Built for AI coding agents

Snap Context to Claude Codex Cursor Claude

The missing macOS screenshot tool for Claude Code, Codex, and Cursor. One capture lands as the right format — file path in your terminal, image in your chat, PNG in your editor.

Download for macOS View on GitHub
macOS 12.3+ · Apple Silicon & Intel · Free & open source · ~2 MB
or install with Homebrew brew install --cask CPPAlien/contextsnap/contextsnap
The problem

One screenshot. Three apps.
Three different formats.

macOS's built-in screenshot tool gives you one representation at a time. So you save to Desktop, hunt the file, drag it in. Every single time.

Claude Code wants a file path. Paste raw pixels into the terminal and you get garbage.

WeChat, Slack, iMessage want a real image. Paste a path and you get a blue link nobody can preview.

Preview, Figma, Notes want PNG bytes. Paste a file ref and you get an icon, not the pixels.

The fix

One capture, every paste format.

ContextSnap writes every shot to the clipboard as file path, image, and PNG bytes simultaneously. The app you paste into picks the format it understands.

Paste into…What lands
Claude Code / Codex / any terminal file path the agent reads the image from disk
Cursor chat panel image inline attachment, ready for vision input
WeChat / Slack / iMessage / Discord image exactly as if you screenshotted in-app
Preview / Notes / Figma / image editors PNG raw pixels
Finder / upload dialogs file ref drops the PNG as a real file
Features

Why you'll keep it open.

Designed so the agent-paste case just works — no more "drag from Finder" detours.

One hotkey

⇧⌘S anywhere — native macOS region selection, no learning curve.

Floating stack

Queue up multiple shots, drag tiles into wherever they need to go, clear when done.

Drop-to-import

Drag images straight from Chrome, Finder, or any app into the stack — auto-downloads web images.

Agent-first

Terminal paste case works on the first try. No more switching to Finder to drag a file in.

Local-first

Every shot archived under ~/Pictures/ContextSnap/. Nothing leaves your machine.

Open source · MIT

~2 MB. Pure Swift + SwiftUI. Read the whole thing in an afternoon.

Install

Up and running in 90 seconds.

Homebrew — recommended

brew install --cask CPPAlien/contextsnap/contextsnap

Homebrew downloads outside Safari's quarantine, so the app launches on first run without needing the xattr step below.

Manual .dmg

  1. Download ContextSnap-x.y.z.dmg from the latest release.
  2. Open the DMG and drag ContextSnap.app into Applications.
  3. This build isn't notarized yet (Apple's $99/yr Developer Program required), so macOS will refuse to open it on first launch. Remove the quarantine flag once:
    xattr -dr com.apple.quarantine /Applications/ContextSnap.app
  4. Launch ContextSnap. On the first capture, grant Screen Recording permission in System Settings → Privacy & Security, then trigger the hotkey again.
The workflow

One capture. Anywhere it needs to land.

  1. Press ⇧⌘S → drag-select a region.
  2. The shot appears in the floating stack (top-right corner).
  3. Switch to Claude Code⌘V → Claude sees the path and reads the image.
  4. Switch to WeChat⌘V → an actual image attachment. Same capture, no extra step.

Stop hunting for screenshots.

Press a hotkey. Paste anywhere. Your AI agent gets the path, your team gets the image.

Download for macOS