canvas-lms-mcp: Skill for Cursor

Adaptive Course Learner

Build reinforcement learning systems that iteratively learn from Canvas course and module content using perceptrons and RL techniques.

Quick Start

Setup:

mkdir adaptive-learner && cd adaptive-learner
uv init --name adaptive-learner --python 3.10+
uv add numpy scikit-learn torch transformers

Project structure:

adaptive-learner/
├── knowledge_base/         # Learned knowledge storage
│   ├── course_{id}/       # Per-course folders
│   │   └── module_{id}/   # Per-module folders
├── models/                 # Trained models
│   ├── perceptron.pkl     # Perceptron model
│   └── rl_agent.json      # RL agent Q-table
├── learner.py              # Main system
├── perceptron_model.py     # Perceptron implementation
├── rl_agent.py             # RL agent
└── canvas_fetcher.py        # Content fetcher

Architecture

Three-component system:

1. Content Fetcher - Retrieves Canvas course/module data via API/MCP
2. Perceptron Network - Extracts features and patterns using MLP
3. RL Agent - Learns optimal context building strategies via Q-learning

Implementation Pattern

Step 1: Content Fetcher

canvas_fetcher.py:

import httpx
import json
from pathlib import Path
from config import load_env_config, get_api_headers

class CanvasContentFetcher:
    def __init__(self):
        self.config = load_env_config()
    
    async def fetch_course_content(self, course_id: int) -> dict:
        async with httpx.AsyncClient(
            base_url=self.config.base_url,
            headers=get_api_headers(self.config.api_token)
        ) as client:
            modules = (await client.get(f"/api/v1/courses/{course_id}/modules")).json()
            for module in modules:
                module["items"] = (await client.get(
                    f"/api/v1/courses/{course_id}/modules/{module['id']}/items"
                )).json()
            return {"course_id": course_id, "modules": modules}
    
    def save_to_folder(self, content: dict, base_path: Path):
        course_path = base_path / f"course_{content['course_id']}"
        course_path.mkdir(exist_ok=True)
        for module in content["modules"]:
            module_path = course_path / f"module_{module['id']}"
            module_path.mkdir(exist_ok=True)
            (module_path / "items.json").write_text(json.dumps(module["items"], indent=2))