Four things are gone forever. Start there.
The classroom layer that sat on top of a code host was decommissioned on 28 August 2026; its APIs answer 410 Gone. The metadata it still held is deleted on 4 September 2026. No import — ours or anyone's — brings that back, so this page states what you lost before it states anything we could do for you.
What no migration can recover
Four classes of data, permanently deleted. This is not a caveat at the bottom of an offer; it is the first thing on the page because it is the true thing about your situation.
Classroom and assignment names
The shape of the course as the product held it — which classroom, which assignment, what it was called. It was never in a repository, so nothing on the code host carries it.
Tests defined outside repositories
Autograding tests configured in the product's own interface rather than committed as files. If a test was never a file in the repo, there is no copy of it anywhere.
Historical test-run and submission data
Every autograding run, every score it produced, every timestamp, for every student, across every term the course ran. This is the largest loss and the one nobody can undo.
Every roster synced from a learning-management system
Not degraded, not partial. Deleted. Rosters existed only inside the classroom layer and were keyed to host handles, so they went when it went.
If you are reading this before 4 September 2026, the most useful thing you can do involves no vendor at all: clone the organisation's repositories, save the membership list, take a grades export while the option still exists, and keep any workflow file that defines your tests. Do that first. Decide what you use next afterwards.
What survived, and the reason it did
Everything on this list lives on GitHub rather than inside the product that was retired. That is the whole distinction — what survived, survived because it was never in the thing that died.
Repositories, with their commit history
Contents and full history. Clone them and you have them.
Accounts and organisations
The GitHub accounts and orgs themselves are untouched — they were never part of the retired product.
The organisation membership list
A list of host handles. Useful, and not the same thing as a roster: a synced roster carried institutional identities, and a handle is not a name.
Any autograding workflow still committed
A workflow file under .github/workflows/ is a file in your repo, so it survived. A test configured in the product's UI did not.
A grades export, if one was taken
A grades.csv you exported before the shutdown is yours. If nobody exported one, there is nothing to export now.
What we could import — and what exists today
Every feature named here carries its status, and the statuses are the honest part of this section. None of it is running this week.
Rescue Import
Coming soonDesigned to take a GitHub org, match per-student repositories by naming convention, infer the assignment from the template, parse a surviving autograding workflow into a grader config, and accept an optional grades export. It is not built. It is a specified, unstarted task in our plan, self-serve only when it lands, and there is nothing you can import with today. We are telling you this on the page rather than after a sales call.
The free instructor tier
Coming soonIntended shape: one active course, up to 50 active learners, one instructor seat, unlimited assignments graded on public repositories, invite links and CSV roster import, and community-forum support. Educator verification before activation. Also unbuilt — the caps are a design decision, not a live limit you can go and test.
LTI 1.3 launch, Deep Linking and grade passback
Coming soonPlanned as part of the free tier rather than as an upsell, because roster-and-grade plumbing is the thing an institution cannot work around. Unbuilt. When it ships, an LTI launch will create or link a miatz account the learner keeps — a code host and an LMS are integrations here, never a way to log in, which is precisely why the deleted rosters could not have happened this way.
What we do not promise
The exclusions are the product decision, not the small print. A free tier that refuses nothing is a free tier that ends.
Any human migration service, onboarding or hands-on support. Community forum only, and there are no exceptions to that — a per-free-user human cost is what has killed every free tier that died of generosity.
Reconstruction of run history or a deleted roster. Not by us, not by anyone. It is not a capability question; the data is gone.
Indefinite retention. The intended free-tier rule is the active term plus twelve months, then archive-or-delete, stated at signup and enforced by a sweeper rather than by good intentions.
Grading on compute we do not pay for, beyond a published quota. If a quota is what makes a tier survivable, the quota gets published rather than implied.
Beyond the free tier, a paid instructor plan is intended to carry unlimited courses and learners, our own grading runners under a published quota, roster sync, similarity detection, extra TA seats and email support. We have not set its price, so we are not printing one.
What we do not know yet
Four things we would have to check before claiming them. Publishing the gaps costs us a tidier page and buys you the ability to check our work.
Do review comments on submission pull requests survive?
Probably — they live on the code host rather than in the retired product. We have not verified it, and the notice we are working from does not say. To answer it we would need the vendor's own decommission documentation.
What time, in what time zone, does the 4 September deletion happen?
We do not know. Treat the date as the deadline and act before it, not on it.
Does an existing grades export carry rubric or per-criterion detail?
Unknown. It may be final scores only. One real export from one real course would settle it, and we do not have one.
Can an org membership list rebuild a usable roster?
It gives you host handles and nothing else. Whether that maps back to your students depends on records you hold, not on anything we or the code host can supply.
We are also not a neutral party. We build in this category, and this page exists partly because we would like you to consider us later. The post-mortem says the same thing at more length, including the part where none of what we decided differently has actually shipped.
Export first. Decide later.
Nothing on this page asks you to commit to us before 4 September. Take your repositories, your membership list and any grades export you can still get, and come back when you are choosing rather than salvaging.
Things that do run today, if you want to judge us on something real: the playground catalog, the published course catalog and what actually shipped, dated.