canvas-mcp

canvas-mcp is an open-source MCP server by Boyucheng Zhang (GitHub: boyuchengz). MCP (Model Context Protocol) is an interface that lets AI assistants call outside tools. You deploy it once, add it as a custom connector in Claude or ChatGPT, and ask in plain language things like "What's due this week, and what have I already submitted?" or "What do I need on the final to keep an A?" Its 26 tools read courses, syllabi, assignment instructions, rubrics, deadlines, your scores and instructor feedback, course files, announcements, discussions and inbox messages. The author states that all 26 tools only read: they can't submit work, post or delete anything, and reading messages doesn't mark them as read. The README is in English and Simplified Chinese.

Who it's for:Students whose school uses Canvas and who want an AI assistant to help keep track of courses and assignments, provided they can deploy a small service themselves. If you don't write code, you won't need it.

Before you start: It uses your own Canvas personal access token and runs on any server with Node.js 20 or later; the author suggests Railway and also covers Render, Fly.io, your own server and local setups. Some schools restrict personal access tokens, so check your school's acceptable-use policy first and make sure it allows course content, grades and messages to be sent to an outside AI. Course content, grades, instructor comments and messages others send you all go to the Claude or ChatGPT account you use. Treat the connector URL like a password: anyone with the full URL can read everything you can see in Canvas. We don't recommend sharing that URL or connecting to a deployment someone else runs for you. If the URL leaks, change the secret and redeploy; if your token leaks, delete it in Canvas. Each deployment serves one person. The code is free under the MIT licence; hosting and AI services may cost money.

Based on the project's public documentation. Not tested by AI Inspiration World.