Canvas LMS MCP. Manage users, courses, and assignments with a single conversation.

Canvas LMS MCP Server manages everything in your Canvas account directly through your AI agent. You can run admin tasks—the kind that usually takes clicking through a dozen menus—using natural language commands. It's built to handle the full lifecycle of courses, users, and content across institutional accounts.

Course Structure and Content Management

The list_courses tool shows every course visible in your primary catalog view; you can then grab the full structure and details for any specific class using get_course. You don't just read courses, either. You can set up an entirely new course shell with create_course, or archive/remove a whole thing using delete_course.

Need to adjust how old classes run? Use update_course to modify overall settings and structure. The system lets you manage content depth by listing structured modules with list_modules, then fetching the full list of items—pages, assignments—inside a module via list_module_items. You can read specific pages using get_page or grab titles and locations for all pages in a course with list_pages.

For assignments, you've got it covered: you can see every assignment active in a course using list_assignments, pull details for one assignment instance using get_assignment, or create brand-new ones with create_assignment. You'll need to update them sometimes; use update_assignment to change due dates, points, or instructions. When students turn stuff in, you can list all their attempts through list_submissions and submit the final grade score using grade_submission.

Quizzes, Discussions, and Files

When it comes to quizzes, you can see every quiz created for a course using list_quizzes. You'll generate new ones with create_quiz, or retrieve all question banks, settings, and results for an existing one with get_quiz. For discussions, the list_discussion_topics tool retrieves all active topics in a specific area; you can then post a brand-new thread using create_discussion_topic.

You can also check out general messaging. Use list_conversations to see your whole inbox, or get detailed message history and metadata from any single chat thread using get_conversation; remember, you'll start new chats with create_conversation.

The content storage is easy too. You can list files saved in a module or folder using list_files, and view the metadata for a specific file by running get_file. If you need to know what's inside a directory, use get_folder after getting the contents via list_folders.

User Accounts, Roles, and Admin Rights

Managing people is key. The create_user tool builds brand-new user profiles in Canvas. You can update specific profile details—like names or permissions—for any existing account using update_user. To check on a person's record, use get_user to see their detailed profile and enrollment status. For admin work, you don't have to guess; the server lets you list all users who hold admin privileges with list_admins, or assign new admin powers directly using create_admin.

You can define roles—like TA or Instructor—by running create_role and retrieving a full list of available roles using list_roles.

The system also helps you manage the institution's structure. Use list_accounts to see all connected institutional accounts, and if your domain has affiliates, get_sub_accounts lists them for you. You can modify general settings across the entire account framework with update_account, or get detailed metadata about a specific institutional account using get_account.

Data Flow and Progress Tracking

Need to bring in student data? Run create_sis_import to start the process, then check its status—running or done—with get_sis_import_status. You can track what's happening across the board by checking the user's current list of administrative or academic 'To Do' items using get_todo, or fetching a timeline of all actions taken in Canvas via get_activity_stream.

Finally, for highly specialized data retrieval that doesn't fit standard commands, you can run custom GraphQL queries against the whole database using execute_graphql.