canvas-mcp

Ask your AI about Canvas assignments — in plain English.

Query assignments, grades, announcements, syllabi, and more — works with Claude, ChatGPT, Gemini, and any MCP-compatible AI.

Download

The desktop installer bundles everything — no Node.js or terminal needed.

| Platform | Download | |---|---| | Mac (Apple Silicon) | canvas-mcp-installer_0.1.0_aarch64.dmg | | Windows | canvas-mcp-installer_0.1.0_x64-setup.exe |

Mac: right-click the app → Open on first launch (one-time Gatekeeper bypass)
Windows: if SmartScreen appears, click More info → Run anyway

Setup (3 steps)

1. Find your Canvas URL

Log in to Canvas and copy the base URL from your browser — it looks like one of these:

https://university.instructure.com
https://canvas.school.edu
https://yourschool.edu/canvas

Copy everything up to (but not including) /courses or /dashboard.

2. Get your Canvas API token

Canvas → Account → Settings → scroll to Approved Integrations → + New Access Token

3. Run the installer

The installer will ask for your Canvas URL and token, then configure your AI client automatically.

Or set up via terminal

npx -y @fireballff/canvas-mcp setup --client claude      # Claude Code
npx -y @fireballff/canvas-mcp setup --client chatgpt     # Cursor / OpenAI
npx -y @fireballff/canvas-mcp setup --client gemini      # Gemini CLI
npx -y @fireballff/canvas-mcp setup --client openrouter  # Cline / OpenRouter

Then restart your AI client and ask:

"What assignments do I have due today?"
"What's due in Physics in the next 48 hours?"

Tools

List all courses you're currently enrolled in.

Try: "What classes am I taking?" · "Show me my courses."

Assignments due across all courses, or for one specific course.

| Parameter | Type | Default | Description | |---|---|---|---| | hours_ahead | number | 24 | How far ahead to look | | course_name | string | — | Required for the single-course version |

Try: "What's due today?" · "What's due in Physics this week?"

Current grades across all courses, or for one specific course.

Try: "What are my grades?" · "What's my grade in Calc?"

All assignments that are past due and not submitted.

Try: "What am I missing?" · "What haven't I turned in?"

Your score, grade, and instructor comments for a specific assignment.

| Parameter | Type | Description | |---|---|---| | course_name | string | Course name or partial match | | assignment_name | string | Assignment name or partial match |

Try: "What did I get on the Physics midterm?" · "Any feedback on my essay?"

Recent announcements from all courses, or one specific course.

| Parameter | Type | Default | Description | |---|---|---|---| | days_back | number | 7 | How far back to look | | course_name | string | — | Required for the single-course version |

Try: "Any new announcements?" · "What did my Bio professor post this week?"

Calendar events (exams, office hours, course events) from all courses.

| Parameter | Type | Default | Description | |---|---|---|---| | days_ahead | number | 14 | How far ahead to look |

Try: "Any exams coming up?" · "What events do I have this week?"

The full syllabus for a specific course — great for asking about policies.

| Parameter | Type | Description | |---|---|---| | course_name | string | Course name or partial match |

Try: "What's the late policy in Bio?" · "How is my Physics grade calculated?"

The module/topic structure for a specific course.

| Parameter | Type | Description | |---|---|---| | course_name | string | Course name or partial match |

Try: "What topics are covered in Calc?" · "What's coming up in week 4?"

Manual config

{
  "mcpServers": {
    "canvas": {
      "command": "npx",
      "args": ["-y", "@fireballff/canvas-mcp"],
      "env": {
        "CANVAS_API_URL": "https://yourschool.instructure.com",
        "CANVAS_API_KEY": "your-token-here"
      }
    }
  }
}

| Client | Config file | |---|---| | Claude Code | ~/.claude/claude_mcp_config.json | | Gemini CLI | ~/.gemini/settings.json | | Cursor | ~/.cursor/mcp.json | | Cline | ~/.cline/mcp_settings.json |

Privacy

canvas-mcp runs entirely on your machine. Your Canvas data goes directly from your school's Canvas to your AI client — it never touches any server we operate. There is no analytics, telemetry, or data collection of any kind.

See PRIVACY.md for full details.

Contributing

Contributions welcome! Planned for v2:

• Announcements — get_all_announcements, get_course_announcements
• Grades
• More AI client integrations

git clone https://github.com/fireballff/canvas-mcp
cd canvas-mcp
npm install
cp .env.example .env   # add your Canvas credentials
npm test

License

MIT