ContextBolt Bookmarks features

Here is everything ContextBolt Bookmarks does, and how each part helps you find what you saved.

Automatic capture

ContextBolt captures two kinds of save. Web pages, one click each, with the page text. And your saves from three social platforms, each of which hides them in its own way.

• Any web page is one click. Open the popup and press Save this page, and it keeps the title, the author, the site and the article text, stripping the navigation, adverts, comments and cookie banners first. Search covers the first 10,000 characters, and ContextBolt Reader keeps up to 100,000. Highlight a paragraph before you save and only that is kept. Because it reads the tab already in front of you, pages behind a login work the same as public ones.
• A PDF you open from a web link saves the same way, with its text. A scanned PDF has no text to keep, so it saves as its title and link, and so does a file over 25 MB. A PDF opened from a file on your computer cannot be saved.
• X / Twitter saves are captured in the background when you open your bookmarks page, and a new one is saved the moment you tap X’s bookmark icon. No API keys and no manual export.
• Reddit saves are read from your saved list whenever you open Reddit. Reddit only lists your latest 1,000 saves, so ContextBolt keeps saves as they enter that list. Ones already past it need Reddit’s own data export.
• LinkedIn posts get a one-click save button in the feed, so the good ones do not vanish before you act. Posts you saved on LinkedIn earlier come in through the Saved Items CSV from LinkedIn’s data export, which you import in Settings.

Everything lands in one library that you control, instead of a browser folder plus three walled gardens.

ContextBolt Reader

Click any save in your dashboard and it opens in ContextBolt Reader, a clean reading view inside ContextBolt. One button at the top takes you to the original. It reads Open original on a web page, Open PDF on a PDF, and Open on X, Open on Reddit or Open on LinkedIn on a post. Press Esc to close it.

• Web pages: open as a formatted article, with the headings, lists, quotes, code and links kept, built from the text saved when you clicked Save this page. Images in articles are not saved. A page longer than 100,000 characters ends with a link to read the rest on the original.
• PDFs: open as text, with the page count.
• X, Reddit and LinkedIn posts: open as the post, with the author and the text. X posts also show up to four photos and the post they quote, and a Reddit post can show one image. Photos load from the platform, so a photo the platform removes stops showing. The text stays.
• Highlights: if you highlighted text before saving, the Reader shows just that, marked as your highlight.

The text is stored in your browser, so a save stays readable after the original page, PDF or post changes or disappears. That copy stays on the device where you saved it. Pro’s cloud sync carries the first 2,000 characters of a page, so on another device a long article ends early, with a link to the original.

Saves made before ContextBolt Reader arrived open in it too, showing the plain text kept at the time. A PDF saved before then shows only its title and link, so open the PDF and save it again to keep its text.

Turning it off: ContextBolt Reader is on by default, on Basic and Pro. To turn it off, open Settings in your dashboard and flip the ContextBolt Reader switch. Clicks then open the original in a new tab, as they did before. Whatever the setting, a Ctrl, Cmd or Shift click, or a middle click, always opens the original.

AI tagging and clusters

The moment a bookmark is captured, AI reads it, tags it, and files it. You never type a tag or pick a folder.

As your library grows, related saves roll up into topic clusters on their own. Browse a cluster to see everything you have saved on a subject, or search across all of them at once. The structure builds itself, so there is no upkeep.

Instant search

Search inside the extension is a local ranked index. It runs on your device, it returns as you type, and it tolerates typos, plurals and word order. It ranks results instead of filtering them out, so the closest match comes first rather than nothing at all.

Because AI tags every save with its topic, the index has more to match on than the raw text. Search “productivity” and you get the thread that never used the word, because the tag carries it. That is a real step up from the substring match your browser does, and it is not the same thing as searching by meaning.

Semantic search, which matches meaning rather than words, runs server-side through the MCP endpoint on Pro. That is what lets an agent find “the thread about pricing anchoring” when nothing in the save says anchoring.

Export

Your library is yours to take with you. On Pro, export your bookmarks to CSV, JSON or Excel at any time, all of them or filtered by platform and by when you saved them, so you have your own copy for backup, migration or use in another tool. On Basic the export button shows with a lock.

Local-first and privacy

ContextBolt is built to keep your library on your machine.

• Search runs in your browser: Your library does not have to leave your device to be searchable.
• We keep no copy of your bookmarks by default: the text of a save goes to our servers, and on to AI providers, to be tagged and embedded, and is not stored after that. What does persist is your account and a log of the license checks the extension makes, which includes your IP. The account starts anonymously; signing in adds your verified email.
• You stay in control: Your collection is yours. You can clear it whenever you want, and export it whenever you want on Pro.

Cloud sync (Pro)

Pro syncs your library across devices, and it is what lets the MCP endpoint answer your AI tools. It comes with the plan and runs on its own; there is no switch. Traffic to us is over HTTPS and the disks it lands on are encrypted, but that is storage security rather than end-to-end encryption. We can read the synced copy, because the endpoint has to read it to answer a question. If you would rather we did not hold it, Basic keeps the library on your machine.

The MCP endpoint (Pro)

Pro adds MCP access so your AI agent can use your library directly. Add https://api.contextbolt.app/mcp in an OAuth-capable client, sign in with your extension account email, and approve the connection. Claude, Cursor or Codex can then search your saves, pull a topic, or save something new, all inside a conversation. Existing personal MCP URLs still work as a compatibility option. See Connect your agent for the steps.

The full list of what your agent can do is on the MCP tools page.