Kaya measures how many hours each industrial machine actually worked by reading its tower light (signal lamp) through the factory's cameras, usually the ones already installed, as long as they can see the light. Computer vision reads the colour of the light and, when it is amber, checks whether anyone is at the machine. No sensors, current clamps or other hardware go on the equipment, and Kaya needs no access to the PLC or machine controller. Each period is recorded as producing, stopped, alarm, unattended, lamp off or no evidence. Amber time with nobody at the machine is reported separately and not counted as runtime, and when there is no usable camera frame the time is left as a gap instead of being guessed. If an hour looks wrong, the factory can open the camera frames it was measured from and check the reading. Reports show floor, day and machine views and can be printed or exported to PDF and CSV. Connecting an ERP is optional; with it (Odoo today), hours can also be assigned to work orders and customers. Kaya covers Availability, the A in OEE; Performance and Quality also need part and scrap counts. It works on any machine with a tower light a camera can see, such as CNC machines, presses, injection moulding machines and packaging lines. Setup takes about an hour and is done remotely, and the customer's own AI assistant, once connected, can guide it and answer questions about runtime. Customer data is stored in the EU.