CUA AI Operation¶
CUA (Computer Use Agent) reads the current KVM screen and executes mouse, keyboard, scrolling, dragging, waiting, and screenshot actions from a natural-language task. It runs from the AI panel in the console toolbar, which shows connection state, reasoning, screenshots, and actions.
Early version
CUA is still evolving; compatibility and stability depend on the selected model and API service.
Confirm before use
CUA directly controls the target machine's mouse and keyboard. Before running tasks that delete files, format disks, change system settings, or submit forms, make sure the instruction is clear and keep the session observable.
Configuration¶
Open AI -> Settings to configure CUA:
| Item | Description | Default |
|---|---|---|
| Enable AI operation | Allows CUA tasks to start | Off |
| Model | OpenAI model name to use | gpt-5.5 |
| API URL | Full endpoint URL. URLs containing /responses use the Responses API; URLs containing /chat/completions use the Chat Completions compatible path |
https://api.openai.com/v1/responses |
| API Key | Key used for model calls | Not configured |
Usage¶
- Open the KVM console and click AI in the top toolbar.
- In Settings, enable AI operation, fill in the model, API URL, and OpenAI API Key, then click Save.
- Describe the task in the input box and press
Ctrl+Enter(orCmd+Enteron macOS). CUA loops over screen capture, model decisions, and supported actions. Click Stop to interrupt or Clear to remove the execution record. A task cannot start until AI operation is enabled and an API key is configured.
