Changelog

• New Clean up the Shelf — a new ✨ button reviews your old screenshots before anything goes. Smart mode proposes exact duplicates, the retakes of a burst, screenshots never reopened after 30 days and anything older than 90 days; Custom mode lets you pick the age yourself. Everything is grouped by reason and ticked, untick what you want to keep, and nothing moves until you confirm — then it goes to the Trash, not into the void. Favorites, titled screenshots, the last week and images macOS didn’t take as screenshots are never proposed, so the photos on your Desktop are safe.
• New Annotate whatever is on the clipboard — ⌥⇧A, or “Annotate Clipboard Image” in the menu, opens the copied image straight in the editor, a file copied in Finder included. Nothing to annotate? A small toast says so instead of an alert stealing focus.
• New Drop an image onto the menu bar icon to open it in the editor.
• Changed The editor and the Shelf now draw their content under a transparent title bar — the first step towards the Liquid Glass look of macOS 26.

• New Every shape can carry its own description now — rectangles, arrows and redacted areas each get a field beside them, so circling a misaligned button finally has somewhere to say why. Until now only a numbered marker could hold a note, and describing a shape meant dropping a marker on top of it or pushing the explanation into the general instructions.
• New Those descriptions travel wherever the capture goes: into capture.md, into an optional note on each shape in capture.json, and into a new ANNOTATIONS section of the baked-in legend. It matters most on a redacted area — “what I hid, and why” is exactly what an agent needs to know, and it was the one part of the handoff that had no way to say it.

• New Install the pinpoint command from the app — Settings ▸ Capture now has a “Command line tool” section that puts pinpoint on your PATH in one click, asking for permission once. It was already there if you installed with Homebrew; this is for everyone who downloaded the DMG and would otherwise have had to symlink it by hand.

• New Pinpoint now hands your capture to an AI agent as files. Every copy also writes capture.png, capture.md and capture.json to a stable folder, so an agent opens the real image with its own Read tool instead of choking on an inline one.
• New Accessibility context under every marker — a pin is no longer just a percentage but “AXButton ‘Login’ in Safari”, turning a spot on your screen into an anchor an agent can act on. Optional, and off until you grant the permission.
• New On-device OCR reads the text under each marker — error messages, logs, class names — and sends it along with the capture. Nothing is uploaded anywhere.
• New A pinpoint command line, an MCP server (pinpoint mcp) and a pinpoint:// URL scheme, so agents, scripts and shell hooks can drive Pinpoint. The MCP server hands over file paths, never inline images.
• New Redaction tool (⌘4) — drag over a token or an API key and it is painted out of the picture, the text, the JSON and the accessibility context alike, before anything is shared.
• New Undo and Redo (⌘Z / ⇧⌘Z) throughout the editor — and arrows and rectangles can finally be selected, moved and resized after you have drawn them.
• New Window capture — hold Space while selecting to highlight a single window and grab it cleanly, rounded corners included.
• New The selection rectangle is now adjustable before you commit it: drag it around, resize it by its handles, nudge it with the arrow keys, Enter to confirm.
• New Task presets (Bug, Review, Implement) prefill the framing you keep retyping, and a structured JSON export documents every capture against a published schema.
• New Shelf: search your library, contextual empty states, and a confirmation before deleting several screenshots at once. Editor: delete the selected marker with ⌫, tool shortcuts ⌘1/⌘2/⌘3, and tooltips everywhere.
• New VoiceOver labels, Dynamic Type and reduce-motion support across the editor and the shelf.
• Fix What you see is now what you export: on-screen marker and stroke sizes follow the same formula as the exported image, so a small capture no longer ships with oversized annotations.
• Fix Exported images were silently rendered at twice the requested size on Retina, which made every pixel coordinate in the copied text describe a grid half the size of the image it shipped with. Files are smaller and the numbers now match.
• Fix A failed copy or save is reported instead of quietly showing “Copied!”, and the Screen Recording permission is checked before a capture, with a button that opens the right System Settings pane.

• New Crop tool in the editor — trim a capture with eight handles and a rule-of-thirds grid; existing markers and arrows are remapped into the cropped frame.
• New Optional capture timer — a 3, 5, or 10-second countdown before the shot so you can arrange your UI or open a menu; press Esc to cancel.
• New A “What’s New…” item in the menu bar opens this changelog, so you can always see what each release added.

Crop tool and capture timer contributed by @ganuong11 — thank you! 🙏

• Fix Markers placed near an edge are no longer clipped in the exported image — the badge is kept fully inside the picture.
• New Shelf: ⌘A selects all screenshots and Esc exits selection mode.
• Fix The editor’s tool picker no longer shifts around in narrow windows.

• New Automatic updates — a new “Check for updates…” menu item plus background checks, so new releases install in one click.
• Fix Multi-monitor capture: the selection overlay now dims and works on every display, not just the primary one.
• Fix Copied images always paste — the “Copy for the agent” image is capped so it reliably pastes into Claude, GitHub, and other targets.
• New “Save image…” (⌘S) exports the full-resolution annotated PNG.
• New The installed version is now shown in Settings.

• New Edit in Pinpoint from the Shelf — reopen any screenshot, including a native ⌘⇧4 capture, as an annotation session.
• New Custom Shelf titles — rename a capture in the Shelf without touching the file on disk.
• Fix Pasting a capture into a terminal now keeps the image instead of dropping it for the text.

• New First public release — capture a region, drop numbered markers, and copy a ready-to-paste prompt for your AI agent.