Table of contents
🤖
AI-generated summary
I’d introduced Karakeep a few months ago . It was good. Automatic AI tagging, mobile apps, full-text search. Except my NAS was starting to struggle.
700 MB of RAM for bookmarks
Karakeep is Node.js + a headless Chrome + MeiliSearch. Three Docker containers running permanently just to save links. On my DS920+ which already runs about fifteen services, that’s a lot. 700 MB of RAM for that alone is too much.
Readeck is a Go binary. A single container, 50 MB of RAM. The ratio is ridiculous.
The real game changer
What made me migrate wasn’t just the RAM. It’s the way Readeck saves articles.
Karakeep works server-side: you send it a URL, and its headless Chrome fetches the page. The problem is that this Chrome isn’t logged into anything. If you save an article from Le Monde or Mediapart, Karakeep fetches the page like an anonymous visitor. You get the free excerpt, not the full article. That happened to me dozens of times — a topic I’d already brought up .
Readeck does the opposite. The browser extension sends the page content exactly as you see it, in your browser. You’re logged into your Le Monde account? The full article goes to Readeck. Paywall, soft paywall, subscriber-only content — it all goes through because your browser does the work, not an anonymous bot on the server side.
What I lost
To be honest, Readeck doesn’t have everything:
- No automatic AI tagging. You have to tag manually. In practice I never used Karakeep’s suggestions anyway, so it makes no difference.
- No native mobile app. The web interface is responsive, it gets the job done.
- No built-in RSS ingestion. I use FreshRSS for that, so not a problem.
What I gained
On top of the RAM savings and paywalled article saving:
- Native ePub export. You can send an article directly to your Kobo or Kindle.
- OPDS. Your e-reader connects directly to Readeck and fetches articles. No manual transfer needed. Though in an ideal world — few e-readers are actually compatible…
- Speed. The interface is instant. No loading time, no spinner.
The migration
I migrated my 103 bookmarks via the API on both sides. A Python script that reads the Karakeep export and posts the URLs to Readeck. Readeck re-fetches each page (so paywalled articles won’t be complete in the migration, only future ones). 5 minutes, zero loss.
If you’re using Karakeep and the RAM consumption is bothering you, or if you’ve ever been frustrated finding a truncated article behind a paywall, try Readeck. It’s exactly what Omnivore should have become before it shut down.
In the end, I do lose the auto-tagging with AI and especially the AI-generated page summary that let you understand at a glance what a saved page was about. So it depends on your needs — pick the service that fits best.
Feel free to reach out if you’d like to react to this article.
