Knowmarks

Knowmarks is a hosted knowledge store you save things into once and actually find again. This extension is the fastest way in: click the toolbar icon, then Save, and the link and title are saved — the default, one click away. Switch modes first when you want more: the full article text, a text selection, page images, or (on Claude/ChatGPT) the whole conversation — so you're not re-fetching a paywalled or client-rendered page later and hoping it still resolves. (plus whatever metadata the page itself declares — author, publish date, that kind of thing — no article text). Article mode adds the readable text of the page. Selection mode saves exactly the text you highlighted — and if you highlight something and hit Save without picking a mode, that highlighted text is what gets saved too, the same as if you'd clicked Selection yourself. Images mode lets you pick which images on the page to save. On a recognized Claude or ChatGPT conversation page, Conversation mode is selected for you automatically — visibly, as a normal mode chip you can switch away from — and pulls the full thread through the platform's own API, using your existing session, so a long chat saves as a real transcript instead of a screenshot of the visible portion. Whichever mode is showing when you click Save is exactly what gets saved: switch to Link/Article/Selection/Images first and the extension makes zero requests to the chat site. If Conversation is still selected and reading the transcript fails, Save stops and tells you inline instead of quietly saving a bare link — retry, or switch modes yourself. No page content is read until you pick a mode or click Save (the popup does show the current tab's URL and title right away, the same way any bookmarking extension would). The full, mode-by-mode breakdown of exactly what's sent is in the privacy policy. attaches a note if you want one, and hands it to Knowmarks' search, Pulse, and Companion — so a save from six weeks ago is something you can actually ask about, not just something that exists. server you've configured — your own local instance, or the hosted alpha's API at api-alpha.knowmarks.app — using an API key you paste into the extension's options page, when one is configured. Three things also talk to a destination other than Knowmarks, worth stating precisely: Conversation mode calls the AI platform's own API (claude.ai, chatgpt.com, or its legacy alias chat.openai.com) using your existing session there, to pull the chat transcript before uploading it to your Knowmarks server. Images mode loads image previews directly from their source sites/CDNs: switching to Images mode renders a picker grid where each candidate image is a live `<img src="...">` pointed at its original URL (`renderImagesGrid()`), so the popup fetches those images from wherever they're hosted in order to show you the grid — a real, if ordinary, third-party network request, the same one your browser already made once to render the page itself. Only the image *URLs* you select (not the image bytes) are then sent to your Knowmarks server on Save; Knowmarks itself fetches the image data server-side. And Catch-up navigates a worker tab to each queued item's URL, in your own logged-in browser, before it captures anything. It doesn't watch your browsing. Opening the popup checks in with your server once, for a pending-saves count — no page content in that request. Page or conversation content is sent only when you pick a capture mode or hit Save. is unlisted — install it once you have an alpha invite and an API key from your Settings page. Public availability follows the product's public launch. Privacy policy: https://knowmarks.app/privacy/