Work visually with industrial data
Organise work as Notebook → Section → Page. Add live blocks and typed relationships such as FEEDS, MEASURES and DEPENDS_ON.
CANVAS + GRAPH + CONTEXTUNS Navigator is the working layer above your namespace — a canvas where plant data becomes dashboards, alerts, forecasts and data products an AI can call.
NO CLOUD ACCOUNT · NO PHONE-HOME · READS YOUR PLANT, NEVER COMMANDS IT

The broker is up. Topics are structured. Data is flowing. Then the layer above the namespace turns into another custom build: a dashboard nobody maintains, a reporting project that takes a quarter, or a data lake that answers questions nobody asked.
The data is available and still nobody is using it, because available and usable are different problems.
Place a source, query, register, machine or document. Wire it. What you build is simultaneously a dashboard, a data-flow definition and the exact context an AI reasons over.
Organise work as Notebook → Section → Page. Add live blocks and typed relationships such as FEEDS, MEASURES and DEPENDS_ON.
CANVAS + GRAPH + CONTEXTRead MQTT, databases, Modbus, torque tools, REST endpoints, watched folders and serial machines through TCP bridges.
READ-ONLY CONNECTORSExpose a pipeline result over REST and, with one checkbox, as a scoped MCP tool an AI assistant can call directly.
ONE RESULT · TWO TRANSPORTSAn agent reads a block only when it is enabled, configured and in scope. Each gate can narrow context, never widen it. Reports identify the blocks used and cite them for factual claims.

Typed, versioned pipelines run on manual, scheduled, streaming or event triggers. Dry-run traces every step and stubs every sink. Exposed results carry provenance and honest freshness.

Anomaly detection reports the method, field, value and threshold that fired. Forecasting projects when a signal will cross a target — and refuses to invent a time when the trend is noisy, flat or heading the wrong way.

UNS Navigator is designed to be additive. It reads existing systems and keeps every deliberate egress path short, named and test-pinned.
One container, with optional TimescaleDB history and FalkorDB graph services. No cloud account, no phone-home and no network requirement for licensing.
Verified on x86-64 and ARM64, including single-board hardware suitable for a cabinet-level deployment.
No key, account or internet connection. If the trial ends, the workspace becomes read-only rather than bricking: pages, models and history remain viewable and exportable, and already-deployed apps keep serving.
No. It connects to yours. The built-in simulator can also generate realistic streams when no broker or plant connection is available.
No. It reads them and provides the working layer above. History is available as an opt-in TimescaleDB service per source.
No. Source connectors are read-only. The deliberate egress paths are named and opt-in: MQTT publish, its own history tables and outbound notification webhooks.
A block must be enabled, configured and inside the agent's scope. The report lists the exact blocks it used. Local inference through Ollama is supported.
An exposed pipeline result can become a read-only MCP tool, allowing an external AI assistant to query curated plant results without direct access to the source system.