cselab — Run CSE autotest from your laptop

The 6-step loop you do 50 times a day

Every edit-test cycle takes 30–60 seconds of overhead.

The SSH Loop

1. Open terminal

2. $ ssh z5555555@login.cse.unsw.edu.au

3. Enter password (again...)

4. $ cd ~/COMP1521/lab01

5. $ scp files from local machine

6. $ 1521 autotest

... fix code, repeat from step 5. Every. Single. Time.

Over a 10-week term, you waste hours on SSH logistics instead of learning.

With cselab: one command.

Save. Run. Done.

cselab

$ cselab run "1521 autotest"

[1/3] Connecting... OK (0.0s)

[2/3] Syncing... OK (0.2s)

[3/3] Running: 1521 autotest

✓ All tests passed!

VS Code, Cursor, Vim — any editor

Watch mode: auto-test on save

Zero dependencies

How cselab compares

Feature	VLAB	SSH FS	Remote-SSH	cserun	cselab
Use local editor	No	Partial	Yes	Yes	Yes
Run autotest / give	Yes	Separate SSH	Yes	Yes	Yes
Server load	High	Low	Very High	Low	Near Zero
Reliability	OK	Good	Gets killed	45%	100%
Watch mode	No	No	No	No	Yes
Install method	Browser	VS Code plugin	VS Code plugin	Rust toolchain	pip install
AI editor support	No	No	No	No	Yes
Offline editing	No	No	No	No	Yes

Built for CSE students

Everything you need, nothing you don't.

Any Editor

VS Code, Cursor, Windsurf, Vim, Emacs — write code in whatever you want. cselab handles the rest.

Sub-Second Feedback

rsync delta sync: 0.2s. SSH execute: 0.15s warm. Your edit-test cycle drops from 60s to under 1s.

Watch Mode

Save your file, autotest runs automatically. No switching windows, no typing commands.

Works Offline

Edit code on the train, at a cafe, anywhere. Sync and test when you're back online.

For Course Coordinators & System Admins

VS Code Remote-SSH spawns persistent Node.js processes on login servers. CSE already deploys reaper scripts to handle the load. Students keep using it because there's no better alternative.

Server Load Comparison

Remote-SSH

~200–500 MB RAM/student

SSH FS

~50 MB/student

cselab

~0 MB idle, pulse connections

If 100 students switch from Remote-SSH to cselab, your login servers free up 20–50 GB of RAM.

Zero server-side components — standard SSH + rsync only

No persistent connections — ControlMaster socket, disconnect after command

Reduces SSH connection churn — ControlMaster reuses one connection

Open source (MIT) — fully auditable, no security concerns

Compatible with all courses using autotest / give / classrun

Review the source code

Get started in 3 steps

Requires Python 3.10+ (pre-installed on macOS and most Linux).

1

Install

$ pip install cselab

2

Configure

$ cselab init

# Enter your zID and password

3

Run

$ cd ~/COMP1521/lab01