Most ERPNext Manufacturing setups still capture production by hand. Operators tap "Complete" on a Job Card at the end of a shift, guess the scrap count, and downtime lives in a notebook that never reaches the ERP. The result is OEE numbers you cannot trust, backdated production entries, and no way to see a stalled line until the daily report. ERPNext core has excellent Work Order, Job Card, and Workstation DocTypes, but nothing that talks to a CNC controller, a PLC register, or an MQTT broker — the machine floor and the ERP are simply not connected.

The IIoT Machine Connector closes that gap. It is delivered in two parts: a lightweight edge agent that runs on-premise (a mini-PC or industrial gateway on the shop-floor network) and a proper Frappe app installed on your ERPNext bench. The edge agent speaks three protocols — an OPC-UA client for modern CNCs and robots, an MQTT subscriber for brokers like Mosquitto/EMQX and Sparkplug B devices, and Modbus TCP/RTU polling for legacy PLCs and energy meters. It reads machine tags on a configurable interval, buffers locally when the network drops, and posts normalized readings to ERPNext through the Frappe REST API using authenticated @frappe.whitelist() endpoints and a scoped API-key role profile.

Inside ERPNext we build new DocTypes — Machine, Machine Signal Map, and Machine Telemetry (plus a Downtime Log) — that map each raw tag to a Workstation and to the fields that matter: good count, scrap/reject count, cycle time, and machine state. hooks.py doc events and Server Scripts translate incoming telemetry into ERPNext transactions: a rising good-count edge increments the linked Job Card's completed quantity, scrap increments the reject quantity, and cycle time is written for downstream OEE. A scheduler_events job runs a state machine that detects when a machine sits in stop/idle beyond a threshold and opens a Downtime Log with a reason code, so unplanned downtime is captured the moment it happens, not reconstructed later. Client Scripts add live tiles and a color-coded machine board to the Workstation and Job Card views. OEE (Availability x Performance x Quality) is then computed from real, timestamped data rather than shift-end estimates.

Because this is build-to-order, nothing is a generic download. We start with a scoping call to inventory your machines, controllers, protocols, and the exact tags/registers to read, then confirm the signal-to-Workstation mapping and downtime reason taxonomy. We build against your ERPNext version (v15 or v16), validate the whole pipeline on a staging bench with a simulated or mirrored device feed, run UAT with your team, and only then install on production with a documented rollback. You receive the full source of your version, a git repository, and a post-go-live support window. Typical delivery is one working week from confirmed scope, depending on the number of machines and protocol mix.