DeepTutor

DeepTutor is an open-source, agent-native learning workspace for lifelong personalized tutoring, with more than 44,000 stars on GitHub. Developed by the HKUDS lab, it turns your documents, books, notes, and videos into a tutor that teaches, quizzes, researches, and tracks your mastery over time, all inside one connected system.

Overview

DeepTutor brings every learning mode into a single runtime: chat, asking questions, quizzes, research, visualization, problem solving, course study, mastery paths, immersive reading, and audio overviews share one session context while keeping purpose-built workflows. You build knowledge bases from your own files, and the tutor answers with citations grounded in those sources. It runs as a web app or command-line agent, supports many model providers, and ships frequent releases. Documentation lives at deeptutor.info, and the project is released under the Apache 2.0 license.

Key Features

• Every learning mode in one workspace: Tutoring, problem solving, quiz generation, research, visualization, and mastery practice run on one capability runtime with shared context.
• Multi-engine knowledge: Versioned retrieval libraries across LlamaIndex, PageIndex, GraphRAG, and LightRAG, plus linked Obsidian vaults, Kiwix archives, and self-hosted knowledge bases, with pluggable document parsing.
• Guided learning and mastery paths: Structured courses and mastery gates track what you know and generate questions from your materials into a personal question bank.
• Immersive reading and video learning: Open documents beside the chat with page-level citations, or add YouTube and Bilibili links for transcript search, timestamp-grounded tutoring, and resumable progress.
• Subagents and Partners: Consult live agent harnesses such as Claude Code, Codex, and Gemini from inside chat, or run persistent tutor companions with their own persona, library, and memory.
• Inspectable three-layer memory: Traces, surface summaries, and synthesis make personalization visible and editable, so you can see why the tutor believes what it believes.
• Extensible tools and skills: Built-in tools, MCP servers, CLI apps, image and voice models, and installable community skills extend the tutor without forking the project.

💳 Pricing

DeepTutor is free and open source under the Apache 2.0 license, and you can self-host it with Docker. You bring your own model credentials, so running costs depend on the models you choose.

Who Is It For

DeepTutor is for self-learners, students, researchers, and educators who study from their own materials and want a tutor that remembers their progress, tests their understanding, and answers from their sources. It suits anyone building a long-term knowledge base rather than asking one-off questions.