DeepTutor: Lifelong Personalized Tutoring with AI Agents

Introduction

DeepTutor, developed by HKUDS, is an innovative open-source project focused on providing lifelong personalized tutoring through advanced AI agents. With over 32,741 stars and 4,274 forks on GitHub, it has quickly gained recognition in the AI and education communities. Written primarily in Python, DeepTutor offers a comprehensive learning workspace that seamlessly connects tutoring, problem-solving, quiz generation, research, visualization, and mastery practice. It's designed to be an extensible system where context moves with the learner across different capabilities, making the learning experience highly adaptive and personalized.

Installation

Getting started with DeepTutor is straightforward, with several installation options available. The recommended path for a full local web application and CLI is via PyPI. Ensure you have Python 3.11-3.13 and Node.js 20+ installed.

Here's how to install DeepTutor using PyPI:

mkdir -p my-deeptutor && cd my-deeptutor
pip install -U deeptutor
deeptutor init     # prompts for ports + LLM provider + optional embedding
deeptutor start    # starts backend + frontend; keep the terminal open

After running deeptutor start, open the frontend URL printed in your terminal, typically http://127.0.0.1:3782. For detailed instructions on other installation methods, including source, Docker, or CLI-only, please refer to the official DeepTutor documentation.

Examples

DeepTutor offers both a rich web interface and a powerful command-line interface (CLI) for interacting with its capabilities.

Web Interface Highlights:
The web application provides a comprehensive suite of tools, including:

Chat: An agent loop for natural conversations, tool calls, and knowledge grounding.
Partners: Persistent AI companions with their own memory, library, and channels.
My Agents: Connect and consult live coding CLIs or import past conversations.
Co-Writer: A Markdown workspace for drafting reports and notes with AI-assisted editing.
Book: Create interactive "living books" from your materials with typed blocks and page-specific chat.
Knowledge Center: Multi-engine RAG libraries for document collections.
Learning Space: Manage skills, personas, notebooks, and question banks.
Memory: An inspectable, three-layer system for personalized learning.

CLI Examples:
The CLI allows direct interaction with DeepTutor's core functionalities:

# Interactive chat REPL
deeptutor chat

# Run a specific capability, e.g., deep solve
deeptutor run deep_solve "Solve x^2 = 4" --tool rag --kb my-kb

# Create a new knowledge base
deeptutor kb create my-kb --doc textbook.pdf

# Show current memory
deeptutor memory show

# List installed skills
deeptutor skill list

These examples demonstrate the flexibility and power of DeepTutor, whether you prefer a graphical interface or command-line control.

Why Use DeepTutor

DeepTutor stands out as a robust platform for personalized learning due to several key advantages:

Unified Learning Environment: All capabilities, from chat to quizzes and research, operate on a single agent loop. This means seamless transitions between objectives and consistent context for the learner, eliminating the need to switch between isolated tools.
Comprehensive Context Management: Knowledge bases, books, drafts, notebooks, and personas are interconnected and available across all workflows. This integrated approach ensures that all learning resources contribute to a cohesive and effective educational experience.
Advanced Agent Capabilities: DeepTutor supports subagents and partners, allowing users to consult live coding CLIs (like Claude Code or Gemini) or persistent AI companions. This multi-agent architecture enhances problem-solving and interactive learning.
Flexible Knowledge Retrieval: It offers multi-engine RAG (Retrieval Augmented Generation) libraries, supporting various retrieval engines like LlamaIndex, PageIndex, GraphRAG, and linked Obsidian vaults. This flexibility ensures efficient and accurate information retrieval from diverse document collections.
Extensible and Community-Driven: With built-in tools, CLI apps, and installable community skills from EduHub, DeepTutor is highly extensible. Its open-source nature encourages community contributions and continuous improvement.
Transparent and Inspectable Memory: DeepTutor's three-layer memory system (L1 traces, L2 curated facts, L3 synthesis) is file-backed and inspectable. This transparency allows users to understand, curate, and audit their personalized learning profile, ensuring accountability and control.

DeepTutor offers a powerful, flexible, and transparent solution for anyone looking to create or engage with intelligent, personalized learning experiences.