Play this game ↗ · About this game · By alphaBETA DEV · Published 2026-09-24 · Updated 2026-10-19 · Editorial standards · Changelog

From a browser tab to a store listing

Room 404 has been installable from the Microsoft Store since 23 September 2026. It is free, published under alphaBETA DEV, product ID 9NKW0PHF7LQH.

Store page: https://apps.microsoft.com/store/detail/9NKW0PHF7LQH?cid=DevShareMCLPCS

There is no native rewrite. What you install is still the game running at room404.gamestrategists.com — wrapped as an MSIX so Windows treats it as an application: searchable in Start, pinnable to the taskbar, openable with no network at all.

This is the whole path, including the parts the official docs mention only in passing. The re-checked figures at the end were measured again on 2026-10-19, not carried over from September.

Where the numbers come from: on 2026-10-19 I took screenshots of the store product page and the game's title screen and checked them against what the listing actually showed; the offline-cache size is an estimate of what the app can load with no network. Every re-checked figure here was measured that day, not carried over from September.

1. Microsoft does not package it for you

The common assumption is that you submit a URL and Microsoft builds the app. It is three steps, and the order cannot change:

1. Reserve the product name in Partner Center, and note the Package ID, Publisher ID and Publisher display name;
2. Generate the package with PWABuilder from your PWA URL, then download the zip containing the .msixbundle and the .classic.appxbundle (plus a signing script — the Store signs on its servers, so you can skip it for submission);
3. Upload both bundles back in Partner Center, fill in the listing and submit.

Product names are unique globally, not per account. Reserve first. Discovering a collision after you have filled in the listing means redoing all of that work.

Review took about 24 hours.

2. The manifest is baked into the package

This is the most expensive constraint. An MSIX freezes the web manifest into the artifact:

• any later change to the app's install manifest — icons, name, display orientation, shortcuts — means repackage and resubmit;
• changes to HTML, CSS, JS or the service worker do not. Users get those on the next launch.

So treat the manifest as a release freeze point and review it properly before generating the package, not as something you can keep editing.

3. orientation: portrait bricks the desktop build

The original manifest declared orientation: "portrait". That is reasonable for a phone, but on Windows it locks the app window to portrait, which makes the desktop experience effectively unusable.

Changed to "any orientation", and added a stable install identity. That identity is how Windows recognises the app: without it, renaming the app or swapping icons can make Windows treat it as a different application, which silently breaks existing installs. Both changes are in the store version note; the line currently shown under “What's new in this version” is: “Window orientation unlocked for desktop use; stable install identity added so updates keep your saved progress.”

Both had to be fixed before generating the MSIX. If you package first and fix after, you run the whole pipeline a second time.

Re-checked on 2026-10-19: the two games that followed, MrQiWinWinWin and Dajin Weight-Loss GoGoGo, also use "any orientation" with a stable install identity in their live builds. The fix has become a template, not a one-off decision.

4. Offline works for the shell, not the images

Before packaging I tested with the network cut in a real browser: after a reload the body rendered, with 16 precached entries and 684 KB.

On 2026-10-19 I measured the offline cache again: 20 declarations, 19 unique entries (one item is listed twice), and summing what actually loads gives about 1,574 KB. Most of the growth is later assets: the main JS bundle at 596 KB, CSS at 88 KB, three PWA icons totalling 797 KB, and four official Microsoft Store badge SVGs totalling 96 KB.

Be precise about what offline covers: the shell, not all resources. The app pre-saves its scripts, styles, page skeleton and some icons, but images are fetched on the first open and then cached (up to 60 images, kept for 30 days) — so the first offline launch shows empty image slots. The page skeleton still loads, so the text renders.

Adding images to the precache would fix it at the cost of pushing install size up by several megabytes. The tradeoff accepted here is "offline playable, artwork missing", because the puzzle layer is text, audio and interaction state; you can finish the game without the pictures.

Store review sometimes tests offline, so this is a known residual risk rather than a solved problem.

5. What the store page actually says

The September submission was a draft; what matters is what the page ended up showing. On 2026-10-19, the store page read:

Three things differ from what we expected, recorded as they are:

1. The category landed as Simulation. In the submission form we chose Games → Adventure as advised; the page shows Simulation. It does not hurt discoverability enough to be worth changing, but it shows that “what you submitted” and “what the page displays” are two different things — check the page.
2. The rating came back ESRB TEEN with a Violence descriptor. We had prepared the questionnaire for 12+. There is no graphic violence in the game, but there is a narrative description involving a tenant's disappearance; the rating decision is not ours to interpret. Record it, and do not describe the game as “all ages” on the site.
3. The screenshot area is age-gated for signed-out visitors. The page shows “Sign in to view this content / We need to verify your age to ensure this content is age-appropriate.” The nine real screenshots we uploaded are therefore invisible to a signed-out web visitor. Changing that means changing the rating or store-side settings, not copy. Noted.

6. The “8 acts” in the listing versus the 6 acts players reach

Update, 2026-10-26: the main-site copy now uses the recomputed 6 acts; the store listing copy still needs the same change.

The Features section of the listing says: 10 simulated websites / Eight acts and four different endings / 21 pieces of evidence to find and connect. Recounted against the 2026-10-26 build: the website count is right (10); the ending count is right (4); the clue count is now 18, while the listing still says 21; the act count is 6, while the listing still says Eight acts.

Before 2026-10-26, the game still carried two unreachable epilogue acts and a player could only reach six. The rewrite trimmed it to six; chapter progression is still capped at act 5 — once you have collected 10 clues, the chapter counter stops at 5 and the later act choices stay greyed out. The playable experience now says six acts; the store listing still says Eight acts.

The main site's introduction uses the recomputed “6 acts” — the game only reaches the finale once you are in the sixth act, so if that boundary moved to five or seven acts this page and the main site's introduction would both have to be re-derived together. The store listing has not been changed, because that means resubmitting store information (not repackaging; the manifest did not move). Either open progression or change the copy to six acts. The decision is on the todo list; it is not fixed and I am not pretending the two numbers agree.

7. An IARC rating is a commitment, not a label

The listing required the IARC content rating questionnaire. Room 404 uses the IARC content rating, dated 2026-09-18, registered under the product name Room 404.

Two constraints worth stating plainly:

• An IARC rating may only be used on digital stores that participate in IARC (Microsoft does). It is not a general-purpose age label — you cannot paste it on physical product or on a store outside the programme.
• If the content changes enough to alter the rating — new downloadable chapters that shift what is in the game, for example — the questionnaire has to be redone. For a narrative game that is still shipping chapters, budget for that. It is not a one-time checkbox.

8. Ads: say exactly what is running

In September I wrote “the web version runs AdSense”. Today that sentence needs to be precise, because it affects both the store questionnaire and the site's own statements:

• the page does load the AdSense auto-ads script;
• but the explicit ad-slot component is switched off — it hard-codes no serving, because the account has not passed review and leaving it on would only produce an empty frame;
• the account is currently not approved, and in testing there is no ad fill.

In other words: a player opening Room 404 today sees no ads running. But the store questionnaire asks whether the app contains third-party advertising, and the script and the intent are present, so we answered yes truthfully rather than changing the page to make the form look better. When the account is approved and ads actually serve, the store's privacy and advertising declarations have to be updated once.

9. What you actually get after installing

• searchable in Start and pinnable to the taskbar;
• opens with no network (except the artwork, see section 4);
• with one Microsoft account, installable on up to ten Windows devices;
• progress stays on-device, with the same localStorage semantics as the web version — no cloud saves, so it does not move with you to another machine.

That last point is worth stating up front, because “it's an installed app now” raises an expectation of account sync. There is none, and there is no plan to trade retention for it.

The badge uses the two official SVGs generated by Microsoft (one per language, one light and one dark), not a redrawn image. It sits immediately after the primary button because the portrait card cannot fit two screens; placing it after the promotional copy would mean nobody sees it.

10. This path has already been reused twice

After Room 404, MrQiWinWinWin followed the same PWA → MSIX route on 2026-10-08, and Dajin Weight-Loss GoGoGo on 2026-10-10. The lesson from reuse: packaging is not the cost; assets and questionnaires are. For each game you still need the 1920×1080 real screenshots, the IARC questionnaire, and a separately checked manifest. None of those can be skipped; the PWABuilder step is the same set of clicks.

The other original ARGs (Midnight Frequency, Tide Sound Archives, Sunken City, Rust River, Eye of the Abyss, Abyss Messenger, Last Online, Unclaimed) share the same manifest and service-worker structure and can queue along the same path. This has been verified, not guessed.

To play in the browser instead, the address has not changed: https://room404.gamestrategists.com/

Change log

• 2026-10-19: added three screenshots taken that day (store header, store detail page, title screen with the official badge). Replaced September's offline figures with the re-measured 19 entries / about 1,574 KB and documented the method; corrected the offline-cache description (cache scope widened to fonts and manifest-type assets) and the two runtime image-cache routes. Added section 5 (the store page's actual category Simulation, ESRB TEEN + Violence, zero reviews, age-gated screenshots), section 6 (the listing's “8 acts” versus the 6 a player reaches; the main site uses the recomputed number while the listing is pending), and rewrote section 8 from “runs AdSense” to: script present, explicit slot off, account not approved, questionnaire still declares ads. Section 10 updated from “the other six will follow” to “two games have already reused it”.
• 2026-09-24: first published.