yarr

Last reviewed Sep 13, 2026 · 774 words

Download one file from the releases page, unzip it, run ./yarr -addr 0.0.0.0:7070 -auth me:longpassword, open port 7070 in a browser, import your OPML. That is a complete yarr deployment, and it took me less time to do than to type this paragraph. There is no database server, no config file, no PHP, and the process sits at a few tens of MB of RAM with 300 feeds loaded. If you are one person who wants to read feeds and nothing else, this is the shortest path in the category.

Everything lives in one SQLite file

yarr keeps feeds, items, read state and settings in a single SQLite database, by default under your user's config directory (~/.config/yarr/storage.db on Linux) and wherever -db points otherwise. That single file is the backup, the migration, and the disaster recovery plan: copy it, and you have moved. Pair it with a systemd unit and you have the whole thing under supervision:

[Unit]
Description=yarr
After=network.target

[Service]
ExecStart=/usr/local/bin/yarr -addr 127.0.0.1:7070 -auth me:longpassword -db /var/lib/yarr/storage.db
Restart=always
User=yarr

Front it with Caddy or whatever reverse proxy you already run, and you get TLS. If you must serve it under a subpath, -base /rss handles the prefix, though a dedicated subdomain is less fuss. The Docker route exists too, but for a static Go binary I find it adds a layer for no benefit; the binary is the container.

The reader is fast because it does little

The web interface is a three-pane layout: folders and feeds on the left, item list in the middle, article on the right, with keyboard shortcuts for moving, marking read and starring that behave the way Google Reader trained a generation to expect. Feed refresh runs on a fixed interval you set in the settings; with 300 feeds at a 30-minute interval the process is idle for all but a few seconds an hour, and page loads stay well under 100 ms because everything is a local SQLite read. There is a "read full content" toggle that fetches and extracts the article when a feed only ships a summary; it works on most blogs and fails on the sites that fight scrapers, same as every other reader's readability mode.

The feature list ends about there. No rules that auto-mark noisy feeds read, no filters by keyword, no saved searches, no scraping of pages without feeds, no integrations that push to a read-later service. yarr is honest about this: the project describes itself as minimal and the maintainer has declined a long list of feature requests to keep it so.

Phone access is through the Fever API

yarr speaks the Fever API, which most iOS and Android feed clients still support for legacy reasons. Enable it in settings, point your app at https://rss.example.com/fever/ with the same credentials, and read state syncs both ways. It is an old protocol with rough edges, notably slow initial syncs for large item counts, and the newer Google Reader-style API that Miniflux and FreshRSS expose is not offered. If you read mostly on a phone via a dedicated app, this is the single strongest argument for picking one of those instead.

Single-user is a hard line

There is one login. Not one admin plus readers, not one login per family member: one. Two people wanting separate subscription lists means two yarr processes on two ports with two databases, which is workable at this footprint but clumsy. FreshRSS handles multiple users natively, and Miniflux does as well with a PostgreSQL requirement attached; both are bigger projects with more eyes on them, and for a household or a small team they are the safer choice. My FreshRSS vs Miniflux piece picks between the two.

What I'd do

If it is just you, and your reading happens in a browser, run yarr as a systemd service behind your existing reverse proxy, back up the one SQLite file nightly with everything else, and stop thinking about it; it is the least operational burden any feed reader can impose. Move to Miniflux the day you want filtering rules, a second user, or a modern sync API for a phone app. Either way, the case for running your own reader at all is made in RSS isn't dead, and yarr is the cheapest way to test whether the habit sticks.