Lextures

The first truly adaptive learning environment

Lextures uses AI to streamline the process of course creation, quiz generation, and content management, enabling educators and learners to get to the content as quickly as possible

Start studying

Lextures

Open-source learning platform for running courses end to end: structured modules, calendars, grading, and enrollments—with AI hooks when you want them—so instructors and students spend less time on tooling and more on teaching and learning. Ships a React web app, native iOS and Android clients, a Tauri desktop shell, a Go CLI, and an MCP server for AI agents.

Features

• Adaptive learning: Quizzes that adjust difficulty in real time using Item Response Theory (IRT 2PL/3PL), plus spaced-repetition review and mastery tracking.
• Course workspace: Build structured modules with TipTap-powered rich content, assignments, vibe activities (interactive AI lessons), and drag-and-drop organization.
• Teaching & grading: Calendars, gradebooks, Speed Grader, optional AI grading agent, enrollment management, and a 1:1 inbox.
• Standards-based grading: Map assignments to NGSS, CCSS, or custom standards and track mastery by objective with full audit trails.
• Communication: Course discussions, activity feed channels, and direct messaging between users.
• Notebooks: Personal and per-course notebooks with markdown, drawings, tasks, and slash commands.
• Integrations: LTI 1.3 provider/consumer (Canvas, Moodle, Blackboard); SAML 2.0, OIDC, and SCIM; Clever and ClassLink for K–12; Canvas course import and QTI; MCP for Cursor, Claude Desktop, and other AI agents.
• AI (optional): Multi-provider BYOK course-grounded tutor, quiz generation, misconception detection, and automated hint scaffolding (OpenRouter, Anthropic, OpenAI, Azure OpenAI, Bedrock, Vertex).
• 14 question types: From multiple choice and essays to live code execution and audio/video responses.
• Cross-platform clients: Web SPA, native iOS (SwiftUI) and Android (Jetpack Compose) apps, Tauri desktop, and a lextures CLI—all talking to the same REST API.
• Accessibility: Immersive reader (read-aloud, captions, translation), accommodations engine, and WCAG-oriented UI work across web and mobile.

Repository

lextures/
├── server/              # Go API, migrations, background jobs
├── clients/
│   ├── web/             # React LMS (primary web app)
│   ├── ios/             # Native iPhone app (SwiftUI)
│   ├── android/         # Native Android app (Jetpack Compose)
│   ├── desktop/         # Tauri 2 desktop shell (wraps web)
│   ├── cli/             # lextures terminal client
│   └── mcp/             # Model Context Protocol server
├── www/                 # Marketing site (lextures.com)
├── e2e/                 # Playwright end-to-end tests
├── iac/                 # Terraform self-host stack (AWS; other cloud modules reserved)
└── docs/                # Architecture, plans, and guides

Tech stack

Layer	Choices
Web app	React 19, Vite 8, TypeScript 6, Tailwind CSS v4, React Router, TipTap, Vitest
Mobile	SwiftUI (iOS 17+), Kotlin + Jetpack Compose (Android)
Desktop	Tauri 2 (Rust + embedded web client)
CLI	Go + Cobra
API	Go 1.25, Chi, pgx, Argon2id passwords, JWT access tokens
MCP server	TypeScript stdio server in clients/mcp (Model Context Protocol)
Marketing site	React 19 + Vite in www/
Data	PostgreSQL 16
Queue	RabbitMQ (async jobs, e.g. Canvas import)
AI (optional)	Provider credentials (BYOK) in Settings → Intelligence → Models — see docs/ai-providers-byok.md

For architecture notes (Compose port layout, dev vs prod web, testing conventions), see docs/ARCH.md. For where new code belongs and structural CI ratchets, see docs/ARCHITECTURE_CONVENTIONS.md.

Getting started

Quick start (Docker, recommended):

# Set first Global Admin email before anyone signs up (repo root .env or shell export)
export [email protected]

make dev   # Postgres, RabbitMQ, API :8080, Vite :5173

Open http://localhost:5173 and sign up with the same email as BOOTSTRAP_ADMIN_EMAIL.

See Getting started for prerequisites, production-style Compose, promoting an existing user to Global Admin, and local development without full Docker. The marketing site also publishes Self-hosting (source: www/src/docs/self-hosting.md). Terraform layouts for single-VM deploys live in iac/.

Native mobile (optional): copy clients/mobile-dev.env.example to clients/mobile-dev.env, run bash clients/scripts/setup-mobile-dev.sh, then open clients/ios/Lextures.xcodeproj or clients/android in Android Studio. See clients/ios/README.md and clients/android/README.md.

Development

Task	Command
Start dev stack	make dev
Lint all apps	make lint
E2E suite (Playwright)	make e2e
E2E against running stack	make e2e-run
Mobile lint + tests	make mobile
Build desktop app	make desktop
Build CLI	make cli

Contributor setup details, env vars, and gotchas: AGENTS.md. Web client conventions (code splitting, bundle budgets): clients/web/CONTRIBUTING.md.

MCP integrations

Connect AI agents in Cursor, Claude Desktop, or any MCP client to your Lextures instance. The MCP server (clients/mcp) runs locally over stdio and calls the Lextures API with a personal access key.

Setup

1. Build the MCP server (from the repo root):

   cd clients/mcp && npm install && npm run build
   

2. Create an access key in the web app under Settings → Integrations. Include the MCP: Connect scope (mcp:connect) plus any data scopes your agent needs (for example courses:read, courses:write, assignments:read, assignments:write, files:read, feed:read, enrollments:read). Copy the key when shown — it is only displayed once. The key’s user must also have course RBAC permission to create/edit items (item:create) for authoring tools.

3. Add MCP config to your client. Open the project with this repository as the workspace root so clients/mcp/dist/index.js resolves. Example for Cursor or Claude Desktop:

   {
     "mcpServers": {
       "lextures": {
         "command": "node",
         "args": ["clients/mcp/dist/index.js"],
         "env": {
           "LEXTURES_API_URL": "http://localhost:8080",
           "LEXTURES_API_TOKEN": "<paste-your-access-key>"
         }
       }
     }
   }
   

   • Cursor: Settings → MCP (or .cursor/mcp.json in the project).
   • Claude Desktop: ~/.claude/claude_desktop_config.json.

   The Settings → Integrations panel also provides a ready-to-copy JSON snippet with your instance’s API base URL.

Tools

Tool	Description
whoami	Authenticated user profile
list_courses	Courses visible to the key (optional term filter)
list_structure	Course modules and items (optional kind filter)
list_assignments	Assignment metadata in a course
list_quizzes	Quiz metadata in a course
list_content_pages	Content page metadata in a course
read_assignment	Full assignment content and metadata
read_quiz	Full quiz (questions, settings, intro)
read_content_page	Content page markdown body
create_module	Create a course module
create_content_page	Create a content page (optional markdown + publish)
update_content_page	Update content page markdown
create_assignment	Create an assignment under a module
update_assignment	Update assignment content and settings
create_quiz	Create a quiz (optional questions + publish)
update_quiz	Modify quiz questions and delivery settings
publish_structure_item	Publish or unpublish any structure item
list_content_tool_catalog	Available content tools for a course
get_content_tool_manifest	Tool manifest + configSchema (e.g. inline_questions)
list_content_tool_instances	Tool instances (optional item filter)
create_content_tool_instance	Create a tool instance with settings
update_content_tool_instance	Update tool config / title / status
attach_content_tool	Create instance and embed a lex-tool fence in host markdown
list_enrollments	Course roster
list_activity_feed	Feed messages from the last N days
list_files	Files and folders in a course file space
read_file	Download a course file (text as UTF-8, binary as base64)

Contributing

Contributions are welcome. Everyone who participates is expected to follow the Code of Conduct (Contributor Covenant 2.1).

1. Fork the repository and create a branch for your change.
2. Make focused commits with clear messages.
3. Open a pull request describing what changed and why.

Report security issues per SECURITY.md — do not open public issues for vulnerabilities.

License

This project is licensed under the GNU Affero General Public License v3.0 — see LICENSE.

---

Lextures — getting to the content, faster.