Canvas MCP

The infrastructure layer that gives 40+ AI coding agents full access to Canvas LMS through natural language.

90+ Tools
8 Agent Skills
290+ Tests

Built for Three Audiences

Students, educators, and learning designers each get tailored tools and skills

🎓

Students

Check assignments, track grades, plan weekly work, and participate in discussions—all through natural language with their AI coding agent.

  • canvas-week-plan: Weekly assignment planner across all courses
  • canvas-discussion-facilitator: Browse, read, and reply to discussions
  • Submission tracking and grade monitoring
📚

Educators

Grade at scale, manage peer reviews, send reminders, and monitor course health—operations that used to take hours now take minutes.

  • canvas-bulk-grading: Grade with rubrics at scale via dry-run workflow
  • canvas-peer-review-manager: Track completion, analyze quality, send reminders
  • canvas-morning-check: Daily course health dashboard
  • FERPA-compliant data anonymization
🛠

Learning Designers

Build and audit courses programmatically. Create modules, pages, assignments, and rubrics through natural language, then audit against quality standards.

  • canvas-course-builder: Create course structure from specs
  • canvas-accessibility-auditor: Scan content for accessibility issues
  • canvas-course-qc: Quality control against design standards
  • Code execution API for 99.7% token savings on bulk operations

Technical Architecture

Open source, modular, and available everywhere AI agents run

🔧

Core

  • Python 3.10+ with FastMCP framework
  • 91 tools across 15 modules
  • Async httpx + pydantic validation
  • 290+ pytest tests
🚀

Distribution

  • PyPI: pip install canvas-mcp
  • MCP Registry (official listing)
  • skills.sh: npx skills add
  • Hosted: mcp.illinihunt.org/mcp
🤖

Agent Compatibility

  • Claude Desktop & Claude Code
  • Cursor, Windsurf, Continue, Zed
  • OpenAI Codex CLI
  • Any MCP-compatible agent
🔒

Safety

  • FERPA-compliant anonymization
  • Dry-run workflow for bulk grading
  • Read-only defaults for student tools
  • Rate limiting and access controls

Impact

Making Canvas LMS the first AI-native learning management system

Open Source

Fully open source on GitHub with comprehensive documentation, guides for all three user personas, and real-world workflow examples for common educator tasks.

Zero Setup Option

The hosted MCP server at mcp.illinihunt.org means anyone can start using Canvas with AI agents immediately—no installation, no configuration, just add the URL.

Token Efficiency

The code execution API enables bulk operations (grading 200 students, updating all pages) with 99.7% token savings compared to individual tool calls.