macOS app brings on-device AI clipboard actions, no cloud needed

apfel-clip is a macOS menu bar application that simplifies text transformation by integrating on-device AI into the copy-paste workflow. The three-step flow is straightforward: users copy any content (text, code, a stack trace, JSON), press ⌘⇧V to open the action popover, and select a transformation. The result lands in the clipboard instantly, ready to paste or review in the integrated result panel. The app detects content type automatically—when you copy code, code-specific actions appear; when you copy prose, grammar and tone actions surface; errors and JSON get their own tailored options. All processing uses apfel, Apple's on-device AI engine bundled inside the app, so nothing leaves your Mac and no API keys are needed. The interface includes a result view that shows the original and transformed text side by side, allowing in-place edits, re-runs, or one-click copy. A persistent local history saves every transformation, so users can revisit or re-copy past results without reprocessing. Beyond built-in actions like "fix grammar," "rewrite tone," "translate," "summarise," and "explain code," users can create custom actions from any prompt and name them; these appear in the action list automatically. The app supports favourites, hiding, and drag-to-reorder, so users can shape their action list to their workflow. An optional auto-copy feature sends results directly to the clipboard the moment transformation finishes. Installation is minimal: download the signed, notarised arm64 zip, unzip, drag the .app to Applications, and run it. The app is free and MIT-licensed open source, available on GitHub, and can also be installed via Homebrew for automatic updates. System requirements are macOS 26 (Tahoe), Apple Silicon (M1 or later), and Apple Intelligence enabled in System Settings with English device language; no other software or configuration is needed.

apfel-clip addresses a practical workflow gap: most AI clipboard tools require cloud connectivity, API keys, and accept the privacy trade-off of sending data to remote servers. By leveraging Apple Intelligence—the on-device ML engine built into macOS 26—the app eliminates that friction. The requirement for macOS Tahoe and Apple Silicon (M1 or later) narrows its audience to recent Mac users, but for that group it removes setup overhead; there are no API keys to manage, no accounts to create, and no network latency. The app's design philosophy is speed and transparency: copy, hotkey, pick, paste. Content-aware detection means users are not shown irrelevant actions, and persistent local history lets users revisit past transformations without rerunning the model. The open-source MIT license invites contribution, which may expand the action library over time.

No comments yet. Be the first to share your thoughts!