FreshRSS vs Miniflux

Both of these read RSS well. Neither is going to disappoint you on the core job, and anyone who tells you one has a materially better reading experience is describing a preference. The decision underneath is about temperament, and it has an operational bill attached.

Miniflux is one static Go binary talking to PostgreSQL. The project publishes a page titled "Opinionated" whose whole purpose is explaining which features will be refused and why. There is no plugin API, no theme marketplace, no second database engine. FreshRSS is a PHP application that defaults to a SQLite file in a single container, ships an extensions directory, per-feed filter rules, XPath scraping and a settings surface deep enough that you will lose an evening to it at least once.

The counterintuitive part: the minimalist one is the heavier thing to operate. Miniflux scores 3 on our ops load scale and FreshRSS scores 2, because "PostgreSQL only, forever" turns a one-container app into two containers, a dump-and-restore procedure and a database major-version upgrade path you eventually have to walk.

What the spec table is actually deciding#

Read three rows and ignore the rest. Datastore is the fork in the road: SQLite-by-default versus Postgres-only is the difference between a file copy and a documented restore order. Container count follows from it. Idle RAM is the row people over-weight, because Miniflux's 40 MB looks decisive until you add the Postgres process sitting beside it, at which point the two land within about 50 MB of each other on a small instance. For context, the median idle footprint across every tool in this catalog is 150 MB, so both of these are at the light end regardless.

Which one works with your existing reader app#

This is what actually decides it for most people, and the honest answer is that it usually does not decide anything, because both sides serve the Google Reader API and the Fever API.

FreshRSS exposes both at https://your.host/api/, gated behind an "Allow API access" toggle in the admin authentication settings plus a per-user API password set in the profile screen, separate from the login password. Its own documentation calls Google Reader "the best option" and describes Fever as limited and less efficient, which is fair.

Miniflux enables each API under Settings then Integrations, where you invent a username and password per protocol. Its Google Reader page names Capy Reader, NetNewsWire, Reeder Classic 5 and later, and RSS Guard as working clients. Its Fever page is blunt about the gaps: JSON only, no links, sparks or kindlings, and Reeder cannot trigger a refresh because it sends no user data.

The uncomfortable line the marketing never includes: Miniflux's own docs say it implements only a subset of the Google Reader API. FreshRSS's implementation is the more complete one, and it is the reason niche clients that break against Miniflux tend to work against FreshRSS. If you already own a paid reader app with an unusual sync implementation, install FreshRSS first and test, because switching servers later means re-syncing every device.

Can either one pull the full article text#

Yes, and the usual claim that this is FreshRSS's exclusive advantage is wrong.

Miniflux has scraper rules in per-feed properties: give it a CSS selector like div#articleBody or article, div.article and it refetches the page and extracts that node. It also has rewrite rules for the practical annoyances, adding image titles, converting Markdown, fixing images that only exist after JavaScript runs, and regex block and keep rules applied globally then per feed.

FreshRSS does the same per-feed selector trick, then goes somewhere Miniflux will not: it manufactures feeds from sites that publish none. The subscription screen lets you pick an HTML plus XPath 1.0 feed type, or a JSON feed, or JSON with dotted paths including string concatenation like meta.title & " some text". You point it at a page, give it XPath expressions for the item container, title, link and body, and you have a feed for a site whose owner never made one.

So the split is:

• Feed exists but truncates the body: either tool, roughly equal effort.
• No feed exists at all: FreshRSS, or you go build a separate scraper service.

What the extension system buys, and what it costs#

FreshRSS extensions are PHP loaded into the running application. That gets you things Miniflux has permanently declined: alternative themes, sharing integrations, per-feed content transformations, and as of 1.29.0 in May 2026 a bundled LLM classification extension for auto-tagging incoming articles.

The cost is stated plainly in our profile and worth repeating: extensions are the usual thing that breaks after a minor upgrade. They hook internals that move. The 1.29.0 release alone changed extension name and version parsing, added a CSP-amending function and added a new entry-state property. A broken extension can take the whole page down rather than degrading quietly, so after every bump you load the extensions admin page before you assume the upgrade succeeded.

Miniflux's answer to the same needs is to hand off. One keystroke pushes an article to wallabag, linkding, Pinboard or Shaarli, and there are roughly thirty such integrations. Nothing runs in-process, so nothing in-process can break.

What 400 feeds actually costs#

Neither of these is heavy. FreshRSS's README benchmark is a Raspberry Pi 1 serving 150 feeds and 22,000 articles with sub-second responses, which is a low floor by any standard. Miniflux's process barely registers between polls.

The number that bites at a few hundred feeds is not memory, it is Miniflux's polling arithmetic. The scheduler wakes every POLLING_FREQUENCY (60 minutes by default) and hands BATCH_SIZE feeds (100 by default) to the fetchers. With 400 feeds and stock settings, each feed is checked roughly every four hours no matter what per-feed interval you set in the UI. People blame the feeds. The fix is arithmetic:

environment:
  LISTEN_ADDR: 0.0.0.0:8080     # default is 127.0.0.1, unreachable in Docker
  RUN_MIGRATIONS: 1
  POLLING_FREQUENCY: 30
  BATCH_SIZE: 200

FreshRSS's equivalent trap is the opposite shape. Its refresh cron fetches and parses every due feed in one pass, so 500 feeds with scraping extensions enabled means a visibly busy few minutes each cycle, and on SQLite with unbounded retention it gets slow before it gets alarming. Set a global and per-feed article retention policy on day one, then run the purge once from the admin page to see what it reclaims.

Backup and restore, which is the real cost difference#

FreshRSS on SQLite: back up data/ and, if you use them, extensions/. Stop the container or use sqlite3 .backup rather than copying a live file, because it writes on every refresh cycle. That is a file copy, and it is why the ops score is a 2.

Miniflux: one pg_dump, and nothing else, because the application keeps no local state at all. Simple, but it is a database procedure with a restore order (database up, restore dump, then start the app with RUN_MIGRATIONS=1), plus an eventual Postgres major-version upgrade that is a dump and reload rather than a container swap. If you are new to that, read Backing up a running database before you commit. Either way, export OPML on a schedule as the format-independent fallback, and test the restore per Backups that actually restore.

The verdict, by situation#

Fence-sitting answer avoided: if you have no strong opinion and no existing Postgres, install FreshRSS. It is the lower-attention deployment despite being the more configurable product, and you can ignore every setting you do not want.

What to do next#

Whichever you pick, put it behind TLS before you enable the API, because both authenticate over plain HTTP quite happily. Reverse proxy and TLS covers the Caddy config, and if you plan to add OIDC on top, both support it natively; Single sign-on for self-hosters explains what that is worth for a single user. If you are moving off a hosted service, Replace Feedly covers the export and the DNS-level details. To see how a reader fits alongside the rest of your stack, run the numbers in Stack planner or browse the rest of Feeds.