Skip to content

Overview

DeepTutor is organized around connected learning surfaces that all share one agent-native runtime. This section walks through the surfaces you touch every day, then the library, memory, and control layers underneath them — in the order they build on each other.

DeepTutor's home — the Chat workspace, with every surface in the sidebar

#SurfaceOne-linerPage
1Chat WorkspaceThe default agent loop: capabilities, sticky context, tools, generation, and the Activity panel in one threadChat Workspace
2PartnersPersistent IM-connected companions with their own soul, library, memory, and channelsPartners
3SubagentsConsult a live Claude Code, Codex, or Partner from a turn — and import their conversationsSubagents
4Co-WriterSelection-aware Markdown editing, grounded in your KB or the web, always shown as a diffCo-Writer
5BookCompile your materials into interactive living books — quizzes, animations, widgetsInteractive Book
6Learning SpaceChat history, notebooks, question bank, mastery paths, personas, and skillsLearning Space
7MemoryL1 traces, L2 per-surface facts, L3 synthesis, and a Memory Graph you can auditMemory
8Knowledge CenterVersioned RAG libraries across a choice of retrieval enginesKnowledge Center
9SettingsAppearance, network, models, parsing, chat tools, subagents, and memorySettings
  • Chat is the default loop. Chat, Quiz, Research, Visualize, Solve, and Mastery Path all run through the same orchestrated runtime — you switch the objective, not the engine.
  • Partners reuse Chat, not a separate bot. A partner message becomes a scoped chat turn inside data/partners/<id>/workspace/.
  • Subagents bring other agents in. A connected Claude Code, Codex, or Partner can be consulted live from a turn, its full run streamed into the Activity panel.
  • Context comes in two kinds. Sticky session context (model, persona, knowledge, subagent) lives on the composer toolbar; one-time references come from the + menu.
  • Memory is inspectable. The graph lets you trace any synthesized fact back to L2 summaries and the raw L1 events behind them.

Start with Chat Workspace for the core interaction model, then read Partners to put DeepTutor inside your IM channels.