One on-site AI layer for your equipment, systems and documents.
Your site’s WatchTower: ask a question in plain English and get the answer from your equipment, your systems and your own documents — on one screen, without leaving the floor.
The problem
Three problems on the floor.
What slows a warehouse down when a line stops.
01 · The problem today
Diagnosis takes hours
Schematics, fault history and SOPs sit in four systems and in the heads of senior techs.
02 · The problem today
Knowledge walks out the door
PLC drawings and recovery steps retire with the people who wrote them.
03 · The problem today
Incidents run over phones and text
No timeline, no post-mortem, and audit evidence assembled by hand.
How it works
Closed loop, from Andon call to runbook.
Agents that execute, not just recommend — every step logged, every answer cited.
01
Signal
A PLC alarm trips or an operator pulls the digital Andon cord. AndonX sees it the moment it happens.
02
Diagnose
Ask Andy in plain English. The live tags, the trend and the cited SOP step come back in seconds, not hours.
03
Act
The right SME is paged and the timeline records itself — no more chasing the fix over phones and text.
04
Learn
The fix becomes a runbook in the Wiki, so what your senior techs know stays on site.
Platform
Five modules. One on-site platform.
Andy, live PLC tags, incident management, the Wiki and forecasting share one data layer: your equipment, your systems and your own documents.
Andy, your Co-Pilot
Challenge
When a line stops, the answer is spread across the WMS, SCADA, the historian and the binder on the shelf.
Solution
Ask in plain English. Andy answers from your tags, systems and documents — with the source cited, never a made-up procedure.
Benefits
Answers in seconds, not screens
No query syntax or dashboards to learn
SOPs and runbooks answerable by anyone on shift, including new hires
Real-time PLC tags & alerting
Challenge
Equipment status is locked in the PLC, invisible to the WMS — and alarms turn into reactive firefighting.
Solution
Hundreds of tags stream live from Allen-Bradley, Siemens, CODESYS, Beckhoff or any OPC UA source, with units, descriptions and alarm status added.
Benefits
Find tags by meaning, not by tag name
Monitor live data from a phone or tablet
Alarms page the on-call team by SMS, phone or email via PagerDuty
Incident management & RCA
Challenge
Incident leads juggle calls and chats while the line is down, and small blips never get logged.
Solution
A digital Andon cord: AndonX opens the incident, pages the right SMEs by type and priority, and builds the timeline from chats, decisions and system events.
Benefits
From outage to operational in minutes, not hours
Objective post-mortems and a searchable incident history
Stakeholder updates on a set cadence — no communication silos
A wiki tied to your assets
Challenge
SOPs, OEM manuals and schematics sit on shared drives — out of date and hard to find on the floor.
Solution
Structured Spaces for SOPs, manuals, PLC and electrical schematics and recovery runbooks, co-edited in real time with full history.
Benefits
Find I/O points and wiring references in seconds
New technicians onboard from curated handbooks
Self-hosted and air-gap capable, with role-based access
Module 05
Forecast agent
Predicts stockout and replenishment risk from live inventory, work-order and machine data.
Key benefits
Fewer stockouts and less overstock — live risk probabilities with confidence intervals.
Urgency-ranked replenishment — continuous reorder recommendations by criticality.
Grounded in your data — live machine data, work orders and SOPs.
Conversational — ask predictive questions without SQL or a data team.
Use cases
Integrates with Manhattan, SAP EWM, Oracle and other WMS or ERP systems via MCP.
Check stock, lead time and substitutes for a part, then raise a PO in the same thread.
Integrations
Plugs into what you already run. No rip-and-replace.
AndonX integrates with your existing equipment and systems. Listed vendors are compatible technologies, not partners.
IoT & edge
MQTT
Ignition Edge
Sensors & gateways
RFID & barcode
OT — control systems
Allen-Bradley
Siemens
CODESYS
Beckhoff
OPC UA
SCADA
Historians
IT — business systems
SAP EWM
Manhattan
Oracle
ERP
WMS / WES
MES
OEM equipment
AGVs
AMRs
AS/RS
Palletizers
Robot mgmt servers
Also
PagerDuty alerting · REST APIs and MCP adapters for new agents · Databricks and Snowflake export (coming soon)
Reference architecture
Your data never leaves your building.
AndonX runs on one industrial server with a GPU inside your network. Here’s the reference architecture for your IT and OT teams.
Field
PLCs & RTUs
Allen-Bradley, Siemens, CODESYS and Beckhoff — through a transmission module or edge gateway.
Edge
Ignition Edge · MQTT
Device data published to an MQTT server on the site network.
AndonX server
Industrial PC + NVIDIA GPU
Models, telemetry, the wiki and every agent run here.
Systems & equipment
WMS · ERP · RMS · AGVs
Line-of-business systems and OEM equipment through their own APIs.
Clients: unlimited web-browser sessions, plus phones and tablets on the site network.
Data provenance
Every PLC tag, transaction and document stays on the on-prem server — no production, IP or safety data leaves the plant network.
Low latency
Inference runs on the local GPU, so answers and alerts return in seconds and keep working if the internet link drops.
No per-user or per-question billing
Models run on hardware inside your building, so usage stays unmetered as adoption grows.
Who it’s for
Built for operations where a stopped line has a cost.
Warehouse and distribution first, where every minute of downtime hits SLAs, overtime and client trust.
01
Third-party logistics (3PL)
Multi-client sites, thin margins and contractual SLAs.
02
E-commerce fulfillment
Peak-season throughput and dense automation.
03
Food & beverage distribution
Cold-chain excursions and traceability pressure.
04
Automotive & pharma distribution
High line-stop cost and strict audit requirements.
Also running on manufacturing lines with the same PLC-driven equipment.
Use cases
What it looks like on a real shift.
Four situations site teams bring to AndonX.
A conveyor sorter faults at 2 a.m.
Challenge: the night shift has no controls engineer on site. With AndonX: Andy reads the drive alarm and tag history, cites the recovery SOP and pages the on-call engineer with the timeline attached.
A cold-chain excursion in a food distribution center
Challenge: temperature drifts in a freezer zone and nobody notices until product is at risk. With AndonX: the excursion triggers an alert, the incident opens with live readings and the recovery runbook, and the timeline becomes audit evidence.
A new technician needs the lockout procedure
Challenge: the dock leveler LOTO steps live in a binder across the building. With AndonX: “@Docs what’s the lockout procedure for the dock leveler?” returns the current procedure with cited steps — nothing invented.
Stockout risk ahead of peak season
Challenge: replenishment runs on weekly manual audits. With AndonX: the forecast agent ranks at-risk SKUs by urgency from live WMS and ERP data, and planners act from the same conversation.
For IT & controls
Technical questions
Security, deployment and maintenance — what technical buyers ask first.
How does AndonX connect to our ControlLogix or CompactLogix setup?
Through native drivers or an edge gateway such as Ignition Edge publishing tags over MQTT. Tags are read, never rewritten, and nothing changes in your PLC programs.
What is the security model?
Everything runs on a server inside your network. Access uses authentication and role-based permissions, guardrails filter content, and no data is sent to a vendor cloud.
Does it keep working if the internet goes down?
Yes. Inference runs on the local GPU, so answers, alerts and the wiki stay available on the site network.
Who maintains it when the AndonX team isn’t on site?
Remote monitoring and support are included for the life of the system. Execution logs, approvals and monitoring give your team full visibility into what agents do.
Next step
Book a 90-day pilot at one site.
One site, one agreed success metric, fixed scope and fixed fee. We measure downtime and time-to-repair against your baseline — then you decide.