Screen capture for Windows, built for the AI age
Show your AI exactly what you're looking at.
Press Print Screen, drag, let go. Capsule drops it straight into Claude, ChatGPT, Codex or Grok, with your secrets covered and the context that explains it.
11fn charge(input: &str) -> Result<u32> {12 let key = "sk-ant-api03-Xk29fQ"hidden;13 let total: u32 = parse(input)?;14}15error[E0308]: mismatched types--> src/main.rs:13:22| let total: u32 = parse(input)?;| ^^^^^^^^^^^^^ expected `u32`, found `i64`
Pasted into Claude
Ready to send. Your key never left the screen.
How it works
One key. Straight into the prompt.
Most screenshots end up in an AI chat. Capsule takes the steps in between away: no saving, no finding the window, no pasting.
- 01
Press Print Screen
Or Ctrl+Alt+S, from anywhere. The screen under your mouse freezes, ready to capture.
- 02
Drag round it
Or click a window to take all of it. Edges snap, and a magnifier shows the exact pixel. Press 1–4 to pick the AI, or T to send the text.
- 03
Let go
It's copied, saved, and pasted into your AI's prompt, ready to send. Hold Shift as you let go to mark it up first.
Ctrl+Alt+R takes the same area again and sends it, with no capture screen at all.
Context
Every capture remembers where it came from.
A screenshot on its own leaves your AI guessing. Capsule keeps what was around it: the app and window you were in, the page a picture came from, when you took it, where it went and what was hidden.
It's saved beside the picture, so it moves and syncs with it, and anything that reads your captures later knows what it's looking at.
// Capsule 2026-10-11 09.41.json
{
"app": "Code.exe",
"window": "main.rs — billing-service",
"taken": "2026-10-11T09:41:07+11:00",
"monitor": "2 of 2, 2560 × 1440",
"sent_to": "Claude",
"sent_as": "text",
"hidden": "1 API key",
"pinned": false
}Made for models
Captured the way your AI reads best.
Send the words, not the picture
Press T and Capsule reads the text on your PC: exact characters for a fraction of the tokens. Best for errors, code and logs.
Sized for the model
The size readout turns amber when your AI would shrink the picture, so you crop tighter and small text stays sharp.
Long pages, in readable slices
Scrolling capture joins a whole page, then sends it in model-sized parts, cut between lines of text, never through them.
Several screens, one prompt
Pick captures from the recent strip and send them together, oldest first, with a note of when and where each came from.
Privacy
Your secrets stay out of the chat.
Pasting a screenshot into an AI shouldn't mean pasting your API key with it.
Hiding errs on the side of covering too much. It's a safety net, not a guarantee, so glance before you send.
- API keys, tokens, passwords, card numbers and email addresses are covered before anything leaves the screen.
- Text is read on your PC, with Windows' own OCR. Capsule never uploads your captures; they sync to a cloud drive only if you turn that on.
- Capsule only talks to an AI if you turn on answers in Capsule with your own key, and it asks before a capture first goes out.
- One key, H, turns hiding off for a single capture when you need to share it whole.
Beyond the screenshot
Your captures, put to work.
An assistant beside every capture
It knows what it's looking at. An error gets Troubleshoot and Explain, a spreadsheet gets Extract table, a web page gets Summarise.
Find anything in plain words
“The error I captured yesterday.” “That chart of monthly sales.” Capsule reads every capture in the background, on your PC, so search just works.
Guides that write themselves
Click through a process and Capsule captures each step and drafts the words. Export to Markdown, a web page or PDF, or paste into Word or Notion.
Use the assistant free through the Claude app, or get answers right inside Capsule with your own Anthropic key. Extract and Share never use AI at all.
Features
And everything you'd want from a pro capture tool.
A real editor
Arrows, boxes, text, pen, highlighter and crop to any shape.
Numbered steps
Click to drop 1, 2, 3, so you can say “fix 2”.
Blur, pixelate, redact
Hide anything by hand, as well as automatically.
Timed capture
3, 5 or 10 seconds to open the menu you need.
Scrolling capture
A whole page or chat history in one picture.
Every screen at once
All your monitors in a single capture.
Gallery and recent strip
Every capture by day. One click to pick one up again.
Pictures from the web
Open any picture on a page full size, with its source.
Sync your way
OneDrive, Dropbox, Google Drive, iCloud or Box.
Any format
PNG, JPEG, WebP or PDF, named the way you like.
Lives in the tray
Out of the way, one key from anywhere, starts with Windows.
Key coach
Learn the shortcuts as you go, with points and streaks.
Be first to get Capsule.
Capsule is in its final stretch. Leave your email and we'll tell you the moment it's ready, and what it costs.
FAQ
Questions, answered.
Which AI apps does it work with?+
Claude (the desktop app, or Claude Code in a terminal), ChatGPT, Codex and Grok. Capsule finds the app's window, brings it forward and pastes for you. For anything else, it copies, and you paste.
Does it send my screen anywhere?+
Not on its own. Captures and their text stay on your PC. They only go where you send them, and the assistant only talks to an AI directly if you turn that on with your own key.
Do I need to pay for an AI as well?+
No. Capsule hands captures to the AI app you already use, free plan or paid. Answers inside Capsule are optional and use your own Anthropic key.
Is there a Mac version?+
Capsule is for Windows today. Join the list and you'll hear if that changes.
What will it cost?+
Pricing comes with the launch. Everyone on the early-access list hears it first.
Who makes Capsule?+
Ships to You Software, a subsidiary of DAC Creative Design. We build software we use every day, then ship it to you.