Bridge AI assistants with Canvas Learning Management System. Up to 101 MCP tools for students, educators, learning designers, and developers. Works with Claude, ChatGPT, Cursor, and more.
Comprehensive tools designed for real educational workflows, from simple queries to complex bulk operations.
Ask questions like "What's due this week?" or "Which students haven't submitted?" and get answers from your Canvas data.
Process bulk operations locally without loading every item into the modelβs context.
Optional data anonymization can reduce identity data before tool results reach your AI client. Compliance depends on your deployment and institutional policy.
Track student performance, assignment completion rates, peer review analytics, and more with AI-powered analysis.
Manage discussions, post announcements, facilitate peer reviews, and grade discussion participation efficiently.
Set CANVAS_ROLE to student, educator, or all to control which tool groups are registered and reduce tool-definition overhead.
Canvas MCP includes role-specific workflows for students, educators, and learning designers.
Get AI-powered assistance with your coursework and stay organized throughout the semester.
Enhance your teaching with powerful tools for course management and student engagement.
AI-powered course design, quality assurance, and WCAG-oriented accessibility review at scale.
Use the code execution API for custom bulk operations.
Tool-by-tool workflows can return every submission to the model's context. The code execution API can process items locally and return selected output.
import { bulkGrade } from './canvas/grading'; await bulkGrade({ courseIdentifier: "60366", assignmentId: "123", gradingFunction: (submission) => { // Analysis happens locally! const notebook = submission.attachments ?.find(f => f.filename.endsWith('.ipynb')); if (!notebook) return null; return { points: 100, comment: "Great work!" }; } });
Install Canvas MCP locally so the server and credentials stay in your environment. Tool results are still sent to your configured AI client under that client's data policies.
Clone the repository and set up a Python virtual environment with all dependencies.
Copy the template and add your Canvas API credentials.
Add the server to your client's MCP config. Most clients read this exact entry and differ only in where the file lives; some use their own schema (Zed nests it under context_servers). See supported clients or the per-client config blocks and file locations for Claude Desktop, Cursor, Zed, Windsurf, Continue, and others.
Test your connection and start interacting with Canvas through your AI assistant.
Local deployment and optional anonymization provide privacy controls for FERPA-conscious workflows; your institution remains responsible for its configuration and compliance.
Local mode reads your token from local configuration. Authenticated institutional deployments pass user credentials per request and do not persist them in Canvas MCP.
Educators can enable identity masking before supported tool results reach the AI client. Review the documented scope before relying on it.
Student-specific tools use Canvas's "self" endpoints. Shared course content follows the permissions Canvas grants each account.
Canvas MCP does not add product analytics. Canvas, your MCP client, and your AI provider may still retain their own logs.
A production-tested Azure deployment guide β campus SSO via Entra ID, per-user Canvas tokens, a FERPA three-tier data model, and copyable CI/CD and config templates β lives in the repository.
Canvas MCP is the only tool I've found that lets me bridge my institutionally-secured Gemini instance with a local server that keeps my credentials and student materials where they belong: on my own system. It turns a massive administrative burden, like auditing hundreds of due dates, into a 30-second conversation, all while maintaining a zero-trust security posture that would make any IT Director proud.
Pre-built skills that teach AI agents how to use Canvas MCP effectively. Install via skills.sh or use as Claude Code slash commands.
One command to add Canvas workflows to any supported agent.
Student weekly planner. Due dates, submission status, grades, and peer reviews across all courses — prioritized by urgency.
Educator morning dashboard. Submission rates, struggling students, grade distribution, and upcoming deadlines in one view.
Smart grading decision tree. Routes to single, bulk, or code execution based on submission count, with safety-first dry runs.
Full peer review pipeline. Completion analytics, quality analysis, problematic review flagging, targeted reminders, and reports.
Discussion forum facilitation. Browse, read, reply, monitor participation, and post — for both students and educators.
Pre-semester quality audit. Checks structure, content, publishing state, and completeness — generates a prioritized issue report.
WCAG-oriented review workflow. Scans content, generates prioritized reports, guides remediation, and verifies fixes.
Scaffold complete course structures from specs, templates, or by cloning. Creates modules, pages, assignments, and discussions in bulk.
Have a repetitive Canvas workflow you'd like automated? Submit an issue describing your use case!
Request a SkillCanvas MCP is designed for MCP-compatible clients; setup details and supported capabilities vary by client.
Built in the open, adopted by educators worldwide.
Join educators and students using AI to make Canvas more powerful and intuitive.