LatestArr

LatestArr is a self-hosted, open-source “what’s new” newsletter for your media server. It connects to Plex, Tautulli, Jellyfin, Emby, and your book, audiobook, and game libraries. It collects everything added recently and emails your users a clean, branded digest on a schedule you choose.

It exists because Tautulli’s newsletter, the closest existing tool, only speaks to Plex and gives you limited control over layout, scheduling, and branding. LatestArr is Plex-aware but not Plex-only, and you design the email with a live preview. No HTML required, though there’s a code mode when you want full control.

How it works

1. Connect a source. Point LatestArr at Tautulli, Plex, Jellyfin, Emby, a BookLore-family app, Audiobookshelf, or RomM.
2. Pick a design. Use the built-in Default design, or make your own in the design editor.
3. Add recipients and SMTP. Import your server’s users or add people by hand, group them, and connect the SMTP server you’ll send from.
4. Create a newsletter. Choose a schedule, a lookback window, its sources, and who gets it. It sends itself from there.

The whole thing runs as a single Docker container with SQLite storage.

Project status

These docs cover the current 0.11 release. Features marked New in 0.11 are new in it; if you’re upgrading from 0.10, read Upgrading to 0.11 first.

Known gaps today:

• No unsubscribe link. An admin deactivates recipients instead. That’s a real gap against CAN-SPAM and GDPR expectations for larger lists; see Unsubscribe.
• No roles yet. You can add more users, but every one is a full admin. Roles are planned (#257 ).
• Limited real-world testing. Testing so far is the maintainer’s own setup. Plex, BookLore, Grimmory, and Audiobookshelf need testers. Jellyfin and Emby are built from their API docs and haven’t been confirmed on a real server.

How it’s built

AI is LatestArr’s lead developer. The code, tests, CI, and docs are written by Claude Code , Anthropic’s AI coding agent. A human maintainer directs the work, reviews every change before it merges, and tests releases on a real setup. That’s worth knowing before you self-host a tool that holds credentials and may face the internet.

What keeps it honest:

• Automated checks on every change: lint, typecheck, build, and tests (including axe-core accessibility checks) on Node 24 and 26, plus a Docker build. Dependabot and CodeQL run continuously.
• Real-world testing: releases run against the maintainer’s own media servers. Where that isn’t possible yet, the docs mark the source “Needs testers”.
• Security by design: credentials encrypted at rest, server-side sessions, rate limiting, CSRF protection, and input validation on every route. See Security.

None of this makes it bug-free. Treat AI authorship as a reason to read the code before trusting it with sensitive data, the same care you’d give any early-stage open-source project. If you can help review or test it, see Contributing.