MANUFACTURING - FULL

Full Track: Author Your Production Intelligence

You chose complete control. Build your production system from first principles.

This track assumes nothing. You design the data architecture, create the surface hierarchy, and author every query.

What you will build

Component	Scope
Connections	Every production data source
Surfaces	A plant, line, cell and station hierarchy
Warm queries	An OEE, quality and downtime library
Team	Access for operations, quality and maintenance
MCP	Any AI client reading the library

Phase 1: Design the production architecture

Step 1.1: Map your hierarchy

Plant
├── Line 1
│   ├── Cell A (Machining)
│   │   ├── Station 1
│   │   └── Station 2
│   └── Cell B (Assembly)
│       ├── Station 1
│       └── Station 2
├── Line 2
│   └── Same structure
└── Support
    ├── Quality Lab
    └── Maintenance

Step 1.2: Identify data sources

Level	Data source	Metrics	How it reaches OpenX
Station	PLC or sensor	Cycle time, state, counts	A connection, through an edge gateway
Cell	Cell controller	Throughput, work in progress	A connection
Line	MES	Yield, batch tracking	An API Source polling the MES
Plant	The queries above	Total output, efficiency	Warm queries over the rest

Step 1.3: Create connections

Ask Azi in the main panel:

"Create connections for the Line 1 PLC data and the cell controllers, and an API source for the MES."

She proposes them. Accept each one, then Deploy: connections and API sources are built only when someone deploys. Each connection's Connection Info tab then shows the credentials your gateway sends with.

Phase 2: Create the surface hierarchy

Step 2.1: Plant surface

"Create a plant-level surface fed by every line's connections."

Azi proposes a Plant-Production surface wired to those connections. Accept it; it is live at once.

Step 2.2: Line surfaces

Surface	Fed by	Purpose
Line-1	Cell A and Cell B connections	Line 1 detail
Line-2	Line 2 connections	Line 2 detail

Step 2.3: Cell and station surfaces

Surface	Fed by	Purpose
Line-1-Cell-A	Station 1 and 2 data	Machining detail
Line-1-Cell-B	Station 1 and 2 data	Assembly detail

Your surface hierarchy decides how finely your queries can break production down.

Phase 3: Build the query library

Query	Purpose	Level
plant-oee	OEE for the plant, by line	Plant
station-cycle	Station cycle analysis	Station
downtime-pareto	Top downtime causes	Plant / line
quality-fpy	First-pass yield	Line
lot-trace	Every step for one lot	Line

Step 3.1: Plant OEE by line

"Create a query for OEE across the plant over the last day, broken down by line."

Data
| where EnqueuedTime > ago(1d)
| extend Payload = todynamic(RawData)
| summarize AvailableSeconds = sumif(todouble(Payload.duration), tostring(Payload.status) != "Planned Downtime"),
            RunSeconds = sumif(todouble(Payload.duration), tostring(Payload.status) == "Running"),
            Units = sum(toint(Payload.units)),
            GoodUnits = sumif(toint(Payload.units), tostring(Payload.quality) == "Pass")
            by Line = tostring(Payload.line)
| extend IdealCycleSeconds = 30.0 // adjust per line
| extend Availability = RunSeconds / AvailableSeconds,
         Performance = (Units * IdealCycleSeconds) / RunSeconds,
         Quality = todouble(GoodUnits) / Units
| extend OEE = Availability * Performance * Quality * 100
| project Line,
          Availability = round(Availability * 100, 1),
          Performance = round(Performance * 100, 1),
          Quality = round(Quality * 100, 1),
          OEE = round(OEE, 1)

Step 3.2: Bottleneck stations

"Build a query that finds bottleneck stations by cycle-time variation."

Data
| where EnqueuedTime > ago(8h)
| extend Payload = todynamic(RawData)
| where isnotempty(Payload.cycleTime)
| summarize AvgCycle = avg(todouble(Payload.cycleTime)),
            StdDev = stdev(todouble(Payload.cycleTime)),
            Cycles = count()
            by Station = tostring(Payload.station)
| extend Variation = StdDev / AvgCycle
| order by Variation desc

Step 3.3: Trace one lot

"Create a query that lists every recorded step for a given lot number."

Data
| extend Payload = todynamic(RawData)
| where tostring(Payload.lotNumber) == "LOT-4582"
| project EnqueuedTime, DeviceID, Step = tostring(Payload.step), Station = tostring(Payload.station),
          Measurement = todouble(Payload.measurement), Status = tostring(Payload.status)
| order by EnqueuedTime asc

A warm query can declare parameters, so the lot number can be passed on each call instead of written in.

Phase 4: Give your team access

The owner adds teammates in Workspace -> Team Members: enter their email and click Invite. They are members at once. No AI can invite members, and the role picker in that window is not saved - what a member can change is set by access rights.

Team	What they need
Engineering	Workspace.Deploy to build connections; Workspace.Access.Manage to manage access
Operations, Quality	Membership: any member can save and accept queries; some changes need a management right (see Permissions)
Maintenance	Membership, to read the downtime surfaces and queries
Management dashboards	An API key from APIs -> API Keys, carrying only the rights the dashboard needs

Access rights apply to the whole workspace; there is no per-surface permission. See Permissions.

Phase 5: Connect any AI over MCP

The OpenX MCP Server is at https://openx.fathym.com/ox-api/mcp. Sign in through the browser: in Claude.ai or ChatGPT, add a custom connector and paste the URL; for coding agents, run fai mcp install https://openx.fathym.com/ox-api/mcp --name openx and sign in when the browser opens. Setup for each client is in Bring Any AI.

Then ask:

"Why did Line 1 miss its production target yesterday? Show me the contributing factors."

The AI reads your saved queries with call_api. If it wants to build something new, it proposes it with manage_proposals, and a person accepts.

What you built

Component	Count
Connections	3 or more
Surfaces	5 or more, plant to cell
Warm queries	5
Team access	Set
MCP clients	1 or more

Before and after

Before	After
A different OEE in every system	One definition, one number
Tribal knowledge about metrics	Saved queries anyone can read
Manual root-cause analysis	Queries that trace a lot step by step

Next steps

If you want to...	Go to...
Try the hybrid track	Hybrid track
See query patterns	Query Patterns
Connect more AI clients	Bring Any AI
Explore pharma	Pharma Factory