A web-based RSS/Atom feed reader you run yourself — for you alone or for several users. Free and open source (MIT).
More in the screenshot gallery.
Reading
- Three layouts: a magazine-style card grid, a compact list, and a two-pane view — with a draggable split — that shows the article next to the entry list.
- Reader view shows the extracted full article text, with images, audio, and video (including streaming video) played inline; you can switch to the feed's own version at any time, or open the original page. If the text is only the free preview of a paywalled article, the reader tells you.
- Read/unread tracking with "mark all read", favorites, and a separate "keep" list.
- Full-text search across your articles, including exact phrases in quotes. Save a search as a named view in the sidebar, each with its own unread count — open one, or open all as a combined list.
- Email digests: a daily or weekly summary of new articles from the saved searches you choose.
- Reading-time estimates, a progress bar, and an optional reading focus that dims everything but the passage in front of you.
Feeds
- Add a feed by its address — or paste the website's address and the app finds the feed for you.
- Preview a feed before you subscribe: recent items, whether entries carry images, and whether the feed delivers full text or only summaries.
- Import and export your subscriptions as OPML.
- Feed health is visible in the settings — active, erroring, or gone, with a plain-language explanation.
- An Organise page gathers every feed and tag in one place: reorder them by drag, select several at once to tag them or change their visibility in bulk, and retry or unsubscribe feeds that are erroring or gone.
- For sites without any feed, an opt-in experimental mode scrapes the article list into a pseudo-feed.
- When a WordPress site's feed carries only summaries, the app detects its REST API while finding the feed and offers it as a richer alternative — full article text, chosen in the same subscribe dialog.
Tags
- Group feeds with tags; each tag has a name, a color, and an icon.
- Assign tags in the feed dialog, or drag a feed onto a tag.
- Unread counts per feed and per tag.
"For you" — AI recommendations (optional)
- A ranked selection of your unread articles, based on what you favorite, keep, and read.
- You choose the provider: any OpenAI-compatible API works, including local models via LM Studio or Ollama — your data can stay on your own machine.
- A free-text guidance prompt steers the picks (topics to prefer or avoid); runs start manually or on a schedule.
- Runs execute on the server: close the tab and come back later, stop a run, or resume a failed one.
- A cost overview shows token usage and cost per run, spend per month, and the total spent — so there are no surprises with paid providers.
Accounts
- Sign in with email and password, with Google or Apple, or with a passkey.
- Email verification and admin approval for new accounts are optional — you decide per instance.
- Registration is protected by an invisible proof-of-work challenge; there is no CAPTCHA to solve.
- Users can export the whole account to a backup file and restore it later, or delete the account entirely, including all its data.
Administration
- Most instances are self-hosted, so you are the admin: approve, suspend, or delete users from the settings.
- Optional trial periods and per-user feed limits.
- Configure outgoing mail (SMTP) from the admin settings, with a built-in connection test.
- Route the server's outbound fetches through an HTTP proxy, set and tested from the admin settings.
- A feed catalog suggests feeds to new users; edit it or import the bundled one.
Interface
- Special effort went into mobile: a collapsible sidebar with swipe gestures, pull-to-refresh, and touch-friendly controls.
- Light and dark themes, following your system setting if you want.
- The interface is available in English and German.
Run your own instance with one command. You need Docker (running) and git. Then:
curl -fsSL https://raw.githubusercontent.com/larspohlmann/simple-feed-reader/main/scripts/install.sh | bashThe installer clones the project into ./simple-feed-reader, checks out the
latest release, generates the secrets it can, asks for the things only you
know (which package to install, how users reach the instance — plain HTTP,
your own certificate, or a reverse proxy — under which hostname and port, and
how to send mail), and starts the production stack. The full guide — TLS,
reverse proxies, mail verification, backups —
is docs/docker-production.md.
Read before you pipe to bash. You can inspect exactly what runs at scripts/install.sh. The installer never deletes data.
The first question the installer asks is which package to install. It decides the database and the search engine together, because the two of them are what the stack costs in memory:
| Package | What you get | Containers | RAM |
|---|---|---|---|
| S | a personal instance. SQLite, title and summary search. | php, worker, web | needs about 250 MB |
| M | several users. MySQL, title and summary search. | php, worker, web, mysql | needs about 1 GB |
| L | like M, plus Meilisearch for full-content search. | php, worker, web, mysql, meilisearch | needs about 2.5 GB |
Every package runs the same application, with every feature: they differ in the containers beside it, not in what the reader can do. Search works in all three. S and M answer it from the database, which matches titles and summaries; L adds a Meilisearch container that matches the full text of every article as well.
Two more keys decide how much you are asked, rather than what runs:
- Q — the quick install, and the default. It runs the S stack and asks
nothing else, which is what the question itself says under it:
It picks http://localhost:3333, and no mail until you add a relay.
Press return at the first question and the instance comes up. The public URL
and mail are changeable afterwards with
./scripts/prod-configure.sh; the database is not. - C — choose everything yourself, database and engine included. S, M and L ask for the public URL and for the mail transport anyway — an SMTP relay (host, port, user, password) or the MTA on this machine, which the app needs before it can send verification, password-reset and approval mail. Without one it sends none, which is the default and a complete answer for a private instance. C adds the database and the search-engine questions to those, which makes it the only way to reach a combination the three packages do not cover, such as SQLite with a search engine.
The figures are measured on an idle, healthy stack holding a real account of 107 feeds and 17,427 articles. S and M do not grow with the number of articles; L adds roughly 45 MB per 1,000 articles on top of its base.
Every path but Quick also asks a separate question about an optional Grafana
log dashboard — off by default, not part of any package, viewing only (no
tracing yet); ./scripts/prod-configure.sh can turn it on or off later.
Both installers take a target directory and --ref <branch-or-tag>, which
installs something other than the latest release — how a change is tried on a
test instance before it ships:
… | bash -s -- --ref feature/430-installer-output my-folder.
Every script in the table below, and both installers, print --help.
Once the stack is running, create the first administrator account — see docs/first-run-setup.md.
Run these from inside the simple-feed-reader directory:
| Task | Command |
|---|---|
| Update to the latest release (prod and/or dev) | ./scripts/update.sh |
| Update to a branch or tag instead | ./scripts/update.sh --ref <branch-or-tag> |
| Start / stop the production stack | ./scripts/prod-start.sh / ./scripts/prod-stop.sh |
| Change the public origin / mail settings | ./scripts/prod-configure.sh |
| Start / stop the dev frontend (:4200) | ./scripts/frontend-start.sh / ./scripts/frontend-stop.sh |
| Stop the dev stack (keeps your data) | docker compose down |
A multi-user RSS/Atom feed reader. Symfony 7.4 LTS JSON API in backend/, with
the Angular 20 SPA that delivers the reader UI and auth in frontend/.
For the development stack — live-reloading frontend, xdebug, and Mailpit catching all outgoing mail locally — use the dev installer instead (it additionally needs mkcert):
curl -fsSL https://raw.githubusercontent.com/larspohlmann/simple-feed-reader/main/scripts/install-dev.sh | bashThe manual walkthrough lives in docs/local-docker.md.
- Frontend workspace — the Angular 20 SPA: dev server, the quality gate, theming, and the bearer-token auth model.
- Architecture: client contract and native-client readiness — the cross-cutting rules for how clients talk to the backend, and the standing constraint that keeps a future native iOS app viable.
- OAuth sign-in (Google and Apple) — provider setup for operators, and the redirect/exchange contract for the SPA.
- Local Docker environment — run the whole stack (MySQL, PHP, nginx with TLS, Mailpit) in Docker.
- How a "For you" run works — what happens after "Get recommendations", closing the browser, stopping, resuming.
- Account backup and restore — what a backup file carries, what it deliberately drops, and what a restore does to the account.
- Running in production (Docker) — the prod stack: MySQL or SQLite, real mail transport, TLS or reverse proxy, updates, backups.
- First-run setup: creating the initial admin — see docs/first-run-setup.md.
- Cutting a release — how a
vX.Y.Ztag onmainbecomes the version the install and update scripts hand to users. - Design spec
- Implementation plans
- Contributing — issue-first workflow, branch conventions, and the quality gate. Licensed under the MIT license; notable changes land in the changelog.