01pin
Pin what matters
Press the hotkey and drag regions over the parts you care about. Type a short label for each one — that label is the instruction your agent reads.
developers.openai.com
macOS native, private and on-device
The Mac screenshot tool
your AI can read
Annotate any app, then send the screenshot, every region crop, and a structured manifest to Claude, ChatGPT, or Cursor — in one paste. So it fixes the right thing the first time — fewer iterations, fewer tokens. One-time, on-device.
checkout.app / pay
Free shipping over $50wrong color
Pay $48.20
misaligned button
produces
SESSION.md
# Screentack capture · 2 regions
display: Built-in Retina · 2560×1600 @2x
▸ region-1 "misaligned button"
position: { x: 71%, y: 64% }
crop: region-1.png
▸ region-2 "wrong color"
position: { x: 22%, y: 41% }
crop: region-2.png
→ paste: screenshot + 2 crops + this manifest
One paste = full screenshot + every region crop + this manifest.
One click — send screen context to your favorite apps, and speak your coding agent's language over MCP
ClaudeChatGPTGeminiCopilotCursorWindsurfCodex
From screen to context
The whole loop, without leaving your screen
Point at what is wrong, and your agent points back at what it understood — before it changes a line.
developers.openai.com
02send
Send it in one paste
Submit, and your agent gets all of it at once: the full screenshot, every region crop in order, and a spatial manifest describing where each one sits.
Claude
03draws back
Your agent draws back
This is the part nothing else does. Before it touches your code, the agent marks up your live screen with what it thinks you meant — labelled boxes, right on the page.
Acme Analytics
04you correct
You correct it — it re-reads
Wrong box? Tag it and move on. The agent polls for your notes, adjusts its plan, and carries on — no re-explaining, and no wrong fix shipped.
Acme Analytics
One tool from capture to context
Screenshot, annotate, record, and redact. Everything your agent needs to see, in a format it can read.
Multilingual OCR
Pull text out of any region. On-device recognition handles many languages and writing systems, straight to your clipboard.
Record like Loom
Record a window or the whole screen with your camera bubble. Click to pick the target, a red border shows what is captured.
Spotlight redaction
Blur everything outside your pins before sharing. Keep the bug sharp, hide the secrets.
MCP agent tools
14 tools any MCP agent can call. Capture, annotate, record, and read text, programmatically.
› capture_window(app: "Xcode", ocr: true)
✓ screenshot + recognized text
✓ screenshot + recognized text
Exact crops plus a spatial manifest mean your agent fixes the right thing the first time — not the fifth. Fewer round-trips, less guesswork, less token churn.
- Semantic recording (soon)
- Direct send to Claude, ChatGPT, Cursor
- Capture history
- Private and on-device
- One-time purchase
Not just another screenshot tool
| Capability | Loom | CleanShot | Screentack |
|---|---|---|---|
| Screen recording | Screenshots only | ||
| Camera overlay (like Loom) | |||
| Region annotations | |||
| App-level recording | |||
| Text recognition (OCR) | |||
| Agent-readable manifest | |||
| MCP agent tools | |||
| Agent annotates your screen back | |||
| macOS native | |||
| Private, on-device by default | |||
| No subscription (one-time) | |||
| Cloud sharing | Coming soon |
Your screen, ready for your agent
Download free and use everything for 7 days. Keep it for a one-time $29. No subscription, no account.
One-time purchase · free updates
Free forever over MCP for your AI agent.
Want a heads-up on new features? Get occasional release notes.