mograder 0.1.8

Semi-automated grading for Marimo notebooks

mograder

Semi-automated grading for Marimo notebooks.

mograder aspires to become a Marimo equivalent of nbgrader. It doesn't yet have full feature parity but should already be useable. mograder has been developed based on my experiences teaching computational modelling and machine learning with Jupyter and nbgrader in HetSys CDT and the Predictive Modelling and Scientific Computing MSc course, both at the University of Warwick, so it is a personal and opinionated take on the nbgrader workflow which may or may not suit other users!

Comparison with nbgrader and otter-grader

Feature	mograder	nbgrader	otter-grader
Notebook format	Marimo .py	Jupyter .ipynb	Jupyter .ipynb
Version control friendly	Yes	No	No
Student UX	Student dashboard	Requires JupyterHub	Minimal
Moodle integration	Native API	None	None
Parallel autograding	Yes	Manual	Docker
Sandboxed execution	Partial (uv + rlimits)	Partial (JupyterHub)	Docker
Partial credit	Yes (weighted)	Limited	Yes
Gradescope/Canvas support	No	Partial	Yes
Community/maturity	Solo/new	Mature	Active
Documentation	Minimal	Good	Good

Three modes of operation are supported:

• Workshop mode which is fully formative (i.e. no marks are assigned). Model solutions can be unlocked automatically by students when they pass automated tests. Solutions can also be released one by one by an instructor from a web dashboard, allowing 'un-sticking' of students during a live workshop.
• Manual grading of a single holistic mark. Instant formative feedback is available to students on coding questions, while reflective interpretation is manually graded.
• Hybrid grading Automated per-question marks on coding exercises, plus a single mark for the reflective component.

Two transports are currently available: Moodle integration, and HTTPS transport for standalone usage.

How it works: Instructors author source notebooks with solution blocks and automated checks. mograder generate strips solutions to create release versions. Students complete the notebooks and get instant formative feedback from check() calls. mograder autograde executes submissions in sandboxed subprocesses, parses results, and stores grades in an SQLite gradebook. GTAs review and add manual marks via the formgrader dashboard. mograder feedback exports annotated HTML for students.

Documentation

• Instructor Quickstart — step-by-step guide to setting up and grading your first assignment
• Grader API Reference — detailed docs for check(), Grader, hint(), and solution markers
• Student Setup Guide — share this with your students (local install, Codespaces, Molab)
• Security — threat model and hardening options for autograde
• Course Repo Template — starter README for your course repository

Thanks to Marimo's WASM support, notebooks with dependencies which are Pyodide compatible can be deployed as a standalone HTML file which runs entirely in students' browsers with no need for a server. It is also possible to run notebooks on MoLab, which allows for full dependencies, or to run your own Marimo edit server. Thanks to Marimo's UV integration and support for PEP 723 script dependencies (--sandbox mode) which automatically installs notebook dependencies in an isolated environment, it is also straightforward for students to install and run themselves locally.

Table of Contents

• Try it
• For students
• Directory Convention
• Workflow
• Installation
• Usage
  • Formgrader dashboard
  • Writing source notebooks
  • Generate release notebooks
  • Hints
  • Workshop notebooks with encrypted solutions
  • Validate a notebook
  • Autograde submissions
  • Export feedback
  • Import student names
  • Sync to remote server
  • Moodle integration
  • HTTPS transport (Moodle-free alternative)
  • Student dashboard
  • WASM deployment
• Configuration
• Comparison with nbgrader and otter-grader
• Development
• License

Try it

A live demo is available with three components:

1. Student Dashboard — WASM app hosted on GitHub Pages. Lists assignments and links to self-hosted WASM notebooks for editing in the browser.
2. Formgrader + Assignment Server — Combined ASGI app. The formgrader UI shows the full grading workflow (assignments, submissions, grading, students tabs) with pre-populated demo data. The same service also handles the assignment API at /assignments. No login required (for a real server, token-based authentication should be used, described below).
3. Notebook Editor — Click "Edit in Browser" in the dashboard to open a notebook as a standalone WASM app with full edit mode or "Edit in Molab" to open a full editor. Each notebook has a submit cell to send your work back to the demonstration assignment server.

There is also a Demo Workshop which is a WASM notebook demonstrating hints and encrypted solutions for formative workshops. The Instructor Dashboard controls which model solutions are visible to students. The workshop key for this demo is mograder.

A demonstration GitHub Codespaces shows how to open this repo in a Codespace for a full development environment with uv, marimo, and the student dashboard pre-configured. Assignments are served from the demo server.

For students

See the Student Setup Guide for full instructions.

Quick start (macOS/Linux — just 2 commands):

curl -LsSf https://astral.sh/uv/install.sh | sh
uvx mograder student <CONFIG_URL>

Or open in GitHub Codespaces — one click, no install needed.

Directory Convention

mograder follows nbgrader's terminology: source → release → submitted → autograded → feedback.

course/
  mograder.toml              ← optional config (dirs, moodle settings, etc.)
  gradebook.db               ← SQLite gradebook (created by autograde)
  source/
    assignment-name/
      assignment-name.py     ← source notebook (with solutions)
      data.csv               ← auxiliary files (copied to release)
  release/
    assignment-name/
      assignment-name.py     ← generated (solutions stripped)
      data.csv               ← copied from source
  submitted/
    assignment-name/
      student1.py            ← student submissions
  autograded/
    assignment-name/
      student1.py            ← output of mograder autograde
  feedback/
    assignment-name/
      student1.html          ← output of mograder feedback

Workflow

1. mograder generate — source/*.py → release/*.py (strip solutions, embed cell hashes)
2. mograder moodle upload — zip release files and open Moodle edit page for attachment
3. Students complete and submit .py files
4. mograder validate — run a notebook in a sandbox and report check results (student self-check); warns if non-solution cells have been modified
5. mograder autograde — submitted/*.py → autograded/*.py
   • Integrity check against source notebook (detects tampered check/marks cells)
   • Runs each notebook via marimo export html
   • Parses check results from HTML
   • Injects verification summary + marker feedback cells
   • Stores results in gradebook.db
6. Markers grade — formgrader Grading tab or marimo edit
   • Marker sets manual mark and feedback per student
   • Grades saved to gradebook.db
7. mograder feedback — autograded/*.py → feedback/*.html
   • Injects mark + feedback callout into existing autograde HTML
   • Removes self-assessment scores cell
8. mograder moodle export — gradebook.db + worksheet.csv → export/
   • Merges grades into Moodle offline grading worksheets
   • Bundles HTML feedback into a Moodle-compatible ZIP
   • Auto-imports student names into gradebook
9. mograder moodle login — obtain and cache a Moodle API token (supports SSO)
10. mograder moodle fetch / mograder moodle submit — students fetch assignments and submit work via Moodle API
11. mograder moodle fetch-submissions / mograder moodle upload-feedback — instructors bulk-download submissions and push grades/feedback via Moodle API
12. mograder moodle sync — syncs assignment metadata from Moodle into mograder.toml (instructor runs this, students get the config via URL or file)
13. mograder student — launches an interactive student dashboard (Marimo app) for downloading, validating, editing, and submitting assignments
14. mograder serve / mograder https * — lightweight HTTPS server + transport for assignment distribution without Moodle (HMAC token auth, atomic timestamped submissions)

Installation

Stable release:

pip install mograder # or: uv add mograder

Development version:

git clone https://github.com/jameskermode/mograder.git
cd mograder
uv venv && uv pip install -e ".[dev]"

Usage

Formgrader dashboard

Launch an interactive grading management dashboard:

mograder formgrader course/

This opens a marimo app with four tabs:

• Assignments — overview table with pipeline status and action buttons for generate, autograde, and export (feedback + Moodle merge). Source and release columns link to marimo edit.
• Submissions — per-student status for the selected assignment with marks breakdown, edit buttons, and auto/manual/total histograms.
• Grading — navigate between students with prev/next, set manual marks and feedback, auto-saved to the gradebook.
• Students — cross-assignment marks table with name lookup from the gradebook.

The formgrader reads mograder.toml from the course directory for directory names, Moodle settings, and gradebook path (see Configuration). Options: --port PORT to set the server port, --headless to suppress the browser.

For deployment as a persistent service behind a reverse proxy, use formgrader-asgi:

mograder formgrader-asgi course/ --host 0.0.0.0 --port 2718 --base-url /grading/
mograder formgrader-asgi course/ --instructors "alice,bob" --trusted-proxies "127.0.0.1"

This runs the formgrader under uvicorn with trusted-proxy authentication middleware. Use --reload for development.

Writing source notebooks

Source notebooks are standard Marimo notebooks (.py files) with a few conventions for marking solutions and autograding checks. Create them with marimo edit and place them in source/<assignment>/<assignment>.py.

Solution markers

Wrap model solutions in ### BEGIN SOLUTION / ### END SOLUTION markers. When you run mograder generate, these blocks are replaced with # YOUR CODE HERE and pass in the release version:

@app.cell
def _(np):
    def finite_diff(x, y):
        ### BEGIN SOLUTION
        dydx = np.zeros_like(y)
        dydx[0] = (y[1] - y[0]) / (x[1] - x[0])
        dydx[-1] = (y[-1] - y[-2]) / (x[-1] - x[-2])
        dydx[1:-1] = (y[2:] - y[:-2]) / (x[2:] - x[:-2])
        ### END SOLUTION
        return dydx

    return (finite_diff,)

For written-response cells, assign the model answer to _response inside a solution block. The generated release version is automatically converted to an editable mo.md() block for the student:

@app.cell
def _(mo):
    _response = "*Write your analysis here...*"
    ### BEGIN SOLUTION
    _response = r"""
    The finite difference method approximates derivatives using nearby
    function values. Central differences achieve second-order accuracy...
    """
    ### END SOLUTION
    mo.md(_response)
    return

Autograding checks

Import check from mograder.runtime and call it with a label and a list of (condition, failure_message) tuples. The result is a coloured callout (green/red/amber) that gives students instant feedback:

from mograder.runtime import check

check(
    "Q1: Palindrome checker",
    [
        (is_palindrome("racecar") is True, 'is_palindrome("racecar") should be True'),
        (is_palindrome("hello") is False, 'is_palindrome("hello") should be False'),
    ],
)

Use mo.stop() with an empty-checks call to show an amber "waiting" state before the student has written any code:

@app.cell(hide_code=True)
def _(check, mo, x):
    mo.stop(x is None, check("Q1: Array creation", []))
    check("Q1: Array creation", [
        (x.shape == (50,), f"x should have shape (50,), got {x.shape}"),
    ])
    return

Holistic vs per-question marks

Holistic mode (single mark 0-100, assigned by a marker): import the standalone check function. This is suited to notebooks where coding questions provide formative feedback only and a marker assigns one overall mark:

from mograder.runtime import check

Per-question marks (automatic + manual): use the Grader class with a marks dictionary. Questions matching a check() label are auto-scored with partial credit: marks are proportional to the weight of passing checks. Questions without a matching check (e.g. written analysis) are scored manually by the marker:

from mograder.runtime import Grader

# === MOGRADER: MARKS ===
_marks = {"Q1": 10, "Q2": 15, "Analysis": 60}
grader = Grader(mo, _marks)
check = grader.check

Each check tuple can optionally include a weight as a third element (default weight is 1). Earned marks are round(available × earned_weight / total_weight, 1):

check("Q2: Finite differences", [
    (isinstance(dydx, np.ndarray), "result should be ndarray"),       # weight 1
    (dydx.shape == x.shape, "shape should match"),                    # weight 1
    (np.max(np.abs(dydx - np.cos(x))) < 0.05, "max error < 0.05", 3),  # weight 3
])
# If only the first two pass: earned = round(15 * 2/5, 1) = 6.0/15

The question key is the text before the first colon in the check label, so check("Q1: Array creation", [...]) maps to the "Q1" entry. Call grader.scores() in a cell to display a reactive score table showing earned/available marks (including fractional values for partial credit).

PEP 723 script dependencies

Include a PEP 723 metadata block at the top of the notebook so that marimo edit --sandbox and mograder validate can automatically install dependencies:

# /// script
# requires-python = ">=3.11"
# dependencies = [
#     "marimo",
#     "numpy",
#     "mograder",
# ]
# ///

mograder generate automatically adds mograder-assignment and mograder-cell-hashes lines to this block in the release notebook, enabling integrity checking during mograder validate.

Generate release notebooks

Strip solution blocks from source notebooks:

mograder generate hw1                             # by assignment name
mograder generate source/hw1/hw1.py -o release/   # by file path
mograder generate hw1 --dry-run                   # preview only
mograder generate hw1 --validate                  # check markers only
mograder generate hw1 --submit-url https://server.example.com  # inject submit cell

Arguments without / or .py suffix are treated as assignment names and resolved to files in the source directory. Auxiliary files (data, helper modules) are automatically copied from the source directory.

Use --submit-url to inject a submit cell into release notebooks, allowing students to submit directly from within the notebook to an HTTPS assignment server.

Hints

Use hint() in notebooks to provide progressive hints as collapsed accordions:

from mograder.runtime import hint

# Single hint — accordion label is "Hint"
hint("Think about what preserves insertion order")

# Multiple hints — numbered "Hint 1", "Hint 2", ...
hint(
    "Think about which data structure preserves insertion order",
    "Consider using `collections.OrderedDict`",
    "Use `OrderedDict.move_to_end()`"
)

Workshop notebooks with encrypted solutions

For formative workshops (ungraded, deployed as WASM on GitHub Pages), mograder supports encrypted solutions with two reveal mechanisms:

• Auto-reveal on check pass — when a student's check() passes and they've entered the workshop key, the model solution appears automatically below the check result
• Instructor release — the instructor can release solutions progressively during a live workshop via keys.json; students click "Check for released solutions" to fetch them

The workshop key is a secret shared verbally by the instructor during the session. It is not stored in the generated notebook (only a SHA-256 hash is embedded), so students cannot extract solutions from the source code.

1. Author a source notebook with a # === MOGRADER: EXERCISES === cell listing exercise keys, alongside regular ### BEGIN/END SOLUTION markers, check() calls, and optionally hint() for progressive hints:

# === MOGRADER: EXERCISES ===
_exercises = ["Q1", "Q2"]

2. Generate the workshop notebook — encrypts solutions, strips them, and injects solution-reveal cells. The command prints the workshop key for the instructor to share verbally:

mograder workshop encrypt source/workshop/workshop.py -o release/workshop/ --salt mykey
# Workshop key (share with students verbally): mykey

3. Export for WASM deployment — generates HTML + keys.json for GitHub Pages:

mograder workshop export source/workshop/workshop.py -o dist/workshop/ --salt mykey

4. Release solutions during a live workshop — incrementally reveal solutions by adding decryption keys to keys.json:

mograder workshop release-key dist/workshop/keys.json Q1 --salt mykey

Students click "Check for released solutions" in the notebook to fetch updated keys. Released solutions appear regardless of whether checks pass or the workshop key has been entered. To release all solutions at once, copy keys_all.json over keys.json.

5. Serve locally with an instructor dashboard — for live workshops, serve the exported directory with a built-in instructor dashboard for releasing solutions in real time:

mograder workshop serve dist/workshop/ --salt mykey

This starts a local server and prints two URLs: one for students (the WASM notebook) and one for the instructor (a dashboard to release solutions incrementally). The instructor dashboard URL includes a randomly generated secret token.

Validate a notebook

Run a notebook in a sandbox and report check results (useful for students to self-check before submitting):

mograder validate hw1.py
mograder validate hw1.py --timeout 600
mograder validate hw1.py --fix --release release/hw1/hw1.py

Installs dependencies in a sandbox, executes the notebook, and prints PASS/FAIL for each check. Exits with code 1 if any check fails. An HTML report is saved alongside the notebook.

If the release notebook was generated with mograder generate (v0.1.1+), cell hashes are embedded in the PEP 723 metadata block. validate compares these hashes against the current cells and warns if any non-solution cells have been accidentally modified. Use --fix to restore them from the release version (found automatically in .mograder/release/ if previously fetched, or specify --release <path>).

Autograde submissions

Run student notebooks and prepare grading copies with injected feedback cells:

mograder autograde hw1                            # by assignment name
mograder autograde submitted/hw1/*.py -o autograded/hw1/
mograder autograde submitted/hw1/*.py --source source/hw1/hw1.py --csv results.csv
mograder autograde hw1 -j 8 --timeout 600

When --source is provided (or auto-discovered from a sibling source/ directory), mograder performs an integrity check — tampered check cells or marks definitions are reinjected from the source before execution. Default values for -j and --timeout can be set in mograder.toml (see Configuration).

Use --force to re-grade all submissions even if the output is already up to date. Use --safety-check to scan submitted code for dangerous patterns before execution.

Autograde directly from Moodle downloads

Instead of manually extracting submissions, you can pass the Moodle offline grading CSV and submission ZIP directly:

mograder autograde --moodle-csv grades.csv --moodle-zip submissions.zip --source source/hw1/hw1.py

This extracts submissions from the ZIP (mapping participant IDs to usernames via the CSV), then runs the normal autograde flow. The output directory and assignment name are inferred from the source notebook path.

Export feedback

Export graded notebooks to HTML and aggregate marks:

mograder feedback hw1                             # by assignment name
mograder feedback autograded/hw1/*.py -o feedback/hw1/
mograder feedback hw1 --grades-csv grades.csv

Import student names

Import student names from a Moodle CSV into the gradebook (used for name display in the formgrader):

mograder import-students worksheet.csv

Sync to remote server

Sync autograded results to a remote server (e.g. a shared formgrader instance) via rsync + SSH:

mograder sync autograded/hw1/ --remote sciml --course-dir /home/svc_user/courses/es98e

This rsyncs .py and .html files to the remote autograded/ directory, then runs Gradebook.import_from_py() on the server via SSH to update the remote gradebook. If the remote uses a uv-managed venv, pass --venv-dir:

mograder sync autograded/hw1/ --remote sciml --course-dir /home/svc_user/courses/es98e --venv-dir '~/marimo-server'

All three flags can be set in mograder.toml (see Configuration) so you can just run mograder sync autograded/hw1/.

Autograded results can also be uploaded via the formgrader UI using the upload button in the Graded column of the Assignments table.

Moodle integration

The mograder moodle command group provides both offline CSV-based workflows and live Moodle API access.

Export grades (offline)

Merge grades into a Moodle offline grading worksheet and bundle feedback:

mograder moodle export "HW1" -o export/
mograder moodle export "HW1" --feedback-dir feedback/ -o export/
mograder moodle export "HW1" --worksheet custom.csv -o export/

The worksheet is auto-discovered at import/<assignment>.csv (matching formgrader convention). Grades are read from gradebook.db by default. The match column and name column can be configured in mograder.toml (see Configuration). Student names are auto-imported into the gradebook when the moodle command runs.

Fetch assignment (student)

Download assignment files from Moodle:

mograder moodle fetch "HW1"                     # download by name
mograder moodle fetch "HW1" -o ~/coursework/     # custom output directory
mograder moodle fetch --list                     # list available assignments

Downloads all attached files (.py notebooks and .zip archives with input data). ZIP files are automatically extracted. Assignment matching is flexible: exact name, numeric ID, or case-insensitive substring.

Submit assignment (student)

Upload a completed notebook to Moodle:

mograder moodle submit "HW1" hw1.py              # upload and finalize
mograder moodle submit "HW1" hw1.py --dry-run    # check without uploading
mograder moodle submit "HW1" hw1.py --no-finalize  # upload draft only

Only .py files are accepted. By default, submissions are finalized (visible to graders). Use --no-finalize to save as draft.

Fetch submissions (instructor)

Bulk-download all student submissions for an assignment:

mograder moodle fetch-submissions "HW1" -o submitted/hw1/

Downloads each student's latest .py submission, named by username.

Upload release files (instructor)

Zip release files and open the Moodle assignment edit page for manual attachment:

mograder moodle upload "HW1"                    # auto-discovers from release/HW1/
mograder moodle upload "HW1" file1.py data.csv  # explicit files
mograder moodle upload "HW1" --dry-run          # preview without creating zip
mograder moodle upload "HW1" --no-open          # create zip without opening browser

Files are zipped into <assignment>.zip in the current directory. If no files are given, all files in release/<assignment>/ are included. The Moodle assignment edit page is opened automatically so you can attach the zip as an introattachment.

Upload feedback (instructor)

Push grades and feedback to Moodle via the API:

mograder moodle upload-feedback "HW1"                          # from gradebook.db
mograder moodle upload-feedback "HW1" --dry-run                # preview without pushing
mograder moodle upload-feedback "HW1" --grades-csv grades.csv  # from CSV
mograder moodle upload-feedback "HW1" --feedback-dir feedback/HW1/  # with HTML feedback files
mograder moodle upload-feedback "HW1" --workflow-state released     # make grades visible immediately

Sync assignment metadata (instructor)

Fetch assignment metadata from Moodle and write it to mograder.toml:

mograder moodle sync                          # sync all visible assignments
mograder moodle sync --include '^A[1-8]'      # only assignments matching regex

Only assignments visible to students are included (hidden assignments are excluded). Students receive this metadata via the config URL or file. Re-run after publishing or hiding assignments.

View feedback (student)

Check your submission status and view grade/feedback:

mograder moodle feedback "HW1"

Shows submission status, grade (if graded), and instructor feedback text.

Login (obtain API token)

Obtain and cache a Moodle API token for subsequent commands:

mograder moodle login                    # username/password prompt
mograder moodle login --sso              # browser-based SSO (CAS/SAML/Shibboleth)
mograder moodle login --url https://moodle.uni.ac.uk

The token is cached at ~/.config/mograder/token.json. For SSO sites, --sso opens the Moodle Security Keys page in your browser where you can copy the token.

All Moodle API commands accept --url and --token flags, or read from MOGRADER_MOODLE_URL / MOGRADER_MOODLE_TOKEN environment variables, or from the [moodle] section in mograder.toml.

HTTPS transport (Moodle-free alternative)

mograder includes a lightweight HTTP server and transport for distributing assignments without Moodle. This is useful for courses that don't use Moodle, for local testing, or as a simple course server on platforms like Molab.

Start an assignment server

mograder serve course/release/        # serve assignment files
mograder serve course/release/ -p 9000  # custom port

The server auto-discovers assignments from the directory structure. Each subdirectory with a files/ subfolder becomes an assignment. You can also provide a manual assignments.json manifest. Use --release-dir to serve files from a flat release/<assignment>/<file> layout instead of requiring a files/ subdirectory.

Authentication

Authentication is enabled by default. The server generates a secret (.mograder-secret) on first start and uses HMAC-SHA256 tokens in the format username:hmac_hex.

Student self-registration (recommended): Set an enrollment code so students can register themselves via the student dashboard or API:

mograder serve course/release/ --enrollment-code "my-course-phrase"
# or via environment variable:
MOGRADER_ENROLLMENT_CODE="my-course-phrase" mograder serve course/release/
# or from a file:
mograder serve course/release/ --enrollment-code-file enrollment.txt

Students enter their username + enrollment code in the dashboard to receive a personal token. The enrollment code can be shared in class or via LMS — it is separate from the HMAC secret.

Generate tokens manually (alternative): Use mograder token to generate tokens directly:

mograder token alice bob carol                      # reads .mograder-secret from CWD
mograder token --secret-file path/to/.mograder-secret alice bob
ssh server "cat /path/.mograder-secret" | mograder token --secret-stdin alice bob

Or from the serve command with a file of usernames (one per line):

mograder serve course/release/ --generate-tokens students.txt

Disable auth for local testing with --no-auth.

Token roles:

• Student tokens — can list/download assignments, submit own work, check own status
• Instructor token — full access: list submissions, download any submission, upload grades

Student commands

Students register via the student dashboard (enter username + enrollment code), or cache a token manually:

mograder https login --token <YOUR_TOKEN> --url https://server.example.com
mograder https fetch --list                              # list assignments
mograder https fetch "hw1" -o hw1/                       # download files
mograder https submit "hw1" hw1.py                       # submit work
mograder https feedback "hw1"                            # check status/grade

The URL and token can also be passed explicitly with --url and --token flags.

Instructor commands

mograder https fetch-submissions "hw1" --url https://server.example.com --token <INSTRUCTOR_TOKEN> -o submitted/hw1/
mograder https upload-grades "hw1" --url https://server.example.com --token <INSTRUCTOR_TOKEN> --grades-csv grades.csv

The URL can also be set in mograder.toml:

transport = "https"

[https]
url = "https://server.example.com"

Server directory structure

server_root/
  .mograder-secret                    # HMAC secret (auto-generated)
  assignments.json                    # optional manifest
  hw1/
    files/
      homework.py                     # assignment files
    grades.json                       # uploaded grades

submitted/                            # submission storage (configurable)
  hw1/
    alice_20260310T200800.py          # timestamped submissions
    alice.py -> alice_20260310T200800.py  # symlink to latest

Submissions are written atomically with timestamped filenames, preserving history across resubmissions. A symlink <user>.py always points to the latest version.

Student dashboard

Launch an interactive course browser as a local Marimo web app:

mograder student <CONFIG_URL>       # first-time setup from URL
mograder student                    # current directory (returning sessions)
mograder student ~/coursework/      # specific course directory
mograder student --port 8080        # custom port
mograder student --headless         # no browser auto-open

The dashboard provides:

• Login — for Moodle courses, paste your Moodle security token (from your Moodle Security Keys page). For HTTPS transport courses, register with your username and the enrollment code provided by your instructor (or paste a token directly). Tokens are cached at ~/.config/mograder/.
• Assignment table — lists all course assignments with due dates, status, check validation results, and action buttons.
• Download — downloads assignment .py files into per-assignment subdirectories.
• Edit — opens the notebook in a new marimo edit --sandbox session.
• Validate — runs the notebook and shows a summary of check results (e.g. "3/5 PASS") with an inline HTML report preview. Warns if non-solution cells have been accidentally modified. Results are cached and marked stale when the notebook changes.
• Submit — uploads the .py file to Moodle and finalizes the submission.
• Status tracking — shows Downloaded, Submitted, or Modified for each assignment.
• Activity log — shows status messages for recent actions with dismiss button.

WASM deployment

mograder provides commands for deploying notebooks as standalone WASM apps (e.g. on GitHub Pages):

mograder wasm-export hw1                      # export a single assignment
mograder wasm-export --all                    # export all WASM-compatible assignments
mograder wasm-export --check-only             # check compatibility without exporting
mograder wasm-export hw1 --mode run           # export in run mode (default: edit)

wasm-export checks each assignment's dependencies against Pyodide and runs marimo export html-wasm for compatible ones.

To inject pre-computed "Edit in Molab" links into a WASM student dashboard app:

mograder wasm-edit-links student_app.py release/hw1/hw1.py release/hw2/hw2.py
mograder wasm-edit-links student_app.py release/*/*.py -o output_app.py

Each notebook is compressed with lzstring and embedded as a Molab URL, keyed by the notebook's filename stem.

Configuration

Create mograder.toml in the course directory to customise settings:

config_url = "https://raw.githubusercontent.com/user/course/main/mograder.toml"
transport = "moodle"   # or "https" — selects the active transport for student/formgrader

# Transport-agnostic assignment list (written by `moodle sync` or `https sync`)
[[assignments]]
name = "HW1"
id = "10"
cmid = "42"
duedate = 1700000000
  [[assignments.files]]
  name = "hw1.py"
  url = "https://..."

[dirs]
source = "source"       # default directory names
release = "release"
submitted = "submitted"
autograded = "autograded"
feedback = "feedback"
import = "import"       # Moodle worksheets for export

[moodle]
url = "https://moodle.uni.ac.uk"  # Moodle site URL (for API commands)
course_id = 12345                  # Moodle course ID (for API commands)
csv = "moodle.csv"                 # default Moodle worksheet (for export)
match_column = "Username"
name_column = "Full name"

[https]
url = "http://localhost:8080"      # HTTPS transport server URL
token = ""                         # cached auth token

[defaults]
jobs = 4
timeout = 300
no_edit = false                    # disable "Edit" buttons in formgrader
no_actions = false                 # disable action buttons in formgrader
headless_edit = false              # open marimo edit in headless mode

[rlimits]                          # resource caps for notebook subprocesses
cpu = 600                          # CPU time limit in seconds
nproc = 64                         # max user processes
nofile = 256                       # max open file descriptors

[gradebook]
path = "gradebook.db"

[sync]
remote = "sciml"                                    # SSH host alias
remote_course_dir = "/home/svc_user/courses/es98e"  # course dir on remote
remote_venv_dir = "~/marimo-server"                 # uv venv dir on remote (optional)

[edit_links]                       # custom "Edit in ..." links for the student dashboard
molab = "https://molab.marimo.io/new/#code/{content_lz}"

Comparison with nbgrader and otter-grader

Feature	mograder	nbgrader	otter-grader
Notebook format	Marimo .py	Jupyter .ipynb	Jupyter .ipynb
Version control friendly	Yes	No	No
Student UX	Student dashboard	Requires JupyterHub	Minimal
Moodle integration	Native API	None	None
Parallel autograding	Yes	Manual	Docker
Sandboxed execution	Partial (uv + rlimits)	Partial (JupyterHub)	Docker
Partial credit	No	Limited	Yes
Gradescope/Canvas support	No	Partial	Yes
Community/maturity	Solo/new	Mature	Active
Documentation	Minimal	Good	Good

Development

uv run pytest              # run tests
uv run ruff check src/     # lint

License

MIT