DeepTutor

Agent-native learning assistant with five modes and persistent memory

DeepTutor is an agent-native personalized learning assistant from HKUDS (Hong Kong University Data Science Lab). Unlike most "AI tutor" products that are just chatbots with educational prompts, DeepTutor was architecturally designed from the ground up for multi-step learning sessions. It offers five integrated modes: Chat (conversation), Deep Solve (step-by-step problem solving), Quiz (adaptive assessment), Deep Research (literature-style investigation), and Math Animator (visual math explanations). Version 1.0.1 shipped April 10. The persistent cross-session memory is the technical differentiator. DeepTutor tracks what you've studied, what you've struggled with, and what you've mastered across sessions, using that context to adapt its approach. This is closer to how a human tutor operates — building a mental model of the student — than the stateless Q&A loop most AI tutors offer. DeepTutor supports OpenAI, Anthropic (Claude), and DeepSeek backends, making it backend-agnostic for institutions with existing API relationships. The Math Animator mode generates step-by-step visual breakdowns of mathematical problems, which addresses one of the weakest spots in current text-based LLM math tutoring. With 1,424 stars gained in a single day and 16.1k total stars, this is clearly meeting a real demand in the education space.

Panel Reviews

The Builder

Developer Perspective

Ship

“Cross-session persistent memory is the missing piece in AI tutoring. Every other tool resets to zero each session. The five-mode architecture also makes sense — different learning tasks need different interaction patterns, not a one-size chatbot. Strong technical foundation from a credible academic lab.”

The Skeptic

Reality Check

Skip

“Academic lab projects often look impressive on GitHub but stall after the paper is published. Support burden for open-source educational tools is brutal — student use patterns are unpredictable and error-prone. The Math Animator mode sounds great but math visualization AI is notoriously unreliable for complex topics.”

The Futurist

Big Picture

Ship

“Personalized education at scale is one of AI's most transformative applications. Cross-session memory is the first step toward a true AI tutor that knows your learning style, pace, and gaps. DeepTutor is early, but the architecture is the right one for where this is going.”

The Creator

Content & Design

Ship

“For self-learners trying to pick up complex topics — design systems, coding, statistics — a tutor that remembers where you left off and adapts the difficulty is a game-changer. The quiz and deep-solve modes in particular map well to how creative professionals actually want to learn new technical skills.”