DeepTutor

Designed as a research project pushing the frontier of AI-assisted education, DeepTutor goes far beyond conventional AI study tools by providing a unified workspace with six integrated learning modes, autonomous TutorBot agents with persistent memory, and deep integration with academic knowledge bases through Retrieval-Augmented Generation.

The platform is available as an open-source project and is designed for learners who want genuinely intelligent, adaptive tutoring rather than a simple Q&A interface.

The six modes in DeepTutor's unified workspace cover the full spectrum of learning activities and share context seamlessly across each other. Chat provides standard conversational assistance.

Deep Solve attacks complex problems using a dual-loop reasoning architecture with RAG retrieval, web search, and code execution for step-by-step solutions with precise citations. Quiz Generation creates adaptive assessments from uploaded materials. Deep Research produces comprehensive literature reviews.

Math Animator visualizes mathematical concepts and proofs dynamically. Visualize generates diagrams and concept maps to represent complex relationships.

TutorBots are DeepTutor's most ambitious innovation: persistent autonomous agents that function as personal tutors rather than stateless chat sessions. Each TutorBot has its own workspace, memory, personality, and skill set that evolves over time as the learner progresses.

TutorBots can set scheduled review reminders via the Heartbeat system, learn new capabilities through tool acquisition, and proactively support learning goalsbehaving more like a dedicated human tutor who knows the learner's history than a general-purpose chatbot.