Operational data foundation
Connects operational, enterprise, event, document, and application context into an agent-ready operating foundation.
Reference architecture for governed agentic operations.
XMPro provides the industrial operating layer for AI agents, applications, recommendations, workflows, and governed autonomous operations. This reference architecture shows how that operating layer connects operational data, enterprise context, applications, agents, governance, and decision accountability into one deployable architecture.
One governed foundation for people, applications, and agents.
The XMPro Agentic Operations Platform sits between operational systems and operational action. It connects data sources, contextualises the operation, governs decisions, and serves both human applications and MAGS-powered agents from the same operating foundation.
XMPro AO Platform
DataStreams orchestrates the operational flow.
OCE enriches the flow with context, identity, trust, governance, and evidence.
Governance controls when recommendations become actions.
Systems of record remain authoritative. XMPro connects and empowers.
Operational & Enterprise Sources
SCADA / PLC / DCS
Historians
MES / EAM / ERP
Work orders
Documents / procedures
UNS / MQTT / Kafka namespaces
AnyLog (where deployed)
Enterprise data platforms
Your existing systems
edge data provider / reads, queries, subscribes
XMPro DataStreams
Data acquisition, normalization, event processing, orchestration — the ingestion and normalization fan-in layer.
Listener agents
Context providers
Action agents
Analysis agents
Recommendation components
Connectors and adapters
Event processing logic
Orchestrates the operational flow
Operational Use-Case Pipelines
Configured for each AO solution
Monitor
Detect
Analyze
Recommend
Coordinate
Simulate
Act
context · identity · trust · governance · evidence
Operational Context Engine (OCE)
Recommended enrichment layer for governed human and agent decisions. Critical for MAGS agents and higher autonomy, because agents need explicit context humans often carry tacitly.
Context
Identity
Trust
Governance
Evidence
AppDesigner / Human Applications
Dashboards, workflows, decision review.
MAGS Agents
Reasoning, recommendations, bounded action.
Front-Running Simulations (FRS)
Tests candidate actions before execution — with live data and context.
Governed Outputs (delivered to humans and systems)
Human decision support
Agent recommendations
Approved action packages
Work orders / workflow actions
Evidence and audit trail
Simulation-backed recommendations
Primary operational flow
Context enrichment flow (OCE)
Simulation flow (FRS)
OCE is a context and governance service. It does not replace systems of record or operational applications.
The same platform pillars, focused on industrial action.
XMPRO CAPABILITY COVERAGE
Foundation
Real-world operating model
Operational Knowledge Graph models assets, sensors, processes, relationships, policies, enterprise context, identity, trust, and provenance.
Application
Application & workflow layer
Supports operational applications, recommendations, workflows, approvals, and decision review experiences.
AI & agent layer
Gives MAGS-powered agents governed access to operational context, recommendations, and decision pathways.
Governance
Governance & control
Evaluates actions through risk, safety locks, source confidence, policies, approvals, evidence packs, and audit records.
Decision accountability
Decision Trace records what was recommended or done, why, what evidence was used, who or what approved it, and what outcome followed.
Reach
Integration & deployment
Designed for brownfield industrial systems, edge/local patterns, enterprise integration, and cross-site scaling.
Portability & standards
Uses open semantic, validation, query, provenance, and industrial standards where appropriate to avoid trapping operational knowledge in a closed model.
XMPro is not only an application layer on top of someone else’s operating model. It provides the operating model, context layer, application layer, agent layer, governance layer, and decision accountability layer required for agentic operations.
CORE OPERATING MODEL
From data to governed action.
The platform is organised around a simple operating loop. The same five steps run whether a human, an agent, or a policy-controlled action does the work.
The architecture is grounded in action, not just data readiness. It exists to help industrial organisations operate better — not to build a more elegant data model.
Every agentic action follows a governed path.
Eight stops between observation and recorded outcome.
AGENTIC DECISION PATH
STEP 01
Observe
Agent or application receives operational context.
STEP 02
Interpret
Asset, process, identity, and enterprise context resolved.
STEP 03
Recommend
A recommendation or action is proposed.
STEP 04
Govern
Risk, policy, source confidence, safety locks, approvals checked.
STEP 08
Learn
Outcomes and feedback improve future recommendations and operating context.
STEP 07
Record
Decision Trace captures what happened, why, who approved, and what evidence supports it.
STEP 06
Act
The approved action is executed or handed off.
STEP 05
Coordinate
Decision routes to a human, workflow, agent team, or policy execution path.
CONTROL MODES
Human-Controlled
Agents recommend. A human decides, plans, and acts.
Human-Approved
Agents prepare and coordinate. A human approves execution.
Policy-Controlled
Agents execute within governed policy limits and escalate exceptions.
Connect existing systems without forcing a full replacement.
XMPro is designed for brownfield industrial environments where critical context is spread across many systems.
- OT & IT data sources
- Live operational streams
- Event & alert data
- Enterprise records
- Documents & procedures
- Operational applications
- AI-accessible context
XMPro doesn’t require every system to become the system of record. It creates an operating foundation that can connect, contextualise, govern, and act across the systems customers already run.
Designed for controlled operation, not unmanaged AI action.
Governance is part of the architecture — not a bolt-on control layer.
Human & specialist overrides
Policy-controlled action boundaries
Approval routing
Evidence packs
Decision provenance
Data lineage
Security & access controls
Audit & compliance reporting
The platform earns the right to support autonomy by making decisions governable, explainable, and reviewable.
Patterns for industrial environments.
Four supported patterns. Pick against connectivity, sovereignty, latency, and governance — deeper detail on the dedicated page.
Cloud
Enterprise-scale analytics, governance, application access, and centralised coordination.
Edge / Local
Site-local context, constrained connectivity, low-latency needs, operational continuity.
Hybrid
Central governance with site-level execution.
Customer-Controlled
Regulated, air-gapped, or sovereignty-sensitive environments.
Operational knowledge as a durable enterprise asset.
Open standards for semantic modelling, validation, query, provenance, industrial taxonomies, and AI access — so operational knowledge doesn’t become a proprietary dead end.
OWL 2
Web Ontology Language — semantic modelling.
SHACL
Validation rules for the operating model.
SPARQL
Cross-domain query across the knowledge graph.
PROV-O
Provenance and decision lineage.
MCP
Model Context Protocol — agent-accessible context.
ISO 14224
Industrial reliability and maintenance taxonomy.
ISA-95
Enterprise–control system integration model.
OPC-UA
Industrial automation interoperability.
Architecture proof for technical evaluators.
Starting points for technical evaluation. Need a downloadable architecture brief, a decision-replay walkthrough, or a worked cross-domain query? Talk to a solution architect.
Live product documentation
Technical reference for installation, deployment, configuration, and component behaviour.
OpenMulti-Agent (MAGS)
XMPro’s public repository for generative AI agents and Multi-Agents — the agent layer in code.
OpenBlueprints & Accelerators
Reference implementations and architecture patterns your team can read, adapt, and reuse.
OpenHelm charts
Kubernetes deployment artefacts — concrete starting point for k8s evaluators and platform teams.
OpenDTDL asset models
Real industry asset modelling — Wind Power Plant & Wind Turbine DTDL models you can inspect.
Open