I Automated My College's Moodle So I Never Have to Open It Again
How I built a Python tool that scrapes grades, announcements, and submissions from IITB Moodle using Chrome DevTools Protocol — and posts everything to Discord.
If you've ever used Moodle, you know the pain. Slow page loads. Clicking through five tabs to find one grade. Checking announcements across ten courses. Downloading submissions you submitted three months ago.
I'm a second year student at IIT Bombay, and I got tired of it. So I automated the whole thing.
The idea
What if I could just write this:
from moodle.moodle import *
connect()
g = grades("data analytics")
post_discord(fmt_grades(g, "Data Analytics"))
And have my grades show up in Discord? No browser tabs, no clicking, no waiting 15 seconds for Moodle's JavaScript to render course cards.
That's exactly what I built.
How it works
The tool connects to a real Chrome session using the Chrome DevTools Protocol (CDP). No headless browser, no Selenium, no Playwright. It talks to your actual logged-in Chrome — the one where you're already authenticated into Moodle.
The stack is dead simple:
- browser-harness — an open source CDP wrapper (~600 lines of Python)
- A thin Moodle layer — ~300 lines that know Moodle's selectors, URL patterns, and quirks
- Discord webhook — because why check grades anywhere else
The architecture looks like this:
Chrome (port 9222) → CDP WebSocket → daemon.py → Unix socket → moodle.py
The daemon holds the WebSocket connection open. My code talks to the daemon through a Unix socket. The daemon forwards CDP commands to Chrome. It's fast and it doesn't break your browser.
What it can do
Check grades for any course:
g = grades("data analytics")
print(fmt_grades(g, "Data Analytics, AI/ML Lab"))
Submission: Lab 01 100.00 100.00 %
Lab 02 Submission 100.00 100.00 %
Lab 03 submission 98.61 98.61 % ⚠ Sklearn train test split not used
Lab 05 submission 65.00 65.00 %
...
──────────────────────────────────────────────────
TOTAL 957.11 95.71 %
Get announcements:
anns = announcements("data analytics", n=5)
print(fmt_announcements(anns))
This navigates to the course, finds the Announcements forum, scrapes the listing, visits each post for the full body, and returns clean structured data.
Download your submissions:
files = download_submission("IE 201", "lab 02")
# -> downloads/24b4506_LabL2.zip (311KB)
The trick here was using fetch() inside the browser context — Moodle requires authentication, so a raw HTTP request from Python would just get a login page. By running fetch through CDP, the browser's session cookies ride along automatically.
Post to Discord:
post_discord(fmt_grades(g, "Data Analytics"))
Reads the webhook URL from .env and posts via httpx. (Fun fact: Discord's Cloudflare blocks Python's urllib — learned that the hard way.)
The hard parts
A few things I discovered while building this:
Moodle renders course cards lazily. The page reports document.readyState === 'complete' almost immediately, but the actual course cards take ~15 seconds to appear in the DOM. You can't just wait_for_load() — you have to poll for the selector.
Grade table cells are messy. Each grade item is prefixed with its type ("AssignmentLab 03 submission") and the grade cell has "Actions Grade analysis" appended to the number. Parsing requires stripping both.
Everything is duplicated in the DOM. Course sections appear twice — once in the main content and once in the sidebar course index. Activities are doubled too. If you querySelectorAll without scoping, you get 2x results.
Course IDs and activity IDs are different things. Courses use ?id=8627 on /course/view.php, activities use ?id=83144 on /mod/assign/view.php. Same parameter name, completely different ID spaces.
I documented all of these in a navigation.md file so I don't rediscover them every time.
Course lookup
Instead of remembering course IDs, every function accepts a name substring:
grades("feedback") # -> Feedback and Dynamics
grades("IE 206") # -> Intro to AI and ML
grades(8627) # -> Data Analytics (by ID)
open_course("optimization") # -> Nonlinear and Discrete Optimization
There's a hardcoded course table for instant lookup, plus a courses() function that scrapes the live listing.
Setup
It's three steps:
- Launch Chrome with remote debugging on a separate profile
- Log into Moodle in that browser
from moodle.moodle import *; connect()
The debug profile persists your session, so you only log in once. The connect() function auto-discovers the WebSocket URL from Chrome's debug endpoint — no hardcoded UUIDs.
What's next
A few things I want to add:
- Deadline tracker — scrape assignment due dates across all courses, post daily reminders to Discord
- Grade diff notifications — run on a cron, compare with last known grades, alert on changes
- Bulk download — grab all lecture PDFs and resources from a course in one go
The whole thing is on GitHub if you want to fork it for your own university's Moodle. The Moodle-specific selectors will vary, but the architecture and the CDP approach work anywhere.
Built with browser-harness and a healthy disrespect for slow web UIs.