Teach a desktop agent by doing a task once. Understudy learns the workflow, turns it into a reusable skill, and can later execute it for a new input across GUI apps, browsers, shell tools, files, and messaging. It’s open source, local-first, and designed for real cross-app work instead of brittle macro playback. Current state: macOS today, MIT licensed, with teach-by-demonstration working and route optimization still early.

Framer — Launch websites with enterprise needs at startup speeds.

Launch websites with enterprise needs at startup speeds.

I built Understudy because too much real work still lives outside APIs and browser tabs.

A lot of workflows still jump across native desktop apps, websites, terminals, and chat tools. Most agents only live in one of those surfaces. Understudy is my attempt to make an agent that can work across all of them in one local runtime.

The part I care most about is teach-by-demonstration. Instead of hardcoding coordinates like a macro recorder, you do a task once, and the system tries to extract the intent and publish a reusable skill.

In the demo, I teach it to search Google Images for Sam Altman, download a photo, remove the background in Pixelmator Pro, export it, and send it through Telegram. Then I ask it to do the same for Elon Musk.

It’s still early and I want to be honest about that: macOS only for now.

Would love feedback on the product direction and on which cross-app workflow you’d teach first.

GitHub: https://github.com/understudy-ai...
Demo:

Framer — Launch websites with enterprise needs at startup speeds.

Launch websites with enterprise needs at startup speeds.

Understudy

Open-source desktop agent that learns from demonstrations

Open-source desktop agent that learns from demonstrations