What you’ll know by the end of this check
- What drag-and-drop actually accepts (files, images, URLs)
- Why this tiny feature kills a surprising amount of friction
- One unexpected workflow it unlocks
The shortest possible answer
Drop a file onto the Claude Code desktop window and it enters the conversation. Images, PDFs, text files, screenshots, URLs — all of it. No upload button. No file picker. Drop and go.
Small feature. High frequency. Adds up fast.
Why it matters
Think about how often you want to show Claude something in the middle of a session:
- A screenshot of a bug
- A design mock from Figma (exported as PNG)
- A log file you just downloaded
- A PDF spec you’re working against
- A URL to a docs page
Before drag-and-drop: find the upload button, click it, browse the file system, select the file, wait for upload, describe what you just uploaded.
After drag-and-drop: drag, drop, keep talking.
The feature is 30 seconds of engineering time per use. Saves maybe 15 seconds each time. You’ll use it dozens of times a week.
The one unexpected workflow
Screenshot-driven debugging. macOS Cmd+Shift+4 takes a screenshot to the clipboard (or Cmd+Shift+Ctrl+4 to clipboard directly). But saving to a file with Cmd+Shift+4 drops it on your desktop — then you drag it straight into Claude Code.
That combo means: see a bug, screenshot the visual, drag it in, say “what’s wrong here.” Loop time is seconds, not minutes.
Same pattern works for:
- Dragging design exports from Figma/Sketch into the chat for iteration
- Dragging error screenshots from Slack threads into the conversation
- Dragging downloaded log files straight from your download folder
The feature is boring. The compound on a full workday of agent work is not.
What drops and what doesn’t
Likely to work:
- Images (PNG, JPG, screenshots)
- Documents (PDF, markdown, text)
- Code files
- URLs (dragged from address bar)
Less likely or uneven:
- Folders (may not recurse; may only take the first file)
- Very large files (check size limits)
- Files with unusual extensions that Claude can’t parse yet
When in doubt, try it. If the drop doesn’t produce what you expected, the UI should tell you why.
Things to try right now (4 minutes)
- Take a screenshot of anything — your terminal, a browser tab, a wireframe.
- Drag it into an open Claude Code desktop session.
- Ask Claude to describe what it sees. Confirm it processed the image.
- Drag a URL from your browser address bar. Ask Claude to summarize what’s at that URL.
The canonical version
Verify behavior against Anthropic’s official release notes.
Ready to verify this check?
You’ve dropped at least two things into a session (an image and something else). You know the screenshot-driven debugging workflow. Mark it cleared — you’ve finished the Claude Code Redesign dispatch.