About Banana News
Our vision: archive every news story of all time, in real time, all topics, in perfect harmony - an encyclopedia of news.
What Banana News does
- Real-time archive. A background worker polls 1157 public RSS/Atom feeds (world, U.S., politics, business, markets, crypto, tech, AI, science, health, sports, entertainment, climate, Rochester & New York) every few minutes, de-duplicates by URL, and tags topics.
- Timeline. Browse by year, month and day like an encyclopedia.
- Same story, many sources. Headlines about the same event are clustered so you can compare coverage.
- Archived copies. Every story links to the original and to its Wayback Machine copy.
Honest limits
- It is not literally every story ever. No one has that. Banana News only knows what its feeds and historical sources expose; coverage grows over time and is uneven across topics, outlets and eras.
- Live coverage starts when we started. Older stories come from historical backfills: GDELT (a rolling window of recent months of global news URLs), and Wikipedia's Current Events portal and year pages (editor-written one-line summaries of major events, licensed CC BY-SA). Chronicling America, Common Crawl News and Internet Archive connectors are designed but not yet built.
- Headlines, snippets and links only. We do not store or republish full article text; articles belong to their publishers.
- Feeds break. Some outlets (e.g. AP, Reuters, Gannett papers such as the Democrat & Chronicle) do not offer open RSS access. Per-feed health is visible at /api/feeds.
- Clustering and topic tags are heuristic (title similarity and keywords) and will sometimes be wrong.
Where political placements come from
The Perspectives view and the coverage bar can place an outlet on a rough left-center-right scale so you can see who is covering a story. We do not use or copy ratings from AllSides, Ad Fontes Media, Media Bias/Fact Check or any other rating project. A placement appears only when an openly licensed fact states one: the outlet's own Wikidata entry (CC0) lists a political position, or the site owner has added a written editorial placement (labelled "Banana editorial"). Wikidata is community-edited and positions are relative to each outlet's own country, so they are not comparable across countries and some will be wrong or out of date. Placements are publication-level and say nothing about any individual article. Most outlets have none, and many well-known outlets are unplaced because no openly licensed source states a position for them. Sports, science, tech, music and trade outlets are never placed. A "Blindspot" means only: at least three placed outlets in our feeds cover a story, at least 75% of them sit on one side, and the other side has none - our feed list is finite, so absence here is not proof of absence elsewhere.
Why not estimate it? We tried inferring a lean from coverage patterns (which outlets cover the same stories). Checked against the outlets that do have a Wikidata position, it picked the right side for only 16 of 25 outlets (64%, not distinguishable from a coin flip) because shared coverage mostly reflects language, region and topic, not politics. So we do not publish estimated left/right labels; we show facts and our own counted statistics instead.
Source transparency chips
Story cards and source pages show facts about the outlet behind a story: type and country, and (from the Bunch plan) owner and reliability. Type, country and owner come from Wikidata (CC0) and our own curation. Reliability is the verdict of English Wikipedia's perennial sources list (CC BY-SA 4.0) - the consensus of Wikipedia editors about citing that outlet in Wikipedia. It covers only a few hundred outlets worldwide, so "not assessed" is normal, and it is not a verdict on any single article. The numbers on source pages (corroboration, how often an outlet is first, typical lag, who it most often covers the same stories as) are counted by us from the last 72 hours of our own archive. None of this is a judgement of truth. Turn the chips off with NEWSIUM_SOURCE_CHIPS=0.
Why it's trending
Why it's trending joins Wikipedia's most-read articles of the previous day (pageview numbers from Wikimedia Analytics, CC0) with the stories in our archive. It is "most read on Wikipedia", not "what the world is searching". Short lead text on topic pages is from Wikipedia under CC BY-SA 4.0, shortened and linked; the explanation sentence is built from our own coverage data and is not a proven cause. Our servers contact Wikimedia, your browser never does.
Today
143,360 stories from 2,508 sources, spanning 1728-11-02 to 2026-10-04. 17,505 ingested today (UTC).
Subscription
Plans start free; paid tiers are $0.55 (Bunch), $1.99 (Crate) and $4.99 (Banana Tree) per month. See pricing. Live TV is free for everyone. Banana News was created by Brooks Faria.
API
JSON endpoints under /api/ - interactive docs. The same free-preview limits apply to signed-out requests.
Crawler
Our fetcher identifies itself as BananaNewsBot, uses timeouts, conditional GETs and per-host spacing, and only reads publicly advertised feeds.