Built for self-learners and exam candidates already working with AI assistants. Consyllus turns authoritative textbooks into Directed Acyclic Knowledge Graphs (DAGs) and feeds your real-time mastery state straight into your AI via the Model Context Protocol.
Live topological derivation: completing a node automatically unlocks downstream concepts when all prerequisites are satisfied.
From a raw textbook syllabus to structured, prerequisite-aware tutoring in three deterministic stages.
Provide your textbook table of contents or exam syllabus to your MCP-connected AI assistant. Consyllus guides the agent through a 4-stage builder pipeline (nodes, concept dictionary, prerequisite projection, and two-level textbook outline).
Every concept dependency is projected into a strict Directed Acyclic Graph (graph.json) verified by Kahn cycle detection and sealed with an RFC 8785 canonical SHA-256 content hash.
Your AI tutor reads your live mastery boundary via MCP—knowing what you have mastered, what is currently unlocked, and which blackbox terms to defer—while you check off nodes and unlock successors.
Every feature is grounded in deterministic data contracts, strict physical separation of objective knowledge and private progress, and local-first ownership.
Node statuses (Completed, Study Target, Available, Locked) are derived at runtime from your completion set and the prerequisite DAG. Unmarking a foundational node automatically triggers a reverse lock cascade to keep causal dependencies intact.
The bundled zero-Node-dependency MCP server exposes your active subject, mastered concepts, prerequisite gaps, and error logs directly to your AI client—eliminating repetitive prompt setup.
Objective curriculum graphs (graph.json & outline.json) are strictly separated from your personal study records (progress.json). Your progress, study targets, and logs stay 100% on your local disk and are never uploaded.
Generate a self-hosted Ed25519 identity secured by your OS keyring (DPAPI / Keychain / Libsecret) with 12-word BIP-39 or AES-256-GCM keyfile backup. Browse and download community graphs with in-memory signature, schema, and DAG verification before disk write.
Connect Consyllus to any desktop or CLI assistant that supports the standard Model Context Protocol (MCP). Copy the generated JSON snippet from Preferences in one click.
Runs out of the box using the bundled runtime—no separate Node.js or Python installation required on your machine.
Direct answers on pricing, local data storage, supported platforms, and AI clients.
Consyllus Desktop and the public Graph Hub are free to use. Because Consyllus connects to your own AI client via MCP, you simply use your existing AI subscription or API key—Consyllus does not charge middleman token fees.
All of your subjects, graphs (graph.json), textbook outlines (outline.json), and personal learning progress (progress.json) are stored as plain, human-readable UTF-8 JSON files inside your local Vault folder. Your personal progress is never transmitted to any server.
Consyllus currently provides a native 64-bit desktop installer for Windows (Windows 10 / 11 x64), complete with OS-level DPAPI keyring protection and a bundled zero-dependency MCP server.
Any AI client that supports the Model Context Protocol (MCP) over stdio works with Consyllus—including Claude Desktop, Claude Code, Cursor, Google Antigravity, Windsurf, and VS Code.
