What it does
An MCP server that reads the screen with OmniParser and operates the on-screen interface automatically. It analyses what is visible and performs actions on the GUI for the AI client. The project is MIT licensed, but the OmniParser models have their own licences, which the README links to. The author reports testing on Windows.
Use cases
- 01Fill in a form in a legacy desktop application
- 02Repeat a routine click sequence in a desktop tool
- 03Test a Windows interface through an AI client