scrape-airing-anime

The one stop shop for currently airing anime. Ask three of the biggest anime databases what is on the air right now, get one normalised table back.

$ scrape-airing-anime --dedupe --limit 5
#  title                                       score  eps  type  season       source
-  ------------------------------------------  -----  ---  ----  -----------  ------
1  Sousou no Frieren 2                         9.12   24   TV    summer 2026  mal
2  Mushoku Tensei: Isekai Ittara Honki Dasu 3  8.42   14   TV    summer 2026  kitsu
3  One Piece                                   8.40   ?    TV    fall 1999    kitsu
...

Sources

No API keys, no HTML scraping, no headless browsers. Scores are normalised to one 0-10 scale across all three. If one source is having a bad day (AniList once switched its entire public API off, mid-development of this very tool), you get a warning on stderr and results from the others: the pipeline never dies with the fleet.

Install

pip install scrape-airing-anime      # or: uv tool install scrape-airing-anime

Or run it without installing: uvx scrape-airing-anime / pipx run scrape-airing-anime.

No dependencies. Python 3.9+ and the standard library, that is the whole stack.

Use

scrape-airing-anime                              # table of everything, all sources
scrape-airing-anime --dedupe                     # collapse duplicates across sources
scrape-airing-anime -s mal -n 25                 # top of MAL's current season
scrape-airing-anime --min-score 8 -f csv -o top.csv
scrape-airing-anime -f json -o ~/Desktop         # airing-anime.json on the Desktop
scrape-airing-anime -f jsonl -o - | jq .title    # stdout carries only data
scrape-airing-anime -f csv -o - | duckdb -c \
  "SELECT source, count(*) FROM read_csv('/dev/stdin') GROUP BY 1"

Every diagnostic (progress, warnings, the final summary) goes to stderr, so stdout is always clean CSV/JSON/JSONL you can pipe.

Full options: --source {mal,anilist,kitsu,all}, --format {table,csv,json,jsonl}, --limit N (per source), --min-score X, --sort {score,popularity,title}, --dedupe, --timeout SECONDS, --version.

Library

from scrape_airing_anime import REGISTRY

shows = REGISTRY["kitsu"](limit=10)
for show in shows:
    print(f"{show.score or '-':>5}  {show.title}  ({show.url})")

Every source returns list[AiringAnime], a frozen dataclass with title, source, url, media_type, episodes, score, popularity, season, year, genres, studios and synopsis.

Behaviour worth knowing

• Politeness built in: a descriptive User-Agent, retries with exponential backoff, Retry-After respected, and a pause between Jikan pages (they ask for at most 1 request/second).
• popularity is source-local (MAL members vs AniList popularity vs Kitsu user count). Compare it within one source, not across them.
• --dedupe keeps the first hit in source order (mal, anilist, kitsu), matching on accent-and-punctuation-insensitive titles.
• Kitsu does not ship genres/studios on its base resource, so those two columns are empty for that source.

Test

python3 -m unittest discover tests

36 tests, all offline: every HTTP interaction is faked, so the suite runs in milliseconds and does not hammer anyone's API.

Heritage

This repo began life in 2021 as scrape-myanimelist-airing: a requests-html scraper of MAL's top-airing page and a pandas notebook. The HTML scraping is gone (that is what always breaks), the ambition grew, the name followed.

License

MIT.