Once you’ve actually found a site’s feed — the URL patterns and autodiscovery tricks for that are their own piece — you still need something to read it in. Browsers used to fill that gap; they don’t anymore. Firefox dropped feed preview and Live Bookmarks in Firefox 64 (December 2018), and no mainstream browser has shipped a general reader since. The reader is now a deliberate choice of separate software, and the choice is bigger than it looks: it’s about who fetches your feeds, who can see the list of things you follow, and who’s on the hook when something breaks.
There are exactly three honest answers to that, and they trade against each other in predictable ways: hosted, where a company runs the infrastructure and you log in; self-hosted, where you run the infrastructure and nobody else is involved; and local, where an app on your device does the fetching and nothing syncs unless you explicitly wire it up. Below is what each one actually costs you, with real examples and pricing checked today — August 27, 2026 — because reader pricing is exactly the kind of thing that goes stale between when an article is written and when someone reads it.
The Real Question Isn’t Which App — It’s Who Holds Your List
Every reader answers three questions:
- Who fetches the feed? A server somewhere polls each URL you subscribe to on a schedule. That server is either a company’s infrastructure, your own infrastructure, or your own device.
- Who can see your complete subscription list? Not one feed — the entire list, which is a more revealing artifact than it sounds: a person’s subscriptions describe their interests more precisely than most social-media follow lists.
- Who is responsible when something breaks? A dead feed, a version upgrade, a sync conflict, a server outage.
A hosted reader answers all three the same way: the vendor. A self-hosted reader answers all three the same way too: you. A local app splits the answer — your device does the fetching, but the moment you turn on sync, a chunk of “who can see the list” reopens, because now some sync provider is in the loop.
None of these is correct in the abstract — it’s a real trade of convenience against control, weighed against how much you mind being your own systems administrator versus how much you mind a company holding a list of your interests.
Hosted: Someone Else Runs It, and Sees Everything You Subscribe To
A hosted reader is the lowest-effort option by a wide margin: create an account, add feeds through a web or mobile app, and the vendor’s servers handle polling, storage, and cross-device sync. Nothing to install, nothing to patch, nothing that goes down because you forgot to renew a certificate.
Inoreader is a clean example of the category. Checked directly against its pricing page today (2026-08-27): the free tier covers 150 RSS feeds, 20 “web feeds” (its term for turning ordinary pages into monitored feeds), 20 newsletter feeds, and a modest set of automation rules and filters. The paid Pro tier is $7.50/month billed annually, or $9.99 billed month-to-month, and raises the feed ceiling to 2,500, removes the cap on web feeds, and adds things like guaranteed hourly refresh, full-content loading, and API access.
NewsBlur is the more interesting hosted option because it’s a hybrid — MIT-licensed and self-hostable via Docker, while also running as a straightforward paid service for people who don’t want to bother. Verified today: Free covers up to 64 followed sites; Premium is $36/year and raises that to 1,024 sites with faster refresh and folder-level reading; Premium Archive is $99/year and adds full story archiving and search across everything you’ve ever subscribed to; Premium Pro is $29/month for people who want feeds checked every 5–15 minutes and a much higher site ceiling. NewsBlur’s signature feature is “training” — teaching the reader to hide or highlight stories by author, tag, title, or full text.
Feedly is the obvious name missing from that list. Its pricing page failed to load on two separate attempts today, so no Feedly number appears here — check it yourself before signing up, and treat any older number you find quoted elsewhere as unverified.
The trade-off is symmetrical and worth saying plainly: a hosted reader is the easiest way to read feeds, and it also means one company has a standing, continuously updated list of every source you follow. That’s not a hypothetical risk — it’s the deal you’re explicitly making the day you create the account.
Self-Hosted: You Run It, You Own It, You’re the Ops Team Now
Self-hosting flips every answer above: you install the software on a server you control, your own IP polls every publisher directly, and the complete list of what you follow lives in a database only you can see. Nobody else’s outage takes your reader down, and nobody else can change the pricing on you.
FreshRSS, checked today via its own GitHub releases feed, is currently at version 1.29.1, released 2026-05-20, distributed under the AGPL-3 license. Its own site (checked today) lists OPML import/export, WebSub support for near-real-time updates instead of pure polling, saved-query search, and the ability to scrape a page into a generated feed for sites that don’t publish one.
Miniflux takes a different shape — a single statically compiled Go binary plus a PostgreSQL database, distributed under Apache-2.0, checked today. It parses the widest format spread of any reader named here (Atom 0.3/1.0, RSS 1.0/2.0, and JSON Feed 1.0/1.1), runs a Readability-style extractor to pull full article text out of summary-only feeds, and strips tracking pixels and utm_ parameters on the way in. If you don’t want to run your own server, the project’s own hosted option is $15/year — cheap enough that it barely counts as “hosted” in the vendor sense; it’s really self-hosting with the ops outsourced for the price of a couple of coffees.
The honest cost isn’t money — Miniflux’s hosted tier makes that obvious by being nearly free. The cost is that you are now the systems administrator: security patches, database backups, certificate renewal, the disk that fills up at 2 a.m., indefinitely, for as long as you want the reader to keep working. It also means your server’s IP is what knocks on every publisher’s door on a schedule — trading “one company sees my list” for “every publisher I follow can see something is polling them” — a different exposure, not a smaller one.
Local: Nothing Syncs Unless You Tell It To
The third option skips servers almost entirely. A local app on your Mac, iPhone, or iPad fetches feeds directly from your device and stores everything there.
NetNewsWire is the clearest example — free and open source, requiring macOS 15 and newer or iOS 26 and newer (checked today against the vendor site), currently at version 7.1.3 as its latest stable release, with a 7.1.4 beta already in testing as of today. Used on a single device with sync turned off, nothing about your subscription list ever leaves that device. No account, no vendor database, no company anywhere holding your list.
The catch is in “single device.” The moment you want read/unread state and subscriptions to follow you from Mac to phone, NetNewsWire needs a sync layer — supported options (checked today) are iCloud, Feedbin, Feedly, BazQux, Inoreader, NewsBlur, The Old Reader, and FreshRSS. Pick iCloud and Apple is in the loop; pick a hosted reader as the backend and you’ve quietly re-added everything the hosted section above describes, with a nicer front end on top. Local-only privacy and multi-device convenience are in direct tension — no app gives you both without someone, somewhere, holding a copy of your list.
Every One of These Still Ranks Something
Here’s the point a lot of RSS advocacy skips: switching to a feed reader doesn’t mean you’ve escaped ranking. It means trading a ranking system you didn’t choose for one you can, if you bother, actually shape.
Inoreader’s paid tier includes automation rules — conditions you write yourself that tag, star, mute, or forward stories by keyword or source. NewsBlur’s “training” does the same by example: hide stories from an author you’ve decided isn’t worth your time, boost a trusted source, and the reader applies that going forward. Even a folder of feeds sorted by priority is a ranking decision — just one you made by hand instead of one an engagement model made for you.
None of that undermines the case for feed readers — it sharpens it. The honest version was never “feeds have no algorithm”; that claim doesn’t survive contact with Inoreader’s rules engine or NewsBlur’s training. The real claim is narrower and holds up: the objection was never to ranking. It’s to ranking you cannot inspect, edit, or export. A rule you wrote yourself is legible — you can read it, change it, delete it, and take the underlying list with you in an OPML file to a different reader entirely. An engagement-optimized timeline gives you none of that: no visibility into why a post ranked where it did, no way to edit the model, no export button for what actually shaped what you saw. Every reader named above supports OPML export precisely so you’re never stuck — that portability is its own story, told elsewhere.
How to Actually Decide
Skip the temptation to rank these like a listicle — the honest answer is that the right architecture depends on what you’re actually optimizing for. A few questions do more work than a scorecard:
- Do you read on more than one device? If yes, pure local-only is off the table unless you’re willing to add a sync backend — at which point you’ve partly re-adopted the hosted model anyway.
- Does it bother you that a company could see your entire list of interests? If that’s a real concern rather than an abstract one, self-hosted or single-device local are the only architectures that avoid it completely.
- Do you actually want to run a server? Self-hosting is rewarding if you already maintain other infrastructure, and miserable if you just want to read feeds. Miniflux’s $15/year hosted tier is a reasonable middle path if the appeal is the software, not the ops work.
- How many feeds, realistically? A dozen sources has very different needs than a few hundred. Free tiers on both Inoreader and NewsBlur cap out fast; a heavy subscriber hits a paywall or a self-hosted server sooner than expected.
- Do you want the reader to rank things for you, or by hand? NewsBlur’s training and Inoreader’s rules are real features, not compromises — decide up front whether you want that assistance or a strictly chronological river.
There’s no wrong answer among the three, only a mismatch between the architecture and what you actually wanted from it.
Frequently Asked Questions
Is a hosted reader like Inoreader or NewsBlur private? No, not in the sense most people mean. The vendor’s servers do the fetching, so the vendor holds a standing, complete list of everything you follow — arguably more revealing than a social-media follow list, because it’s concentrated in one company and rarely audited by the person it describes.
Is self-hosting actually more private, or does it just move the exposure? It moves it. Instead of one vendor seeing your whole list, every publisher you follow sees requests from your own server’s IP, on a predictable schedule, indefinitely — a different trade, not a strictly better one.
Can I switch readers later without losing my subscriptions? Yes — this is one of the format’s genuine strengths. FreshRSS, Miniflux, NewsBlur, and NetNewsWire all support OPML import and export (checked against each vendor today), which is a plain file listing your subscriptions and folder structure. Migrating readers is a file download and a file upload, not a data-export fight with support.
What does a reader’s “training” or “rules” feature actually do? It reintroduces ranking on your terms. NewsBlur’s training hides or highlights stories based on author, tag, title, or full text as you mark examples; Inoreader’s rules let you write conditions that tag, star, or mute stories automatically. Neither is “no algorithm” — both are an algorithm you wrote, can read, and can turn off.
Is NetNewsWire really free with no catch? Yes — it’s free and open source with no paid tier, though as of today it requires a fairly recent OS (macOS 15+ or iOS 26+ per the vendor’s current requirements). The only “catch” is that it’s Apple-platform only, and multi-device sync means picking a sync backend, which reopens the privacy question the app otherwise avoids.
Do I have to pick just one architecture forever? No. NetNewsWire alone demonstrates the point — it runs purely local on one device or syncs through a hosted or self-hosted backend, so the same app answers differently depending on configuration. Plenty of people run a local reader day-to-day and keep a self-hosted instance as an archive.
Recommended Resources
Start free. Every price and version claim in this article came straight from the vendors themselves — Inoreader’s pricing page, NewsBlur, Miniflux, FreshRSS, and NetNewsWire all publish their own terms, features, and (where relevant) source code. If you’re choosing between architectures, an hour reading those five pages directly will tell you more than any book on the subject — because the subject changes every few months and a book can’t.
If you’re leaning self-hosted, you’ll eventually be looking at raw feed files and wondering why one won’t parse. The RSS/Atom book shelf hasn’t been updated since the format itself stopped changing, but the mechanics in these titles are still accurate:
- RSS and Atom in Action — the most complete side-by-side treatment of both formats; useful once you’re running FreshRSS or Miniflux and want to understand what they’re actually parsing.
The more durable question this article raises isn’t about feed formats at all — it’s about who controls the ranking you see and whether you can inspect it. That argument has a real shelf, unlike the RSS-specific one:
- The Filter Bubble — Eli Pariser’s original case against algorithmic curation deciding what you see without your input. Worth reading, but treat its central thesis as contested rather than settled; later empirical work pushed back on how strong the “bubble” effect actually is.
- Breaking the Social Media Prism — the empirical counterweight to Pariser’s argument, and the fairer way to read the two together rather than taking either one as the final word.
Read them as a pair, not a verdict — the honest position on ranking is closer to “make it inspectable” than to a flat “algorithms bad,” and these two books argue opposite sides of how bad the un-inspectable version actually is.