Canvas LMS MCP Server

An MCP (Model Context Protocol) server that connects Claude Code to your Canvas LMS account. Ask Claude about your courses, assignments, grades, and files — no manual downloading needed.

What it does

This server exposes 13 tools that let Claude Code interact with your Canvas account:

Tool	Description
list_courses	List all your enrolled courses
get_course_modules	Get the module structure for a course
get_module_items	Get items within a specific module
get_page_content	Read a Canvas page as clean text
search_course_content	Search across all pages in a course
list_course_files	List files in a course or folder
download_file	Download a file to your local disk
get_assignments	List assignments with due dates and status filters
get_assignment_details	Full assignment details with submission status
get_syllabus	Get the course syllabus
get_announcements	Get recent course announcements
get_grades	Get your grades and submission scores
export_course_content	Bulk download an entire course as a ZIP

Setup

1. Generate a Canvas Personal Access Token

1. Log into your Canvas instance (e.g. https://uncch.instructure.com)
2. Go to Account > Settings
3. Scroll to Approved Integrations
4. Click + New Access Token
5. Give it a name (e.g. "Claude Code MCP") and click Generate Token
6. Copy the token — it looks like 7006~abc123...

2. Install dependencies

cd /path/to/canvas-mcp
npm install

3. Configure in Claude Code

Add the server to your Claude Code MCP config. For global access (all projects), create or edit ~/.claude/.mcp.json:

{
  "mcpServers": {
    "canvas-lms": {
      "command": "npx",
      "args": ["tsx", "/absolute/path/to/canvas-mcp/src/index.ts"],
      "env": {
        "CANVAS_DOMAIN": "your-school.instructure.com",
        "CANVAS_ACCESS_TOKEN": "7006~your-token-here",
        "CANVAS_DOWNLOAD_DIR": "/absolute/path/to/canvas-mcp/downloads"
      }
    }
  }
}

For project-level access, create .mcp.json in your project root with the same content.

4. Restart Claude Code

Restart Claude Code and run /mcp to verify the server is connected and all 13 tools appear.

Environment Variables

Variable	Required	Description
CANVAS_DOMAIN	Yes	Your Canvas instance domain (e.g. uncch.instructure.com)
CANVAS_ACCESS_TOKEN	Yes	Personal access token from Canvas Settings
CANVAS_DOWNLOAD_DIR	No	Directory for downloaded files (defaults to ./downloads)

Usage Examples

Once configured, just ask Claude naturally:

• "List my Canvas courses"
• "What are my grades in COMP 455?"
• "Show me the modules for course 104811"
• "Download the latest homework for my CS class"
• "What assignments do I have coming up?"
• "Get the syllabus for STOR 435"
• "Search my COMP 211 course for 'linked list'"
• "Export all content from my COMP 301 course"

How it works

The server uses the Canvas LMS REST API with Bearer token authentication. It handles pagination, rate limiting, and converts Canvas HTML content to clean readable text for Claude.

Built with:

• @modelcontextprotocol/sdk — MCP protocol implementation
• cheerio — HTML to text conversion
• adm-zip — ZIP extraction for course exports
• zod — Input validation

Project Structure

canvas-mcp/
├── src/
│   ├── index.ts              # Entry point, registers tools, stdio transport
│   ├── canvas-client.ts      # Canvas API client (auth, pagination, rate limits)
│   ├── html-to-text.ts       # HTML to clean text conversion
│   └── tools/
│       ├── courses.ts        # list_courses
│       ├── modules.ts        # get_course_modules, get_module_items
│       ├── pages.ts          # get_page_content, search_course_content
│       ├── files.ts          # list_course_files, download_file
│       ├── assignments.ts    # get_assignments, get_assignment_details
│       ├── syllabus.ts       # get_syllabus
│       ├── announcements.ts  # get_announcements
│       ├── grades.ts         # get_grades
│       └── export.ts         # export_course_content
├── package.json
└── tsconfig.json