Docs / source reviewed
jev-desktop
Choose controls and actions in desktop apps
How this project uses Jev
- Input
- A native accessibility tree and an interaction goal
- Jev decides
- Target, action, risk, and stopping conditions
- Code executes
- Local desktop tools execute and observe again
Evidence and limitations
Jev support lives in separate agent-desktop skill/scripts. Capturing snapshots also costs time; inference latency is not total task latency.
This project has not been run independently here. Author-reported results are not independently verified results.