Canvas Student

GitHub
Canvas Student MCP

A Model Context Protocol (MCP) integration for interacting with Canvas LMS. This repository was created as a test of how to interface Canvas with LLM clients through the MCP standard. It was "vibe-coded" during an evening as an exploration of the MCP capabilities and Canvas API for students.

Note on PDF Handling: Unfortunately, there's a limitation regarding PDF parsing seemingly due to restrictions on Claude's internet access. A complete solution would require an additional API service for text extraction from PDFs, which wasn't implemented in this project. The current implementation provides links to PDFs but cannot fully extract and display their content within Claude.

Setup

Prerequisites

Installation

Option 1: Simple Installation (Recommended)

  1. Clone this repository
  2. Run the installation script:
# Navigate to the canvas-student directory
cd src/canvas-student

# Run the installer Python script
python install.py

This script will:

  • Detect if you have UV or pip available
  • Install the package with PDF support
  • Provide next steps for configuration

Option 2: Manual Installation

  1. Clone this repository
  2. Install dependencies:
# Navigate to the canvas-student directory
cd src/canvas-student

# Install with UV (including PDF support)
uv pip install -e ".[pdf]"

# Or with standard pip
pip install -e ".[pdf]"

Configuration for Claude Desktop

Edit your Claude Desktop configuration file:

  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Add the following configuration:

{
  "mcpServers": {
    "canvas": {
      "command": "python",
      "args": ["ABSOLUTE_PATH_TO/src/canvas-student/main.py"],
      "env": {
        "CANVAS_API_TOKEN": "your_canvas_api_token",
        "CANVAS_BASE_URL": "https://your-institution.instructure.com"
      }
    }
  }
}

Replace:

  • ABSOLUTE_PATH_TO with the absolute path to your project
  • your_canvas_api_token with your Canvas API token
  • your-institution.instructure.com with your institution's Canvas URL

Note: The tool now uses Canvas's preview capabilities for files like PDFs, providing URLs for direct access when previews aren't available.

Troubleshooting

If you encounter issues:

  1. Check the Claude Desktop logs:

    • Windows: %APPDATA%\Claude\logs\mcp-server-canvas.log
    • macOS: ~/Library/Logs/Claude/mcp-server-canvas.log

  2. Verify the path in your configuration is correct

  3. Make sure you're using Python 3.10 or newer

Authentication Issues

If you encounter authentication problems:

  1. Check your API token: Make sure your Canvas API token is valid and not expired. You can create a new token from your Canvas account settings under "Approved Integrations".

  2. URL Format: Ensure your Canvas URL is correctly formatted:

    • Should start with https://
    • Should NOT include /api/v1/
    • Should NOT have a trailing slash
    • Example: https://canvas.institution.edu (correct)
    • Example: https://canvas.institution.edu/ (incorrect - has trailing slash)

  3. Environment Variables: If you're getting "CLIENT_ID not configured" errors but are using API token authentication, this is likely because your API token isn't being recognized. Check that your environment variables are correctly set in either:

    • Your .env file
    • Your system environment
    • The Claude Desktop configuration

  4. Permissions: Ensure your Canvas user account has sufficient permissions for the actions you're trying to perform.

Features

  • Retrieve and search courses
  • Find and filter assignments
  • Access course content (files, modules, pages)
  • Search across all Canvas content
  • OAuth authentication
  • View text file content directly in Claude
  • Get links for other file types (PDFs, images, etc.)

File Content Handling

The Canvas Student MCP tool provides capabilities for handling different file types:

  1. Text Files (.txt, .md, .csv, .json): Content is displayed directly in Claude
  2. PDFs and Images: URLs are provided for viewing or downloading these files
  3. Canvas Previews: When available, Canvas preview content can be shown

This approach doesn't require external libraries and works reliably across all platforms.

Example usage:

Show me the content of the assignment description file from course 27849

Configuration

Create a .env file with your Canvas credentials:

CANVAS_API_TOKEN=your_token_here
CANVAS_BASE_URL=https://your-institution.instructure.com
CANVAS_CLIENT_ID=your_oauth_client_id  # Optional for OAuth
CANVAS_CLIENT_SECRET=your_oauth_client_secret  # Optional for OAuth
REDIRECT_URI=http://localhost:8000/oauth/callback  # Optional for OAuth

Usage

Running the server directly

For standard usage:

python canvas_student.py

For guaranteed PDF support:

python run_server.py

Using with Claude for Desktop

  1. Find your Claude Desktop configuration file (usually in ~/.config/Claude/claude_desktop_config.json on macOS/Linux or %APPDATA%\Claude\claude_desktop_config.json on Windows)
  2. Add this server to your configuration:
{
  "mcpServers": {
    "canvas": {
      "command": "python",
      "args": ["ABSOLUTE_PATH_TO/src/canvas-student/run_server.py"],
      "env": {
        "CANVAS_API_TOKEN": "your_canvas_api_token",
        "CANVAS_BASE_URL": "https://your-institution.instructure.com"
      }
    }
  }
}

Replace:

  • ABSOLUTE_PATH_TO with the absolute path to your project
  • your_canvas_api_token with your Canvas API token
  • your-institution.instructure.com with your institution's Canvas URL

Note: The run_server.py script automatically ensures PDF extraction libraries are installed before starting the server. This provides a more reliable solution for PDF support.

  1. Restart Claude Desktop
  2. Look for the MCP icon when chatting with Claude

Examples

Here are some things you can ask Claude:

  • "Show me all my courses"
  • "Find exams in my Data Communication course"
  • "What assignments are due in the next week?"
  • "Search for content about 'machine learning' across all my courses"
  • "Extract and show me the content of Lab1P.pdf from course 27849"
  • "Show me my to-do items and missing assignments"
  • "What quizzes do I have coming up in my courses?"
  • "Show me the details for quiz 12345 in course 67890"
  • "What assignments are due in the next 14 days?"

Project Structure

  • canvas-student.py: Main MCP server
  • main.py: Entry point for the MCP server
  • tools/: Tools organized by category
    • courses.py: Course-related tools
    • assignments.py: Assignment-related tools
    • content.py: Content/files related tools
    • search.py: Search tools
    • utils.py: Helper utility tools
    • file_content.py: PDF and text content extraction
    • todos.py: To-do items and missing assignments tools
    • quizzes.py: Quiz information and preparation tools

License

MIT