What Is KŌJŌ Stack
The data plane between OT and cloud
An industrial data plane is the layer where data from physical systems is acquired, structured, and prepared before it reaches any downstream system. This layer does not exist in traditional industrial architectures. KŌJŌ Stack establishes it at the first mile.
KŌJŌ Stack sits between operational systems-PLCs, SCADA, sensors, historians-and enterprise/cloud consumers. It ingests raw industrial data, applies structure and context, and routes it deterministically. Without this layer, every downstream system must independently reconstruct data from raw, protocol-bound sources.
Protocol-Native Ingestion
OPC UA, OPC DA, Modbus, S7, BACnet, DNP3, Sparkplug B, MQTT, Kafka-ingested natively at the source, not through translation layers.
Real-Time Transformation
CEL expressions, report-by-exception filtering, and unit conversion-applied deterministically at the edge before data leaves.
Unified Namespace Alignment
ISA-95 compliant hierarchy structures all data into a canonical namespace. Every tag carries context, provenance, and quality.
Reliable Data Delivery
Durable local buffering, ordered replay on reconnect, and guaranteed delivery to cloud, lakehouse, and streaming targets.
The First-Mile Data Problem
You cannot scale AI or analytics without solving this layer
Every enterprise investing in industrial analytics, predictive maintenance, or operational AI hits the same wall: the data between machines and cloud systems is fragmented, unstructured, and unreliable. This is the first-mile data problem.
Protocol Fragmentation at the Edge
OPC UA, OPC DA, Modbus, S7, BACnet, MQTT-each with different transports, semantics, and timing models. Point-to-point integrations multiply with every new system.
No Deterministic Data Pipelines
Without bounded-latency processing at the edge, data arrives late, out of order, or not at all. Polling-based collection misses critical state transitions.
Missing Context at the Source
Raw tag values without hierarchy, provenance, or quality metadata are meaningless to downstream consumers. Context must be applied where data originates-not reconstructed later.
Cloud-First Architectures Failing OT
Sending raw telemetry to cloud platforms ignores bandwidth constraints, latency requirements, and the reality that OT networks cannot depend on internet connectivity.
No amount of cloud compute compensates for missing, late, or context-free data.
The first-mile data plane is the prerequisite for every downstream system-analytics, AI, digital twins, and operational intelligence.
The Solution
A deterministic edge data plane
KŌJŌ Stack is built on four pillars that define the first-mile data plane: ingest natively, transform deterministically, structure canonically, and deliver reliably.
Protocol-Native Ingestion
Each protocol adapter speaks the native language of the device. No translation gateways. No protocol converters. Data is ingested at the source with full fidelity.
- OPC UA with subscription mode and security policies
- OPC DA for legacy Windows-based SCADA and DCS systems
- Modbus TCP/RTU for legacy PLC communication
- Siemens S7 native protocol-no OPC server required
- BACnet IP with Change-of-Value subscriptions
- EtherNet/IP (CIP) for Rockwell ecosystems
- Sparkplug B - 100% spec-compliant
- DNP3 (IEEE 1815) for utility SCADA
- MQTT and Kafka ingress for message brokers
Real-Time Transformation
Transformations execute at the edge with predictable latency. Filter noise, convert units, compute derived values, and enrich context-before data leaves the plant.
- CEL expression engine for deterministic transforms
- Report-by-Exception (RBE) deadband filtering
- Unit conversion and scaling at the edge
- Derived value computation
- Quality indicators and metadata enrichment
- Bounded-latency pipeline execution
Unified Namespace Alignment
All data is structured into an ISA-95 compliant namespace. Tags carry identity, timestamp, quality, and source context. The namespace is the contract between OT and IT.
- ISA-95 hierarchy: Enterprise → Site → Area → Line → Cell
- Automatic topic generation from hierarchy
- Canonical schema with provenance metadata
- Decoupled producers and consumers
- Single source of truth for all operational data
- Protocol-agnostic addressing for downstream systems
Reliable Delivery with Buffering
Data is buffered locally before transmission. Network outages, cloud unavailability, and destination failures do not cause data loss. Replay ensures completeness.
- Durable local buffering for zero data loss
- Ordered replay on network reconnection
- Configurable retention policies
- Multi-destination fan-out
- S3, GCS, Kafka, MQTT, TimescaleDB, InfluxDB, Iceberg targets
- Durable delivery with buffering and replay
Architecture
Where KŌJŌ Stack sits in your stack
The first-mile data plane between operational technology and every downstream consumer-analytics, AI, historians, and cloud platforms.
OT Systems
OPC UA
Subscriptions + Browse
OPC DA
Legacy SCADA / DCS
Modbus TCP/RTU
Register polling
Siemens S7
Native protocol
BACnet IP
COV subscriptions
EtherNet/IP
CIP protocol
DNP3
Utility SCADA
MQTT / Kafka
Broker ingress
KŌJŌ Stack
Protocol-Native Ingestion
CEL Transforms & RBE Filtering
UNS / ISA-95 Contextualization
Durable Buffering & Replay
Transform · Buffer · UNS
Cloud / AI / Historians
S3 Tables / Iceberg
Lakehouse
Apache Kafka
Event streaming
TimescaleDB
Time-series historian
MQTT Brokers
Cloud IoT
AWS IoT Core
Cloud services
S3 Data Lake
Batch export
Google Cloud Storage
BigQuery compatible
InfluxDB
Time-series historian
Why KŌJŌ Stack
Purpose-built where general tools fall short
Dashboards, middleware, and cloud IoT platforms were built for other problems, then stretched to fit the first mile. KŌJŌ Stack was engineered for it from the ground up-protocol-native, edge-resident, and deterministic.
KŌJŌ Stack is not a dashboard or SCADA replacement. It operates below the visualization layer-structuring and routing data that dashboards consume.
Event-driven pipelines with bounded latency, executing transformations at the edge where data originates.
Traditional middleware passes data through without understanding it. It cannot filter, transform, or contextualize industrial telemetry at wire speed.
Every data point is ingested natively, enriched with ISA-95 context, filtered for significance, and routed to the correct destination.
Cloud IoT platforms assume reliable connectivity, unlimited bandwidth, and centralized processing. OT networks have none of these.
Processing, buffering, and decision-making happen at the edge. Cloud is a destination, not a dependency. Operations continue without connectivity.
Platforms that try to do everything-device management, rules engines, dashboards, app hosting-do none of it well for industrial use cases.
One job, done right: own the data plane between machines and every system that needs their data. Protocol-native, deterministic, industrial-grade.
Architectural Differentiators
What makes KŌJŌ Stack fundamentally different
These are not features-they are the core responsibilities of an industrial data plane. Industrial systems do not produce usable data by default. Data must be structured at the source. These responsibilities define how.
First-Mile Data Ownership
Data is structured and prepared at the point of origin-before any downstream system touches it. No reconstruction required.
Deterministic Edge Execution
Pipelines execute with bounded latency and predictable sequencing. Events are delivered in a consistent, deterministic order.
Protocol-Native Ingestion
OPC UA, OPC DA, Modbus, S7, BACnet, EtherNet/IP, DNP3, Sparkplug B, MQTT, Kafka-each ingested natively. No translation gateways.
Contextualization at the Source
ISA-95 hierarchy, canonical schema, and provenance metadata are applied at ingestion-not reconstructed downstream.
Edge-Native Processing & Reduction
CEL transforms, RBE filtering, and unit conversion execute at the edge. 90%+ volume reduction while preserving every meaningful state transition.
Composable Module System
External modules extend the data plane over typed, versioned interfaces. Deploy, update, and remove protocol adapters or custom transforms independently.
Co-Located Edge Execution
Workloads execute alongside data pipelines in a shared runtime at the edge. Compute and data are unified-no movement to external systems for processing.
Data Plane Observability
Per-pipeline throughput, latency percentiles, error rates, and backpressure metrics. Prometheus-compatible. Hash-chained audit trail.
Enabling Physical AI
The data foundation AI systems require
Predictive maintenance, process optimization, and autonomous operations are impossible without clean, contextualized, low-latency industrial data. The first-mile data plane is the prerequisite for Physical AI.
Clean, Contextualized Data
AI models require structured, timestamped, quality-annotated data. KŌJŌ Stack delivers data that is ready for feature engineering and model training-not raw telemetry dumps.
Low-Latency Edge Inference
Deterministic pipelines with bounded latency enable real-time inference at the edge. Predictive maintenance, anomaly detection, and quality control operate on live data.
Deterministic Replay for Validation
Replay historical scenarios through the same pipeline to validate AI agent behavior. Test against known outcomes before deploying to production operations.
Policy-Scoped Agent Access
AI agents interact with operational data through structured, auditable interfaces. Namespace-level policies define what agents can observe, query, and request.
Cross-Vertical Data Plane
One architecture, every vertical
The same first-mile data plane deploys across manufacturing, energy, automotive, and process industries. Different protocols and domains-identical architecture.
Manufacturing
Ingest data from PLCs, CNC machines, and quality systems. Normalize across protocols, apply ISA-95 context, and route to historians, lakehouses, and analytics platforms.
High-throughput ingestion
Automotive
Unify data from welding robots, paint systems, and assembly PLCs into a single namespace. Enable line-level analytics without point-to-point integrations.
Deterministic edge execution
Energy & Utilities
Collect from substations, renewable assets, and grid infrastructure via Modbus, DNP3, OPC UA, and OPC DA. Buffer locally for unreliable WAN connectivity.
Durable delivery over constrained networks
Food & Beverage
Capture batch parameters, CIP cycles, and environmental data with full provenance. Route to compliance systems with deterministic, auditable pipelines.
Auditable, deterministic pipelines
Own Your First-Mile Data
Deploy the data plane
your operations need
KŌJŌ Stack deploys at the edge, alongside your operational systems. Ingest natively, transform deterministically, and route reliably-from day one.