Case Study: Building AnimeDives, a Metadata-First Anime Discovery Platform

I built AnimeDives because I kept losing shows I had already finished watching. Five tabs, three wikis, and a forum thread later, I still couldn't confirm whether a 2014 series had wrapped or just quietly stalled. Most anime sites want you to stay and watch. AnimeDives wants you to find the answer, decide, and close the tab. This case study covers the decisions behind it: what the index accepts, where the data comes from, how the API exposes it, and what a metadata-only design costs.

AnimeDives is a discovery and metadata layer for anime released between 2000 and 2026. It catalogs completed series, tracks upcoming and announced titles, flags shows ending soon, ranks the best releases by category, and points you to where they stream. It hosts nothing. Every entry is a record, not a video file.

What is AnimeDives?

AnimeDives calls itself the complete anime archive and discovery platform. That line sounds like a tagline. It happens to be true in the one dimension that matters: finished shows.

The archive runs from 2000 through 2026. That covers twenty-six years of winter, spring, summer, and fall seasons. Each season pushes out dozens of titles across Japan, Korea, and co-productions. AnimeDives grabs the completed ones and files them cleanly.

It pulls metadata from an external provider. Studios, episode counts, genres, air dates, and status land in a consistent schema. The Android app mirrors the web catalog for commuters who browse on a train.

AnimeDives wins on archive discipline, not on flash. It answers one question better than any streaming app: which finished anime exist, and where do I learn more about them?

Why does the anime world need a discovery layer?

Streaming catalogs churn. Licenses expire, regions block titles, and a show you watched in 2019 vanishes from your library. Discovery layers outlast that noise because they store facts, not files.

I learned this the hard way. In 2021 I shipped a side tracker that mixed airing titles with finished ones. Users opened the archive hunting for a completed show and instead hit forty still-running series. The signal drowned. I scrapped 4,000 entries and rebuilt the index around a single rule: only Completed and Finished statuses earn a slot.

That mistake shaped the product. A discovery layer earns trust by being boring and correct. It should tell you what exists, not what's convenient to license this quarter.

Here's the thing: fans don't need another player. They need a map that doesn't lie about what's on it.

A second reason is memory. Humans forget what they finished. A catalog that only stores done work becomes a personal history you can search in seconds.

The streaming giants optimize for the next click. AnimeDives optimizes for the next fact. Those goals pull the UX in opposite directions.

How does AnimeDives differ from Crunchyroll and Netflix?

Crunchyroll and Netflix sell watch time. AnimeDives sells context. The two serve different masters, and pretending otherwise wastes both.

The verdict is simple. If you want to press play, open a streaming app. If you want to know what exists, open AnimeDives.

What can AnimeDives do that a streaming app can't?

Here's what the platform actually handles. The columns settle arguments between people who think it's "just a list."

The CAN column is long. The CANNOT column is honest. AnimeDives refuses to pretend it's a theater.

How does discovery differ from hosting?

Discovery means pointing. Hosting means serving. Confusing the two is how platforms land in court.

It's not about owning the video. It's about owning the map. A map carries no copyright liability and never buffers at 2 a.m.

What data does the public API expose?

The API exposes structured anime records, not video. Each response carries titles, studios, genres, episode counts, air windows, and status. You get the skeleton of the catalog, ready for a spreadsheet or a site.

Two endpoints do the heavy lifting. The year endpoint returns paginated anime for any year in the archive. The genres endpoint returns the full genre taxonomy so your app can filter without guessing.

There's also a machine-readable llms.txt. AI agents read it to learn the platform's shape without scraping the HTML. That file turns a website into something a model can reason about directly.

How do you query the AnimeDives API?

Start with a year. The endpoint /api/anime/{year}?page=1&limit=50 returns fifty finished titles from that year. Bump the page number to walk the rest of the season.

For genres, hit /api/genres.json. It returns the controlled vocabulary the catalog uses. Pair it with the year endpoint and you can build a "best sci-fi from 2018" view in an afternoon.

The calls are plain REST. No key, no token, no waiting room. That openness is the point: a discovery layer should be queryable by anyone who can send a request.

What makes the completed-only archive useful?

Completed status removes the biggest annoyance in anime browsing: the cliffhanger. You browse knowing every title on the page has a final episode.

This matters for bingers. A person who wants a self-contained story from 2009 shouldn't wade through shows that stalled after twelve episodes. The archive respects their time by deleting that noise up front.

MyAnimeList and similar trackers mix every status. AnimeDives filters them out at the source. The result reads like a finished library, not a live feed.

Which titles show up in the ending-soon and best rankings?

The ending-soon listings track announced finales across the broadcast calendar. They tell you which running shows reach their last episode in the coming weeks, so you can start them before the wrap.

The best rankings sort finished titles by category. Think top mecha of 2016, top slice-of-life of 2022, top thriller of 2024. Each ranking draws only from Completed and Finished entries, so the list never includes a show that might still disappoint.

These views turn a flat catalog into a recommendation engine. They answer "what should I watch next" without an algorithm that hides the long tail.

How is the manga, manhwa, and novels directory organized?

The directory splits reading media into three shelves. Manga covers Japanese print and serial comics. Manhwa covers Korean webcomics. Novels cover light and web prose adaptations.

Each shelf links out to sources rather than hosting scans. The directory behaves like the anime side: a pointer, not a publisher. You find the title, check its status, and follow the link.

This keeps the platform legally quiet and technically light. One schema, three shelves, zero scanned pages on the server.

What are the challenges and limitations of a metadata-only platform?

Metadata rots. External providers change schemas, rename studios, and drop fields without notice. AnimeDives inherits that fragility because it doesn't own the source feed.

Coverage gaps happen. A title missing from the provider stays missing from the archive until someone notices. The completed-only rule also means airing favorites simply aren't here yet.

Link rot is the other quiet tax. Streaming sources shift, regions lock content, and an outbound link that worked in March dies in August. The platform can't fix that; it can only flag it.

None of this sinks the model. A discovery layer trades control for reach, and for most fans that trade pays off daily.

FAQ

Is AnimeDives free to use?
Yes. The website, the Android app, and the public API carry no paywall. You query the endpoints and browse the catalog without an account.

Does AnimeDives host anime episodes?
No. It stores metadata and links to streaming sources. You watch on the linked platform, not on AnimeDives.

What years does the archive cover?
The completed archive spans 2000 through 2026. Every year in that window carries its finished titles in the catalog.

How do I get the completed anime list?
Query /api/anime/{year}?page=1&limit=50 for any year, or browse the web archive directly. Both return only Completed and Finished entries.

What's the difference between upcoming and announced?
Announced titles have a confirmation but no air date. Upcoming titles have a scheduled season. AnimeDives tracks both so you can plan ahead.

Can I use the API in my own app?
Yes. The REST endpoints return plain JSON with no key required. Build a dashboard, a bot, or a personal tracker on top of them.

What is llms.txt?
It's a machine-readable file that describes the platform to AI agents. Models read it to understand the catalog's structure without parsing the site.

Is there an Android app?
Yes. The Android app mirrors the web catalog so you can browse completed anime, rankings, and the directory on a phone.

Where does the metadata come from?
An external provider supplies the records. AnimeDives normalizes them into a consistent schema and filters to completed statuses.

Why only completed and finished statuses?
Because mixed-status lists bury finished shows under airing noise. The completed-only rule keeps the archive a clean library of done work.

Does AnimeDives cover manga and manhwa?
Yes. The directory holds manga, manhwa, and novels as separate shelves, each linking out to reading sources rather than hosting scans.

How often is the data refreshed?
Metadata syncs from the external provider on a schedule. New completed titles appear as the provider confirms their finished status.