Self-Operating Computer

Open-source framework that operates your computer with multimodal models, captures the screen, and acts.

Self-Operating Computer, by HyperWrite, is one of the earliest published implementations of computer-use as an open-source framework. It reads the screen with multimodal models, decides on actions, and issues mouse and keyboard commands.

Because it is a framework rather than a packaged consumer app, the audience leans developer. The reference implementation is useful precisely because it makes the agent loop visible and modifiable. Anyone building a desktop agent from scratch can read this codebase first.

The project runs on macOS, Windows, and Linux and works with several model backends. It predates most of the polished commercial products in this space, which makes it a useful historical reference as much as a usable tool.