AI assistant that can actually do things
Not a chatbot bolted onto the docs. It reads your live namespace and calls real platform tools — bounded by the same permissions as the person asking.
Most "AI in industrial software" is a search box over a manual. This one is wired into the platform: it runs as its own service with a defined tool surface, so it can browse the namespace, inspect a pipeline, query history and take scoped actions rather than describing how you might do those things yourself.
The important part is the boundary. The assistant executes as the signed-in user, through the same role-based permission checks as the UI. It cannot read a branch you cannot read, and it cannot write anything you are not allowed to write. An operator asking it to change a setpoint gets the same refusal the operator would get from the interface.
That makes it useful for the thing engineers actually lose time to: not writing code, but answering "what is this tag", "which pipeline writes here", "what changed since Tuesday" — questions whose answers are spread across a namespace, a pipeline graph and an audit log.
What you get
- Executes real platform tools, not just chat
- Reads the live namespace, pipelines and history
- Bounded by the signed-in user's RBAC — no privilege escalation
- Runs as an isolated service over the message bus
- Every action it takes lands in the audit log