quizzes, mastery practice and guided courses
You want to study from books, papers or videos with tutoring grounded in the source
You prefer to bring your own model providers and keep settings and data local