DeepTutor

DeepTutor is a self-hosted AI tutor for learners who want to study their own materials through a browser or CLI. It supports Docker deployment and uses configured model providers for its AI capabilities. Hosting the app locally is separate from model inference; requests to a remote provider leave the host.

Tutoring, problem solving, quizzes and research share session context with course study and mastery practice. Knowledge bases, books, notebooks, question banks and writing drafts can carry material between supported learning activities. Separate workspaces keep conversations, study materials and progress apart, including in multi-user deployments.

Personalization is inspectable. Learners can edit memory and follow links between its summaries and supporting evidence. A workspace task board tracks study tasks and notes, with an archive for restoring cards.

For document-based study, DeepTutor connects to Obsidian vaults, Kiwix archives, MarginNote 4 and self-hosted WeKnora libraries. Retrieval engines include LlamaIndex, PageIndex, GraphRAG and LightRAG. YouTube study adds synchronized captions, tutoring tied to timestamps and saved progress; playback can use a self-hosted Invidious instance.

The code is open source under Apache 2.0. MCP servers and community skills extend its tools, and configured models provide image, video and voice generation. Chat can consult agent systems such as Claude Code and Codex, while the CLI exposes the same capabilities and knowledge bases through an interactive terminal interface or structured JSON.