A native RSS reader for Linux and Omarchy, written in Rust. A three-pane library, article list, and reading view keeps your subscriptions and articles on your computer.
- RSS, Atom, and JSON Feed; manual refresh and scheduled refresh while the app is open.
- Nested folders: create, rename, move, and remove. Move feeds between folders without losing article history or read state. Removing a folder promotes its feeds and children to its parent.
- Content filters: hide articles by keyword or
/regex/, matched against the title and preview. Starred articles stay visible. - OPML import/export with duplicate detection and preserved folder hierarchy.
- A command line to list, search, star, and subscribe without the window, with
--jsonoutput for scripts and LLM agents. - All Unread, Today, Starred, All Articles, folder and feed views; full-text search and unread filtering.
- Read/unread, stars, bulk mark read with undo, and offline article text.
- A context menu on sidebar feeds, folders and views (right-click, Menu key or Shift+F10): open or copy a feed's home page, copy its feed URL, refresh, mark all as read, add a feed or folder inside a folder, edit, and remove.
- A context menu on articles: mark read or unread, star, mark the articles above or below as read, open in the browser, copy the link, and go to the article's feed.
- Sanitized HTML articles with images, code blocks, adjustable typography, copy link, and open original.
- Omarchy palette following across the top bar, menus, sidebar, and reader; light/dark modes, persisted window/pane sizes, keyboard navigation, and favicon caching.
- A companion Omarchy bar plugin: an RSS icon that lights up when something is unread, with the newest articles one click away.
- Conditional HTTP requests, bounded concurrency, timeouts, retry backoff, and visible per-feed errors.
Download the package for your system from GitHub Releases. Every option keeps your articles and settings when you update or uninstall. Releases before 0.2.5 have only the archives and the PKGBUILD; check that release's notes for what they need.
Install omafeed from Omarchy menu → Install → Package, or from a terminal:
omarchy pkg add omafeedIt updates with the rest of the system.
omafeed-bar puts an RSS icon in the Omarchy bar. Its dot lights up when something is unread; click it for the newest unread articles, filter them by feed, and open, star or mark them read without opening the window. Install omafeed first, then:
omarchy plugin add https://github.com/pch/omafeed-bar --enableRemove it with omarchy plugin remove io.github.pch.omafeed. The plugin uses the command line below and never touches the database directly.
Point your coding agent to GitHub Releases and tell it to install the latest version for your Linux distribution, or give it this prompt:
Install the latest release of omafeed for my Linux distribution from https://github.com/pch/omafeed/releases. Check my distribution and CPU architecture, choose a compatible installation method from the README, verify downloads against the release's SHA256SUMS, install it with its dependencies, and tell me how to launch it.
Install the prebuilt package for x86_64 from the latest release:
sudo pacman -U ./omafeed-VERSION-1-x86_64.pkg.tar.zstOr build the tagged source as a regular pacman package (this works on ARM too):
mkdir omafeed && cd omafeed
curl -LO https://github.com/pch/omafeed/releases/latest/download/PKGBUILD
makepkg -siTo update, repeat either when a new release is out. To uninstall: sudo pacman -R omafeed.
To follow the latest development code, build the development package instead:
git clone https://github.com/pch/omafeed.git
cd omafeed/packaging/omafeed-git
makepkg -siRerun makepkg -si in that folder to update. To uninstall: sudo pacman -R omafeed-git.
Each release has .deb and .rpm packages for Debian 13 or newer, Ubuntu 26.04 or newer, and Fedora 43 or newer. Ubuntu 24.04 is too old: omafeed needs libadwaita 1.6. Download the one for your system and architecture from the latest release, then:
sudo apt install ./omafeed_VERSION_amd64.deb # Debian, Ubuntu (arm64.deb on ARM)
sudo dnf install ./omafeed-VERSION-1.x86_64.rpm # Fedora (aarch64.rpm on ARM)They pull in GTK, libadwaita and WebKitGTK from your distribution. There is no package repository yet: to update, install the next release's package the same way. To uninstall: sudo apt remove omafeed or sudo dnf remove omafeed.
Installs to ~/.local without a package manager or root (on Omarchy or Arch; elsewhere install Rust, pkg-config and the GTK 4, libadwaita and WebKitGTK 6.0 development packages):
omarchy pkg add rust gtk4 libadwaita webkitgtk-6.0 pkgconf
git clone https://github.com/pch/omafeed.git
cd omafeed
make
make installMake sure ~/.local/bin is on your PATH. To update: git pull && make && make install. To uninstall: make uninstall.
For a system-wide install, build as your user and install with sudo make install PREFIX=/usr (uninstall with sudo make uninstall PREFIX=/usr).
Each release includes omafeed-VERSION-x86_64-linux.tar.gz and omafeed-VERSION-aarch64-linux.tar.gz. From 0.2.5 both are built on Debian 13 and need glibc 2.34 or newer, gtk4, libadwaita 1.6 or newer, and webkitgtk-6.0; earlier releases' notes say what theirs need. Extract the one for your architecture (uname -m) into ~/.local:
tar -xzf omafeed-*-$(uname -m)-linux.tar.gz -C ~/.local --strip-components=1After installing, open omafeed from your launcher or run omafeed.
Use the menu → Import OPML, or:
omafeed import /path/to/subscriptions.opml
omafeed refresh
omafeedOPML contains subscriptions, not historical articles or read/star state. The first refresh downloads the items currently published by each feed, initially unread. Use the checkmark above the article list to mark the current view read; Ctrl+Z undoes that action.
Click + in the header to Add feed (Ctrl+N) or create a New folder (Ctrl+Shift+N); either goes into the folder you are viewing. Subscribe with a website or feed URL: omafeed finds the site's RSS, Atom, or JSON feed, lets you choose when there are several, and uses the feed's title if you leave the name blank. Drag feeds and folders in the sidebar to move them: drop on a folder to put it inside, or on a smart view to move it to the top level. Right-click a feed or folder and choose Edit… to rename it, move it, or correct its URL without losing saved articles, or Remove… to remove it. Refresh in the same menu downloads just that feed, or every feed in the folder. Removing a feed deletes its articles after confirmation; removing a folder keeps them.
Search looks through the current view, including nested folders. It forgives typos (agentc finds agentic), treats regular plural and singular forms alike (box and boxes, story and stories), and matches the start of the last word you type, so results appear as you go. Words shorter than five letters are not typo-corrected; the last word can match prefixes from three letters onward. Numbers are matched exactly, and a word that appears in your articles is never swapped for a similar one. Summary-only feeds show the summary; open the original for the rest. Podcast episodes appear as download links.
Not yet supported: full-article extraction, podcast playback, sync between devices, and notifications.
omafeed is both the app and its command line. With no arguments it opens the window; with a command it does the job and exits. omafeed --help lists every command under the version number. It needs the GTK libraries installed (they come with the app) but no display, so it runs from a terminal, a script, or an SSH session. The commands read and write the same library as the app, and an open window picks up their changes within a couple of seconds.
| Command | What it does |
|---|---|
import FILE / export FILE |
Import or export subscriptions as OPML |
refresh |
Fetch every feed now |
status |
One line of totals, then any feed errors |
discover URL |
List the feeds a website advertises |
feeds |
Folders, feeds, unread counts and feed errors |
articles [OPTIONS] |
List articles, newest first |
search WORDS [OPTIONS] |
Find articles containing all the words; searches read and unread articles alike |
article ID [--html] |
One article's text, or its sanitized HTML |
read, unread, star, unstar ID... |
Change article state |
subscribe URL [--folder ID] [--title NAME] |
Find a site's feed and add it; run refresh to fetch its articles |
unsubscribe ID --yes |
Delete a feed and all its articles, starred ones included |
articles and search share these options; search also takes the words to find:
| Option | Meaning |
|---|---|
--scope S |
unread (the default for articles), today, starred, all (the default for search), feed:ID or folder:ID (folders include their subfolders) |
--search WORDS |
articles only: the same as omafeed search WORDS --scope unread |
--unread |
Only unread articles, whatever the scope |
--exact |
Disable typo correction, plural expansion and prefix matching |
--since 24h |
Published within 90m, 24h, 2d or 1w. --scope today means since local midnight; --since is a rolling window |
--limit N, --offset N |
Page through results (default 50, at least 1). Every page comes from one snapshot, so the app writing meanwhile cannot repeat or skip articles |
search searches an article's title, author and text, ignoring case and accents as supported by SQLite’s Unicode tokenizer. By default it uses the same forgiving search as the app; use --exact for whole-word matching without expansion, which scripts that need precise results should prefer. Every word must match, in any order (by default a close misspelling, the plural or singular, or for the last word a longer word starting with it also counts), and the words can be quoted or not: omafeed search "agentic engineering" and omafeed search agentic engineering do the same. Typed search syntax is quoted, so OR, *, quotes and title: mean nothing special (there is no OR or exact-phrase search), and a dash is ignored. Put -- before a word that starts with a dash. Results stay newest first, as in the app's own search field.
Failures print a message to stderr and exit 1; a mistyped option exits 2. Feed and folder IDs come from feeds, article IDs from articles.
Output is plain text by default. feeds prints tables, articles prints one tab-separated line per article (ID, read state, starred, local date, feed, title) so cut, awk and grep work on it, search prints the same lines as articles, and article prints a short header and then the text. Commands that change articles name every article they changed, so you can see what an ID was:
$ omafeed feeds
1 feeds · 0 folders · 9 unread · 1 starred
ID UNREAD FEED FOLDER ERROR
1 9 Rust Blog
$ omafeed articles --scope all --limit 2
1 read starred 2026-09-22 Rust Blog Announcing a Maintainer in Residence: Scott Schafer for the Cargo team
2 unread - 2026-09-21 Rust Blog GitHub Actions leaking secrets when Miri output is cached
More articles match; raise --limit or use --offset.
$ omafeed search miri secrets
2 unread - 2026-09-21 Rust Blog GitHub Actions leaking secrets when Miri output is cached
$ omafeed star 2
Starred 1 article
2 GitHub Actions leaking secrets when Miri output is cachedThe "More articles match" and "No articles match" notes go to stderr, so they never land in a pipe.
article keeps the article's shape: blank lines between paragraphs, # headings, - and 1. lists (nested ones indented), > quotes, and code in fenced blocks exactly as written. Links keep their text but not their address, and images appear as [image: description].
$ omafeed article 2
GitHub Actions leaking secrets when Miri output is cached
Rust Blog · Manish Goregaokar · 2026-09-21 · unread
https://blog.rust-lang.org/2026/09/21/github-actions-leaking-secrets-when-miri-output-is-cached/
The Rust Security Response Team was notified that Miri stores all environment variables to target/…
## Overview
GitHub Actions makes it possible to cache directories between runs…Add --json to feeds, articles, search, article, read, unread, star, unstar, subscribe, unsubscribe, refresh or discover to get one line of JSON instead, for scripts and LLM agents:
$ omafeed articles --since 24h --limit 1 --json
{"articles":[{"author":"Manish Goregaokar","feed":"Rust Blog","feed_id":1,"id":2,"preview":"The Rust Security Response Team was notified that Miri stores all environment variables…","published":"2026-09-21T00:00:00+00:00","read":false,"starred":false,"title":"GitHub Actions leaking secrets when Miri output is cached","url":"https://blog.rust-lang.org/2026/09/21/…"}],"count":1,"truncated":true}
$ omafeed search "miri secrets" --json
{"articles":[{"id":2,"title":"GitHub Actions leaking secrets when Miri output is cached","preview":"The Rust Security Response Team was notified that Miri stores…", …}],"count":1,"truncated":false}
$ omafeed refresh --json
{"failed":0,"feeds":[{"error":null,"title":"Rust Blog"}],"total":1}previewis the first 220 characters, not the passage that matched a search.article ID --jsonreturns the whole text intext, formatted as above, or the sanitized HTML inhtml.truncatedistruewhen more articles matched than--limitreturned. Raise the limit or page with--offset.read,unread,starandunstarcheck every ID first, so one wrong ID changes nothing, and return the ID and title of each article they changed.unsubscribewithout--yesdeletes nothing and reports how many articles, and how many starred ones, it would remove.
A daily digest, for example, is four commands:
omafeed refresh --json # which feeds updated, which failed
omafeed articles --since 24h --unread --limit 200 --json # titles, previews, IDs
omafeed article 42 --json # full text of the ones worth reading
omafeed read 42 43 44 # mark what was coveredFeed content is written by strangers. A tool that hands article text to a model should treat it as untrusted data, never as instructions.
The Omarchy bar plugin is built on this output: it reads feeds --json and articles --json and runs read, star, unstar and refresh. Renaming or removing those fields breaks it.
| Key | Action |
|---|---|
| J/K or ↓/↑ | Next/previous article |
| N | Next unread on the current page |
| Space | Scroll article; next unread at the end |
| Shift+Space | Scroll back |
| M | Toggle read |
| S | Toggle star |
| O | Open original in browser |
| Ctrl+R | Refresh |
| Ctrl+F | Search |
| Ctrl+N | Add feed |
| Ctrl+Shift+N | New folder |
| Menu or Shift+F10 | Context menu of the focused sidebar item or article |
| Ctrl+Shift+A | Mark current view read |
| Ctrl+Z | Undo bulk mark read |
| Ctrl+Q | Quit |
| ? | Shortcuts |
Articles, subscriptions, read state, stars, and the search index live in $XDG_DATA_HOME/omafeed/omafeed.db (default ~/.local/share/omafeed). Settings live in $XDG_CONFIG_HOME/omafeed/settings.toml, and website icons are cached under $XDG_CACHE_HOME/omafeed.
No account, telemetry, cloud sync, or background daemon. Closing omafeed stops scheduled refreshes. Requests go to your feed servers, their redirects, favicon URLs, and article image servers when enabled. Remote images can be disabled in Settings. Feeds on your own machine or network work when you enter their address, but a site on the internet cannot send omafeed to localhost or a private, link-local, or loopback IP address, whether by a redirect, a feed link found during discovery, or an icon link. Host names are not resolved for this check, and images in articles are loaded by WebKit without it. Article JavaScript, frames, forms, and embedded media are disabled; links open in your browser.
Article text works offline; images need a connection unless they were already loaded. Articles are never deleted automatically. To back up everything, copy the data directory while omafeed is closed; OPML export saves subscriptions only.
Theme following reads $XDG_STATE_HOME/omarchy/current/theme/colors.toml, with the older config location as fallback. Missing/invalid palettes use a built-in dark theme. The top bar, search field, menus, and reader update as soon as the theme files change. No Hyprland configuration is changed.
cargo run -p omafeed
make checkGTK4, libadwaita, WebKitGTK 6.0, a C compiler, and pkg-config are required. The Rust core is independent of GTK:
cargo test -p omafeed-coreThe native desktop smoke test (run headlessly in CI) also verifies WebKit rendering, read/star state, folder creation, and moving a feed through the real dialogs:
cargo test -p omafeed --features test-support --test desktop_smoke -- --ignoredTests use temporary SQLite databases and a localhost HTTP server; they do not depend on live blogs. Use OMAFEED_HOME=/tmp/omafeed-test to isolate data, settings, and cache during manual testing. The command line also answers --help and --version.
The workspace contains omafeed-core (database worker, migrations, OPML, fetch scheduling, sanitization) and omafeed-app (GTK interface and CLI). Database work runs on a dedicated worker; HTTP work runs on Tokio. The GTK thread receives results asynchronously.
Command output is checked byte for byte against transcripts in crates/omafeed-app/tests/snapshots/, because scripts and agents parse it. After an intended change to the output, rewrite them with OMAFEED_BLESS=1 cargo test -p omafeed --test cli_output and review the diff. ARCHITECTURE.md describes the modules, data flow, storage, and known limits, and CONTRIBUTING.md the rules that keep scripts, settings and libraries working across versions.
CI runs formatting, Clippy, tests, the documentation build, the headless desktop smoke test, a build with the minimum supported Rust version, an Arch package build, and cargo deny; changes to packaging also build the DEB and RPM packages and install them on each supported system. A vX.Y.Z tag publishes a release with its notes from packaging/release-notes/. packaging/RELEASING.md describes the downloads, how to release, and how to write the notes.
NetNewsWire, Omarchy Meeting Recorder, and Spotifast. omafeed is an independent Rust implementation; no source or visual assets were copied from those projects.
Content filters were contributed by UnderStackDev, modelled on the filters in Aura News.
MIT licensed.