1. LLM Generation 2. Parser & Mapper 3. AppKit Execution Generate Token Extract Screen 1 Screen 2 USER_PROMPT [POINT:X,Y:L:2] TOKEN_PARSER COORD_MAPPER APPKIT_SCREEN_1 APPKIT_SCREEN_2 CURSOR_OVERLAY MONITOR VIEW BTN User requests UI interaction Model outputs coordinate token Extracts screen & x,y data Normalizes to OS coordinates Primary display bounds Secondary display bounds Animates cursor to target