About Streamsight

The diagrams you already build, made into context for an agent.

Industrial facilities run on visualization. Engineers at refineries, chemical plants, pulp and paper mills, mines, and pharmaceutical manufacturers use tools like PI Vision to understand and monitor their processes, and over time they build hundreds of views that map equipment, relationships, and live process data across a plant. Streamsight turns that visual context into a machine-readable process graph: a digital representation of how the plant actually operates. Then it operationalizes that knowledge layer with an engineering agent, bringing the AI-driven productivity gains seen in software development to industrial operations.

The difference from the usual industrial-AI playbook is where the context comes from. Instead of standing up a separate data platform and running a long project to mirror your plant inside it, Streamsight reads the operational views engineers already build and maintain. There is no separate source of truth to keep in sync: the same diagram operators read is the context the agent reasons over.

What the agent does

Describe a symptom in plain language and the agent investigates. It traces the dependency graph to find what feeds the trouble, pulls the relevant tag history, and runs the analysis in a real Python sandbox: correlations, regressions, change-point detection, whatever the question needs. When it reaches a conclusion, it builds a live dashboard, KPI tiles and trend charts wired to your tags, so the finding is a view you keep rather than a paragraph you forget.

It can also extend the model itself: declare a new tag, add a piece of equipment, drop a node onto a diagram, write a status rule. Every one of those writes is a proposal you review and approve in the UI. For multi-step changes the agent outlines a plan first and waits for your go-ahead. Corrections you give it persist across conversations, so a lesson learned once surfaces again when it's relevant.

A model worth reasoning over

The agent is only as good as the model beneath it. You build a diagram out of nodes and edges. Nodes can represent steps, equipment, services, or sub-systems; edges describe how they connect. Each node holds data items: a tag from your historian, a derived expression over other values, or a synthetic source for development. Status profiles map those values to color based on thresholds, so the whole flow communicates state at a glance, and when a single symbol represents something more complex, that node can open into its own sub-diagram.

Every node, edge, binding, and sub-diagram is structured information, a graph of your system, not just pixels. That's what makes the agent's reasoning possible: it queries real structure to know what's upstream and downstream and how operational components relate, rather than guessing from a picture.

An open platform

Everything the in-product agent can do is a typed tool in a shared catalog, and that catalog is published over MCP. The agent runtime is pluggable, so you can bring your own model, or attach an external agent such as Claude Code or Cursor. Whoever is driving gets the same governed tools and the same propose-then-approve safety.

Who it's for

Engineers and operations teams, especially groups already running an industrial historian or process data system, that want an agent working alongside them on the systems they already model.

How it's deployed

On-prem and single-tenant. Installed in your environment, behind your network perimeter. User authentication runs through single sign-on with Microsoft Entra ID (Azure AD). Streamsight is built for mainstream historians and process data protocols; PI Web API is the first integration that's shipped, with additional historians and protocols in active development. Streamsight is pre-1.0 and currently shown as an early preview to interested teams.