The governed intelligence node at the plant floor.
Industrial AI needs more than a brain. It needs arms and legs -- a governed unit that can act at the OT boundary, through OT systems, within OT safety constraints. That is IndustrialClaw.
IndustrialClaw is the OT deployment module for XMPro MAGS -- installed at the ISA-95 Level 2/3 boundary, pre-wired for OT systems, hardened by default, operating with a blast radius of zero.
Not a platform you configure for industry. An appliance that arrives ready for it.
Three things, understood clearly.
IndustrialClaw is often described in terms of what it replaces or competes with. It is better understood on its own terms.
A self-contained, deployable unit
Like a historian server or a SCADA collector, IndustrialClaw is deployed at the process interface -- not centrally. One node per production unit, compressor train, or functional zone. It installs into your network, connects to your systems, and operates within a declared IEC 62443 security zone.
The OT module for XMPro MAGS
MAGS is the cognitive layer -- the brain that reasons, orchestrates, and sets objective functions. IndustrialClaw is the module that gives MAGS arms and legs in the physical world. MAGS without IndustrialClaw can reason about OT. With it, MAGS can act in OT -- safely.
The only governed path to the plant floor
Enterprise agent frameworks have no governed path to OT systems. Any action that needs to reach the physical process -- from any agent, on any platform -- is routed through IndustrialClaw. That boundary is not a limitation. It is the architectural guarantee that keeps the plant safe.
Three operating modes. Matched to your control environment.
IndustrialClaw does not require a full SDA environment to deploy. You start where your infrastructure is, and the action surface expands as your OT environment evolves.
Correlates alarms, creates work orders, notifies operators, coordinates across systems. IndustrialClaw reads from historians, SCADA, and CMMS -- and writes back to work order systems and notification channels. The PLC never knows IndustrialClaw exists. This is available from day one, in any OT environment.
Proposes setpoint changes through normal SCADA and DCS supervisory paths -- exactly as a human operator would. Every proposal is validated against the symbolic governance layer before submission. Human approval remains in the loop at whatever autonomy level your organisation has declared.
Proposes and deploys certified control logic changes from the approved function block library. Every proposal is hash-verified and governance-validated before execution. In Software Defined Automation environments -- IEC 61499 function blocks, OPC-UA, containerised control -- IndustrialClaw can reconfigure the control layer itself, through supervised paths, with an immutable audit trail above every action. SDA creates the execution surface. IndustrialClaw is the intelligence that uses it.
SDA is not a prerequisite for starting. Modes 1 and 2 are available in any OT environment from deployment. Mode 3 becomes available as your infrastructure migrates toward Software Defined Automation. The governance architecture is in place from day one -- the action surface expands over time.
For your stack. For your operations. For your team.
IndustrialClaw does not replace what you have. It adds the governed intelligence layer that was always missing.
XMPro MAGS -- the brain
Multi-agent orchestration, objective functions, fleet-wide governance. Coordinates across all nodes and facilities. Can activate enterprise agent frameworks as subagents for IT-side tasks. Does not touch OT directly.
IndustrialClaw -- the arms and legs
The governed OT deployment module. One node per zone. Connects to your historians, SCADA, DCS, and SDA runtimes. Executes within declared boundaries. The mandatory gateway between the cognitive layer and the physical process.
Your existing OT infrastructure
DCS, SCADA, PLCs, historians, SDA runtimes, Safety Instrumented Systems. Unchanged. IndustrialClaw acts through your existing supervisory paths -- never around them.
Agents that act on exceptions at 2am without waking anyone -- unless the exception requires a human decision, in which case it escalates with full context already gathered.
Every agent action goes through supervised paths. No direct agent-to-actuator path exists. The control layer behaves exactly as it did before IndustrialClaw -- it just has a governed intelligence layer above it now.
Default-deny network posture. OT-native connectors pre-wired. Agent scope declared at deployment. Immutable audit trail. IEC 62443 zone architecture respected. Nothing in the agent layer can self-promote its own authority.
Start at Mode 1 with a single use case. Expand to Mode 2 as operational trust is earned. Mode 3 becomes available as your SDA infrastructure matures. The governance architecture is in place from day one -- it does not need to be retrofitted as autonomy grows.
Clarity on boundaries matters as much as the definition.
MAGS is the cognitive layer -- the brain. IndustrialClaw is the OT deployment module that gives MAGS arms and legs at the plant floor. MAGS without IndustrialClaw can reason about OT but cannot act in it. Together they are complete.
General platforms arrive open and require industrial hardening through configuration. IndustrialClaw arrives hardened. Default-deny is not a setting you enable -- it is the starting state. OT-native connectors are pre-wired at deployment, not bolted on.
IndustrialClaw operates above and alongside your control infrastructure, communicating through it via supervised paths -- exactly as a human operator would. Safety Instrumented Systems remain the authority for safety-critical actions. IndustrialClaw does not cross that boundary.
IndustrialClaw does not wait to be asked. It operates autonomously within its declared scope -- observing, reasoning, proposing, validating, and acting continuously. When a human decision is required, it escalates with full context. It does not generate a summary and wait.
Designed from the DTC Industrial AI Agent Manifesto.
The ten-law governance framework for AI agents in safety-critical industrial operations, developed with and distributed by the Digital Twin Consortium. IndustrialClaw was not aligned to the Manifesto after the fact. The Manifesto describes what an industrial agent must be. IndustrialClaw is the implementation of that standard.
Law 1
Deterministic Validation
Law 2
Physics-Aware Intelligence
Law 3
Symbolic Primacy
Law 4
Separation of Control
Law 5
Kill Switch and Override
Law 6
OT Protocol Fluency
Law 7
Auditability
Law 8
Progressive Autonomy
Law 9
Multi-Agent Safety
Law 10
Safe Continuous Learning
Ready to deploy the node?
We are onboarding a limited cohort of enterprise operators in 2026. Applications are reviewed -- acceptance is at XMPro's discretion. We prioritise heavy asset and mission-critical organisations.