A while ago, I had the plot of a manga stuck in my head, along with a few specific details about the main character and how the story started but I completely blanked on the title. I spent three straight days digging through MAL, AniList, Reddit threads, and tag databases trying to track it down. Every traditional database failed because they only work if you already know the exact title or the exact combination of rigid tags, and every general LLM I asked just hallucinated fake titles or mixed up three different series.
Once I finally found that manga, I decided to build a real solution for the problem.
So I built Anime Akashic Records ( https://aarecords.tech ) a semantic discovery engine and dense vector database designed to find anime, manga, light novels, and visual novels from pure natural language descriptions. When I finished building the engine and tested it with the exact vague description of that manga I spent three days looking for, AAR pulled it up as the very first result in under 300ms.
What AAR offers:
Hi Product Hunt!
Ever thought about an anime or manga whose title you completely forgot, even though one specific scene is still vivid in your head?
Introducing Anime Akashic Records (AAR)—a semantic discovery engine and dense vector database built to find exact works and scenes from pure natural language descriptions.
The whole idea started when I forgot the name of a manga and spent three straight days digging through forums and tag databases trying to find it. Traditional databases like MAL, AniList, and VNDB only work when you already know the exact title or rigid tags, and general LLMs constantly hallucinate titles and episode numbers. Once I finally found that manga, I decided to build a real fix for the problem—and when I finished building AAR, it found that exact manga on the very first search in under 300ms.
What powers AAR under the hood:
- 146,140+ indexed works for general discovery (69.8k+ Manga, 37k+ Anime, 20k+ Visual Novels, 11.7k+ Light Novels, and 7.5k+ Manhwa) with cross-platform ID mapping (MAL, AniList, TMDB, VNDB, AniDB).
- 102,264+ indexed anime episodes across 7,535+ series (broken down into 762,318+ timestamped scenes) and 10,565+ story arcs for deep scene-level search.
- High-precision, sub-300ms retrieval powered by dense vector embeddings and a two-stage neural reranking pipeline.
- A full Developer REST API, recommendation engine, and MCP server for AI agents and third-party apps.
To celebrate our Product Hunt launch, we are offering 3 exclusive first-month discount codes (redeemable in the dashboard or directly at checkout):
- AAR1ST — 50% OFF the Starter Plan ($4.99 instead of $9.99)
- PHPRO20 — 20% OFF the Pro Plan ($23.99 instead of $29.99)
- PHBIZ40 — 40% OFF the Business Plan ($35.99 instead of $59.99)
You can test the search right away on the web with the free daily sandbox (no signup required) at https://aarecords.tech, or grab a free sandbox API key at https://aarecords.tech/dashboard.
Would love to hear your feedback—try stumping the engine with the most obscure anime or manga scene you remember!