Educational Tutor

An experimental system that transforms documentation repositories into interactive educational content using AI and the Model Context Protocol (MCP).

🌟 Overview

This project consists of two main components:

1. 📚 Course Content Agent - Generates structured learning courses from documentation repositories

2. 🔧 MCP Educational Server - Provides standardized access to course content via MCP protocol

Related MCP server: repocks

🏗️ Architecture

Documentation Repository → Course Content Agent → Structured Courses → MCP Server → AI Tutors

The system processes documentation, creates educational content, and exposes it through standardized tools for AI tutoring applications.

📂 Project Structure

tutor/
├── course_content_agent/    # AI-powered course generation from docs
│   ├── main.py             # CourseBuilder orchestration
│   ├── modules.py          # Core processing logic
│   ├── models.py           # Pydantic data models
│   ├── signatures.py       # DSPy LLM signatures
│   └── about.md           # 📖 Detailed documentation
├── mcp_server/             # MCP protocol server for course access
│   ├── main.py            # MCP server startup
│   ├── tools.py           # Course interaction tools
│   ├── course_management.py # Content processing
│   └── about.md           # 📖 Detailed documentation
├── course_output/          # Generated course content
├── nbs/                   # Jupyter notebooks for development
└── pyproject.toml         # Project configuration

🚀 Quick Start

1. Install Dependencies and Create Virtual Environment

This project uses uv for fast Python package management.

# Create a virtual environment
python -m uv venv

# Install dependencies in editable mode
.venv/bin/uv pip install -e .

2. Generate Courses from Documentation

# Generate courses from a repository
.venv/bin/uv run python course_content_agent/test.py

Customize for Your Repository: Edit course_content_agent/test.py to change:

• Repository URL (currently uses MCP docs)

• Include/exclude specific folders

• Output directory and caching settings

3. Start MCP Server

# Serve generated courses via MCP protocol
.venv/bin/uv run python -m mcp_server.main

# Or customize course directory
COURSE_DIR=your_course_output .venv/bin/uv run python -m mcp_server.main

4. Test MCP Integration

# Test server capabilities
.venv/bin/uv run python mcp_server/stdio_client.py

📖 Detailed Documentation

For comprehensive information about each component:

• Course Content Agent: See course_content_agent/about.md

  • AI-powered course generation

  • DSPy signatures and multiprocessing

  • Document analysis and learning path creation

• MCP Educational Server: See mcp_server/about.md

  • MCP protocol implementation

  • Course interaction tools

  • Integration with AI assistants

🔌 MCP Integration with Cursor

To use the educational tutor MCP server with Cursor, create a .cursor/mcp.json file in your project root:

{
    "mcpServers": {
        "educational-tutor": {
            "command": "/path/to/tutor/project/.venv/bin/uv",
            "args": [
                "--directory",
                "/path/to/tutor/project",
                "run",
                "mcp_server/main.py"
            ],
            "env": {
                "COURSE_DIR": "/path/to/tutor/project/course_output"
            }
        }
    }
}

Setup Steps:

1. Create a virtual environment: python -m uv venv

2. Install dependencies: .venv/bin/uv pip install -e .

3. Update the command path and the path in args to your project directory.

4. Restart Cursor or reload the window.

5. Use @educational-tutor in Cursor chat to access course tools.

🔧 Development Status

Current Status: ✅ Functional MVP

• Course generation from documentation repositories

• MCP server for standardized content access

• Multi-complexity course creation (beginner/intermediate/advanced)

Future Enhancements:

• Support for diverse content sources (websites, videos)

• Advanced search and recommendation systems

• Integration with popular AI platforms

🛠️ Technology Stack

• AI Framework: DSPy for LLM orchestration

• Content Processing: Multiprocessing for performance

• Protocol: Model Context Protocol (MCP) for standardization

• Models: Gemini 2.5 Flash for content generation

• Data: Pydantic models for type safety

📄 License

This project is experimental and intended for educational and research purposes.