
Pick three things you want to keep track of: a Bilibili creator who uploads twice a month, a Weibo account that announces restocks, and a software release page that only changes when there's a new version. None of them publish an RSS feed. Check each site manually and you'll miss updates; rely on the recommendation algorithm and you'll see them whenever the platform feels like showing you. RSSHub is the open-source project that closes this gap — it generates RSS feeds for websites that never bothered to ship one. Created by developer DIYgod and released under the AGPL-3.0 license, it has collected more than 46,000 GitHub stars since 2018, and its tagline states the whole mission in four words: "Everything is RSSible."
This piece grew out of a scouting run where our usual forum source was unreachable, so there's no single thread to credit — instead it's a decoder on a tool the Chinese developer community has kept alive and central for years. Everything factual below is checked against the project's own repository and docs; community color is marked as such.
RSS is the oldest "follow anything" protocol on the web: a site publishes a machine-readable list of its updates, your reader polls it, and you get a chronological feed with no algorithm in between. Google Reader made it mainstream — Google launched it in 2005, then retired it on July 1, 2013 as part of its "spring cleaning," declaring usage had declined. The obituaries for RSS were written that summer.
But the protocol never died; the feeds did. The modern web's most-followed destinations simply stopped shipping them. Social platforms, video sites, and app-era services keep updates behind logins and infinite scroll, which means the original RSS dream — one reader for everything you follow — broke exactly where it mattered most. That break is the entire reason RSSHub exists: instead of waiting for sites to publish feeds, it builds the feeds itself.
RSSHub's core abstraction is the route: a URL path on your RSSHub instance that maps to a rule for extracting updates from a specific site. You run an instance, you append a route, you get a feed. The project's repository lists the services its routes cover, and the list reads like a map of the Chinese internet next to the global one: Bilibili, Weibo, Zhihu, Douban, WeChat, Ximalaya, Lofter, V2EX alongside TikTok, Instagram, Twitter, YouTube, Spotify, Telegram, Pixiv, and Dribbble.
Writing a route is ordinary open-source plumbing — fetch the page or API, normalize the items, emit RSS — but the project's README is explicit that this is a community effort: "our vibrant open source community is ensuring the deliver of RSSHub's new routes, new features and bug fixes." With 18,000+ commits and 10,000+ forks, the maintenance burden of a thousand site-specific scrapers is distributed across the people who actually use them. When a site redesigns and a route breaks, someone who depends on that feed usually fixes it. That feedback loop is the real product.
For discovering routes without reading docs, the ecosystem ships RSSHub Radar, a browser extension that detects which feeds — native RSS or RSSHub routes — are available on the page you're currently visiting. One click subscribes. There are mobile companions too: RSSBud for iOS and RSSAid for Android.
RSSHub's own README describes it as "the world's largest RSS network, consisting of over 5,000 global instances" — that's the project's claim, not an independent census, but it tells you how the project wants to be used: self-hosted, everywhere. The official deployment docs walk through Docker, Docker Compose, Kubernetes, and manual Node.js installs. The Compose route is recommended; the plain Docker route is one line:
docker run -d --name rsshub -p 1200:1200 diygod/rsshub
Images are published as diygod/rsshub on Docker Hub and ghcr.io/diygod/rsshub on GitHub's registry. Two tags matter: latest for the standard build, and chromium-bundled for sites that need a real headless browser to render — the docs note that the Chromium variant supports a wider range of routes at the cost of more resources. Configuration is by environment variable (e.g. -e CACHE_EXPIRE=3600), and the instance serves everything on port 1200 by default. It's the same one-container discipline as the self-hosted file tools in this scene — deploy it next to something like OpenList and it slots right into a homelab.
A word on etiquette, straight from the project's posture: feeds are cached, and heavy polling of someone else's public instance is poor citizenship. Run your own instance if you're subscribing to a lot — on the kind of cheap VPS Chinese homelabbers favor, a container this light barely registers.
A feed generator is only half the stack; you still need somewhere to read. The Chinese self-hosting community's two default answers are both open source and both worth knowing:
| Reader | Stack | License | Character |
|---|---|---|---|
| FreshRSS | PHP, multi-user | AGPL-3.0 | Full-featured: themes, extensions, OPML import/export, a Google-Reader-like API for mobile apps, built-in website scraping for sites without feeds |
| Miniflux | Go + PostgreSQL, single binary | Apache 2.0 | Minimalist and opinionated: deliberately few features, fast, keyboard-driven |
FreshRSS's own docs sell it plainly: "an RSS aggregator and reader" that lets you follow sites "without the need to browse from one website to another," with multi-user support so one instance can serve friends and family. Miniflux's pitch is the opposite virtue — a single statically-compiled binary, PostgreSQL only, no frameworks, features voluntarily limited. Pick FreshRSS if you want a platform; pick Miniflux if you want an appliance.
There's also a newer third option from the same orbit: Folo, an AI-assisted RSS reader that the RSSHub README now recommends as a pairing ("pairs especially well with RSSHub, an AI RSS reader for feed discovery and modern reading workflows") and whose source is open on GitHub. The AI-reader wave — summarization, feed discovery — is where the maintainers clearly think reading goes next.
Look again at that services list: Bilibili, Weibo, Zhihu, Douban, WeChat, Ximalaya, Lofter, V2EX. RSSHub didn't become essential in China by accident — it became essential because China's biggest platforms are precisely the ones that never shipped open feeds, and the developer community routed around the damage. In an information environment built on super-apps and algorithmic timelines, a tool that turns walled gardens back into plain chronological lists is quietly subversive. It doesn't fight the platforms; it just refuses to read on their terms.
That's the pattern worth noticing. The Chinese self-hosting scene keeps producing the same shape of project: take a dependency on a platform you don't control, replace it with a small server you do. OpenList did it for cloud drives; RSSHub did it for the feed. The community doesn't wait for openness — it builds the adapter.
Routes are scrapers, and scrapers rot. Every site redesign is a potential outage for the routes that depend on it, which is why the project's survival depends on that contributor flywheel rather than any technical trick. The Chromium-bundled image exists because some sites only render in a real browser — heavier, slower, but sometimes the only way in.
There's also the question nobody in the docs fully answers: many of these sites' terms of service restrict automated access. RSSHub lives in the same gray zone as every scraper, ad-blocker, and read-it-later tool — widely tolerated for personal use, never officially blessed. If you're generating feeds from someone's API at scale and reselling the result, you're not the user this was built for.
RSSHub will eventually be one entry in a longer history of feed protocols, but the instinct behind it is evergreen: own your information diet. An RSS reader with feeds you chose, polled on your schedule, served from a container you control, is the closest thing the modern web has to the old promise of "follow anything." The algorithm shows you what keeps you scrolling; a feed shows you what you asked for. That distinction is the entire pitch — and it's why, thirteen years after the obituaries, a Chinese open-source project with 46,000 stars is still minting new feeds every day.