What people use it for
Stop re-finding what you already found.
Research has a cruel shape: the moment you need a source is always months after the moment you read it. Bookmarks hold the URL, but the URL is not the argument, and half the time the page has changed, moved, or died since. What you actually want to ask is "which of the things I read said X?", and a folder of links cannot answer that.
Save it while you read
One click in the extension, or a paste on the dashboard. The moment you save, Norget fetches and stores the readable text of the page itself: not a screenshot of the link, the words.
The text is yours now
If the page moves, paywalls itself, or disappears, your copy stays. Link rot stops mattering the day the text lives in your own archive.
Search by meaning, not keyword
You do not need the title or the site. Describe what it was about, like "the piece arguing write-ahead logs are everywhere", and semantic search finds it from the content.
Ask, and check the answer
Questions get answers assembled from your saves, and every part of the answer names the save it came from. You are never trusting a model’s memory. You are reading your own sources, faster.
What this does not do
Honest limits: Norget archives web pages, not files: no PDF or EPUB ingestion, so a papers-as-PDF workflow needs the HTML version. It does not do highlights or margin notes, and it does not sync to Obsidian or Notion; if that is your system, Readwise Reader is the better tool and our comparison page says so. Free plan: 100 saves, 50 questions a month.
Questions
- How is this different from asking ChatGPT?
- A general model answers from its training data and cannot tell you where a claim came from. Norget answers only from what you saved, and cites the specific saves, so the answer is checkable, and "I never saved anything about that" is a possible and honest response.
- What if a page is behind a paywall?
- Honestly: often only partially. Norget’s archiver fetches the page from our servers, without your login, so a hard paywall archives the public teaser rather than the full article. The exceptions are X and Instagram, where the extension captures the text from the page you are actually looking at. The save always works; how much text sits behind it depends on what the page shows a stranger.
- Can I trust the answers?
- Trust the citations, not the prose. Every answer names the saves it drew from and links back to them, so the check is one click into your own library. That is the design: the model assembles, your sources verify.
More