Capturing

What ⌃⌥C saves from each app, notes, images, duplicates, and enrichment.

Press ⌃⌥C anywhere. cap looks at the frontmost app and saves the most useful thing it can read — always without touching what you’re doing. The hotkey always captures: press it five times quickly and you get five captures.

A toast appears in the top-right corner for a few seconds confirming what was saved. Click the toast to attach a note — ⏎ saves, Esc dismisses.

What gets captured where

Safari, Chrome, and Arc give the full experience: cap asks the browser for the front tab’s URL and title over Apple Events, grabs your selection, and extracts the readable page body directly from the rendered tab. Because the body comes from your own tab, paywalled or logged-in content you can see is captured without any network request.

Firefox doesn’t support the Apple Events interface cap uses, so the hotkey can’t see the tab’s URL — it captures your selected text only. To capture a Firefox link, copy the URL and press the hotkey, or use cap add <url>. Bodies for those captures come from cap’s own hardened fetch instead of the tab.

Everything else — editors, terminals, chat apps, PDF viewers — captures the selected text, remembering which app it came from.

Selection and the clipboard fallback

cap first reads your selection through the Accessibility API. When an app doesn’t expose its selection that way, cap falls back to a synthetic ⌘C: it snapshots your clipboard, copies, captures, and restores the clipboard exactly as it was.

The fallback refuses to run when the clipboard holds promised files (like a drag-in-progress) or very large data — cap then captures what it can without it, and the toast says why.

If secure input is active — a password field, or an app like a password manager holding the secure-input flag — cap refuses to capture at all. Nothing is read, nothing is stored, and the toast says capture was blocked.

What a capture is

Each capture is a link, a text, or an image:

• Link — a URL was found (from the browser tab or in the captured text). Stores URL, title, your selection, and later the readable page body.
• Text — no URL; the selected text itself is the capture.
• Image — an image on the clipboard with no URL or text taking precedence. The image is stored as a file, and any text inside it is recognized on-device and made searchable. Images come from the clipboard only — there is no screenshot mode.

Duplicates

Capturing the same thing twice doesn’t create a second entry. cap normalizes URLs — dropping fragments and tracking parameters like utm_*, fbclid, and gclid — and hashes the content, so a re-capture bumps the existing entry’s seen count and timestamp instead. The toast tells you: “Already captured 3w ago”.

Re-capturing something whose page fetch previously failed retries the fetch.

Enrichment

Link captures are enriched in the background by cap-agent: the page body is extracted (from the tab where possible, otherwise fetched), cleaned to readable article text, and indexed for search.

• A body under about 120 words — or one that looks like a login or paywall wall — is marked thin rather than treated as the article.
• Failures keep the capture (URL, title, selection) and are retried up to three times; after that the capture is marked failed and the menu-bar icon shows a badge. cap refetch requeues failures, as does simply re-capturing the page.
• On battery the agent enriches one capture at a time; on mains power, three.

From the terminal

cap add captures without the app: URLs, text, stdin, with --note, --title, --no-fetch, and --wait. See the CLI overview.