YOUR EXECUTION ENVIRONMENT

Settings

Manage input delivery, OCR providers and shortcuts in one place.

General

Configure keep-on-top, automatic activation, dry run and the default AI provider. The AI timeout limits provider response time; it is not a visual similarity threshold.

Foreground input

Uses physical mouse/keyboard delivery and requires the correct focus. Your own input can interfere with the run, so keep initial tests short and observable.

Background window messages

Sends messages to compatible windows without physical mouse input or automatic activation. Supports single-key Tap and text/variable input, not Hold/Down/Up, multi-key combinations or Ctrl/Alt/Shift modifiers.

Use Window scope for image/OCR operations. A covered window differs from a minimized one: do not expect normal capture from a minimized target. Compatibility with Raw Input/DirectInput games is not guaranteed.

Gemini

Enter your own key in the protected field. Refresh the model list, select a suitable model and edit the reading prompt. Test knowing that the selected OCR area is sent to that provider. Never put keys in macro text.

OpenRouter

Configure its key, model, base API URL and OCR prompt separately. The model must support image input; its name alone does not establish that capability. Provider pricing and quotas apply separately.

Shortcuts

Set run, stop, fallback stop and recording-stop keys. Resolve conflicts with group shortcuts. Verify the emergency-stop key before executing a real task.

Local versus remote processing

Windows and Paddle OCR are local; AI OCR uses a remote API. MCP workflow design and AI OCR are separate capabilities. Using one does not require using the other.

Actual interface images

Settings · General · actual application, sample data
Settings · Gemini · actual application, sample data
Settings · OpenRouter · actual application, sample data
Settings · Shortcuts · actual application, sample data
Next stepBuild a workflow with these modules ↗Resolve a problem ↗