Hero
Connect the Canvas LMS MCP to your AI agent and hand off the repetitive parts of teaching. Grade submissions against your rubric, message a whole section in seconds, schedule announcements, and pull the analytics you actually need. It runs on your own instructor login, on the courses you already teach, with nothing new to learn.
Set CANVAS_ROLE=teacher and your agent gets the same tools you have in Canvas, driven by plain language.
What your agent can do
Grade submissions
Point your agent at an assignment and let it work through the ungraded pile. It reads each submission, scores it against the rubric you define, drafts per criterion feedback, and posts grades back to the Canvas gradebook. You review before anything goes live.
Message students in bulk
Send a targeted message to a whole section, a group, or just the students who missed a deadline. Your agent pulls the right recipient list from the course roster and sends through Canvas Conversations, so replies land in your inbox like normal.
Post and schedule announcements
Draft an announcement once and have it posted now or scheduled for later. Reuse a template across every section you teach, adjust the date per course, and keep your whole roster in sync without copy and paste.
Pull course and student analytics
Ask for the numbers that matter this week: submission rates, score distributions, students trending below a threshold, participation gaps. Your agent queries Canvas analytics and hands back a plain summary you can act on, no dashboard digging.
Manage assignments
Create, update, and organize assignments from a prompt. Change due dates, adjust points, move items between modules, set availability windows, or clone an assignment across sections. Your agent makes the edits in Canvas and confirms what changed.
Export the gradebook
Pull a full or filtered gradebook export whenever you need it, for progress reports, department reviews, or your own records. Get current grades, missing work, and category breakdowns in a clean file ready to share or archive.
Ask your agent
> "Grade all ungraded submissions in CS101 against the rubric, draft feedback for each, and show me the grades before you post them." > Your agent pulls every ungraded submission, scores each one on your rubric criteria, writes feedback, and waits for your sign off before writing anything back to Canvas.
> "Message every student in Bio 210 who hasn't submitted the Week 6 lab, reminding them it's due Friday at 11:59pm." > Your agent builds the recipient list from actual submission status, drafts the reminder, and sends it through Canvas Conversations once you approve.
> "Show me every student in my three sections who dropped below 70% this month, and export the gradebook so I can flag them for advising." > Your agent pulls analytics across all your sections, surfaces the at risk students, and generates a gradebook export you can hand to your advising team.
Authorization and student privacy
The MCP acts only as you, on the courses your Canvas account is authorized to teach. It uses your instructor credentials and the standard Canvas API, so it can never reach a course, section, or student record you couldn't already access yourself. There is no separate data store and no shared access.
Student records are protected education data. Under FERPA, the grades, feedback, and roster information you work with here are covered records, and the responsibility for how they are used stays with you. Keep exports on approved, institution sanctioned systems, share student data only with people who have a legitimate educational interest, and follow your school's data handling policy. Treat every AI drafted grade or message as a draft: you review and approve before it reaches a student or leaves your control.
Get started
Install the Canvas LMS MCP, add your Canvas API token, and set CANVAS_ROLE=teacher. You'll be handing off busywork in a few minutes, on the courses you already run.
[Read the teacher setup guide](/docs/teachers) · [Get your Canvas token](/docs/token)