UWaterloo MCP

A read-only MCP server for UWaterloo Learn and Piazza. It reads; it never submits or posts.

Deploy your own. Anyone with your server's URL and token can read everything you can read in Learn.

Setup

1. Deploy

With Vercel

The form asks for five values:

Locally — needs Node 22.17+, and Chrome for the Duo sign-in.

git clone https://github.com/syed-zayd/uwaterloo-mcp.git
cd uwaterloo-mcp
npm install
cp .env.example .env   # edit it — the server won't start without MCP_AUTH_TOKEN
npm start

Open your server's root URL in a browser. It prints the two things steps 2 and 3 need.

2. Sign in to Learn

Piazza works already. Learn needs a Duo approval, which only you can give.

Open /setup, enter your MCP_AUTH_TOKEN, and press Sign in with UWaterloo and Duo. It shows a three-digit code — type that into Duo Mobile.

The session is held in memory, so expect to repeat this after a restart, after 180 idle minutes, and on Vercel whenever a new instance starts. To make one last, use Or paste a session cookie on that page and save the same value as a D2L_COOKIE environment variable.

3. Connect your client

Add a remote MCP server pointing at /mcp on your server. When the approval page asks for a token, paste your MCP_AUTH_TOKEN.

Related MCP server: unsw-mcp

Tools

Learn — server_info, list_courses, get_upcoming, get_grades, get_rubric, list_assignments, get_submissions, get_submission_file, get_submission_file_url, get_course_content, list_page_links, get_rubric_definition, resolve_link, get_file, get_file_url, get_quiz_attempts, get_classlist, list_groups, get_group, list_discussions, list_discussion_posts, get_discussion_thread, get_announcements

Piazza — piazza_list_courses, piazza_list_folders, piazza_search_posts, piazza_get_posts

Start with list_courses or piazza_list_courses — everything else takes a course.

get_rubric_definition returns the blank rubric for an assignment — every criterion, what it is worth, and the descriptor for each level — so you can see what is being marked before you write. get_rubric is the other one: a rubric already filled in on work that has been graded.

get_course_content lists topics, which is not everything a course holds: templates, policy PDFs and handouts usually sit in the content directory and are reached only by a link inside some page. list_page_links reads one page and lists those files with the paths get_file takes, along with the quicklinks the page carries.

Course pages link things by quicklink (quickLink.d2l?...&rcode=...) rather than by id. resolve_link follows one and says what it points at — a content topic, a rubric, an assignment folder, a discussion, or a file path — so the id can go straight into the tool that needs it. Quiz quicklinks are reported, never followed.

get_file and get_file_url name a file in either of two ways. Most files are content topics, so topic_id from get_course_content is the usual one. Some are not: an assignment template, a policy PDF or a handout linked straight from a course page has no topic of its own, and for those pass path instead — the link as it appears, e.g. /content/enforced/123456-CS_247/media/template.docx, or the full https://learn.uwaterloo.ca/... URL. A path must sit inside the named course's own content directory; anything else on Brightspace is refused.

MIT licensed.