KAYA - 2026 Pricing, Features, Reviews & Alternatives

Kaya is a cloud-based machine runtime measurement platform built for contract manufacturers, job shops, and any industrial operation that invoices by the machine hour or uses machine hours for cost accounting. Rather than fitting sensors or accessing PLCs, Kaya reads each machine's tower light through the factory's existing camera infrastructure, classifying light states as green (running), amber (stopped), red (alarm), or off, and converting those readings into verified runtime hours per machine per day. Every reported hour links back to the specific camera frame it was measured from, making the data fully auditable. Setup takes approximately one hour and is completed entirely remotely, with no factory visit or production interruption required.

Kaya operates on AWS infrastructure using Greengrass, S3, DynamoDB, Cognito, and CDK. If the site's NVR or DVR is internet-reachable, no on-site installation is needed. Where the recorder is only accessible from within the factory network, a small Linux virtual server is placed on the local network; this server has no connection to any machine and cameras are not exposed to the internet. The platform supports any industrial machine fitted with a tower light, regardless of make, model, or age, covering CNC machining centres, presses, injection moulding machines, packaging lines, assembly cells, and robotic cells.

Reporting spans monthly per-machine summaries down to individual minutes, with drill-down access to the source camera frame at any point in time. When connected to an ERP (Odoo supported; others planned), work orders appear alongside runtime hours, and per-customer machine-hour reports can be exported as PDF or CSV for invoicing purposes. An optional part-counting add-on uses camera footage to count parts and cycles per machine. A shop-floor screen displays live machine status and runtime data for floor teams. Kaya also supports historical analysis of past recordings already held on the NVR or DVR, and can be connected to an external AI assistant so users can query machine data in plain language. Computer vision models output a confidence percentage for each tower light reading; if a camera drops out or a lamp is obscured, the system marks that period as no evidence rather than zero. Custom lamp training is available for non-standard colours or blink patterns.