Make to stock
The everyday order. Built against a forecast or a planner's demand, then released to free stock.
One lifecycle
MFG - production system of record
MFG defines one controlled production record connecting demand, materials, capacity and shop-floor evidence to work orders, WIP, quality-gated completion and auditable decisions.
production order
release readiness

Production capabilities
Connect planning, readiness, execution, quality and completion without splitting the production truth across disconnected systems.
Orders
Every order runs one lifecycle, from draft to closed. Not a status field people edit.
Recipe
Releasing an order freezes its BOM and routing. You always know how a batch was built.
Readiness
Material, capacity and asset checks run before work reaches the floor.
Execution
Operators log confirmations, moves, yield, scrap, rework and downtime per operation.
Quality
A failed mandatory check blocks completion. Only an inspector can clear it.
Handoff
Declare output by lot, batch or serial. WMS receives it, FMS values it.
The production lifecycle
An order moves through one sequence. Each step adds to the record instead of overwriting a status, so you can always see what happened and when.
01
Planned
Version + readiness checked
02
Released
Planner releases the order
03
Staged
Materials staged from WMS
04
In progress
Operators confirm on the floor
05
Completed
Output declared, quality-gated
Production modes
You pick the mode when you create the order, and it never changes. Every mode runs the same lifecycle, the same readiness checks and the same material and quality steps. The mode just adds the extra checks that kind of work needs.
The everyday order. Built against a forecast or a planner's demand, then released to free stock.
One lifecycle
Records which component went into which unit. A component serial can only be installed in one place, and you cancel a mistake with a removal that points at it.
Coverage checked before receipt
Raised against an open defect on another order, and capped at the units still unresolved. The original order cannot close while its rework is open.
Two planners cannot split one defect
Names a serial an earlier order already made, and returns that same serial. One live repair per item and serial.
Quantity is always one
One order is one batch, named by batch reference instead of serial. A weak ingredient lot gets charged at a higher weight, and a yield outside its limits needs a separate review.
Potency and yield checks
Workbenches
A planner, a supervisor, an operator and an inspector all need different things. So they get different screens, each limited to what that person is allowed to do. What you can see and what you can change are the same list.
Your queue, your priorities, and whether everything you need is connected.
Demand, coverage, draft plans, readiness, release and the exceptions that need a decision.
What is running late, what the projected dates are, and which work centres are overbooked.
Released orders, holds, progress, yield and scrap, plus WIP and downtime per operation.
The work assigned to you. Confirmations, movements, scrap, rework, time and notes.
The hold queue, inspections, defects, what to do about them, and inspector release.
What is needed, what the warehouse has staged, what is short, and what was substituted.
Look up any order. See why it changed, what went into it, and where the output went.
The problem it targets

Readiness controls

Finite-capacity scheduling

Sustainability evidence
We record energy, waste, emissions and water against the order that used them. A process batch counts as an order too, so every batch gets its own figures.
Unless it is an estimate, every number names the meter, bill or document behind it. A calculated number also shows the factor used, so you can judge it yourself. We do not ship a factor library, follow a reporting standard, or certify anything. We record what was measured and add it up.
Nobody can type in a number and call it a meter reading. Real readings come through a separate door that only an approved system can use, and the same reading never lands twice.
Everything converts to one unit before it is added up. A bill in MWh and a meter in kWh still agree, and litres of water never get added to kilos of steam.
MFG in the production stack
MFG connects demand, material, capacity, execution and quality. Every other system keeps owning what it already owns.
Architecture
The architecture combines a controlled make-to-stock lifecycle, a governed agent surface and clear ownership boundaries across the Splenta family.
The make-to-stock flow
A production order begins with demand and moves through material, capacity and quality controls. Each step writes append-only evidence, with corrections represented as reversals, adjustments or rework loops rather than silent edits.
Demand picks a production version. That is a BOM snapshot plus a routing snapshot, with the operations, work centres and quality checks it needs.
ProductionVersion
Material readiness reads warehouse stock and incoming supply. Capacity reads work-centre calendars and asset availability. Shortages show up first.
MaterialReservationRequest
A planner releases once readiness passes, or records an approved shortage reason. Release locks in the inspection plans in force that day.
ProductionOrder
Operators confirm start, stop and partial completion. Moves, labour and machine time, yield, scrap, rework and downtime are all logged per operation.
OperationConfirmation
Inspection records its measurements, then releases, holds or sends the work back. A mandatory step stays blocked until an inspector clears it.
QualityCheckpoint
Declare output by lot, batch or serial. The handoff freezes what went into it, and the warehouse confirms receipt. Nothing closes until every unit arrives.
OutputDeclaration
production-order lifecycle
The dashed state is the exception branch - work could be held for quality, material, safety, maintenance, or engineering reasons and resumed by the owning function. Cancelled, scrapped, split, merged and reworked provide explicit paths for other outcomes.
The Arin-over-MCP surface
Arin reads production readiness, timelines, quality holds, shortages and variance evidence, explains the exception and prepares an approval-bound action. MFG validates policy and human approval before executing any change.
mfg:schedule:read
mfg:schedule:write
mfg:operation:write
mfg:material:write
mfg:assembly:write
mfg:process:approve
mfg:quality:write
Every mutating tool declares its required mfg:* scope, carries an idempotency key, and high-impact actions require an approval state and a reason. Reads and writes share one service layer behind REST, the UI and MCP.
The family boundary
MFG owns the production record without replacing WMS, SCM, FMS, EAM, PLM, an external MES or Arin. Systems exchange information through REST, MCP and events, with no shared database tables.
Identity and tokens for the family.
MFG checks every token ICE issues. It never issues one.
Agent and decision-support layer.
Reads MFG facts and prepares actions for approval. It never owns the record.
Procurement, suppliers, purchase orders.
Tells MFG what supply is coming and when it slips. MFG owns the effect on production.
Physical stock and locations.
Reserves, stages and issues material, then receives finished goods.
Asset registry and maintenance.
Tells MFG what is available and what is down. MFG owns the production impact.
Demand and financial truth.
A customer order arrives from RMS. Costing goes to FMS. MFG calls neither directly.
the family boundary
Arin explains, classifies, recommends and prepares actions. MFG validates the action, enforce approval and shop-floor controls, and execute through its own services. This is the intended arrangement, not running software.
Shape the concept
If demand, materials, capacity and shop-floor evidence are difficult to reconcile, tell us how your production stack works today. Your operating priorities will help shape the first MFG implementation.