Pincho

To show Claude something, I took a screenshot, opened it in another app, marked it up, saved it and dragged it over. Five steps to say “look at this”.

I wanted to go from key to clipboard in almost no steps, annotate with dictated text, and have the image reach the AI ready to be understood.

• Capture and annotate on top. Arrows, shapes, numbered steps with notes, text, highlighter, magnifier and pixelation to hide what shouldn’t be seen.
• Copy for AI. The image comes out optimised and with its context: the app, the window and the web page it came from.
• Reads the text. It recognises what each screenshot says, on the Mac itself, and you can search it later in the history.
• Burst. It watches an area and, when you press, keeps the three seconds before and the two after. To catch the exact moment.
• Long screenshots. You scroll and it stitches the whole page together as you go down.
• Records the screen. Video or GIF, with your voice transcribed and the key frames ready to paste into Claude.
• Claude uses it on its own. Pincho connects to Claude, which can ask for a screenshot when it needs one without me touching anything.

I started by writing down what it had to do and what it didn’t. Then I built it in phases, testing each one before moving on: first capturing and annotating, then reading text and burst, and finally video and the connection to Claude.

I use it every day. It has replaced the Mac’s screenshots and every other app I had.

It’s for me. It isn’t for sale.

Swift, with AppKit and SwiftUI. ScreenCaptureKit to capture and record. Vision to read text on the Mac itself. And an MCP server to talk to Claude.

• Evolve Performance · Gym apps, in real use.
• Pastor de Agentes · Watches my AI agents while they work.
• My TV · My Fire TV’s system, rebuilt to explore films.