Building automation / Interface prototype
BAS Agent Console
An operator workspace prototype that shows what building agents read, why they acted, and whether the equipment responded.
- Status
- Prototype
- Year
- 2026
- Role
- Product design + development
- Built with
- Next.js · React · TypeScript · shadcn/ui
The problem
An agent can make a building easier to operate only if its work remains legible. A simple chat transcript does not show which station values were trusted, which policy allowed an action, or whether the equipment actually responded.
This prototype treats the event trace as the primary operator surface. It is intentionally built with synthetic demo data, so the interaction and evidence model can be explored without presenting it as a live Niagara control system.
What it does
Trace every action
Organizes reads, reasoning, control actions, approvals, and verification into one chronological record with the evidence captured at each step.
Expose the boundary
Shows how Niagara roles and categories establish station authority, then layers narrower agent policies, command limits, and operator-review rules on top.
Require approval
Keeps consequential requests pending until an operator can review the proposed command, affected equipment, supporting evidence, and downstream impact.
Return to equipment
Connects the activity record to AHU and VAV views so an operator can move from an agent decision back to the equipment context it touched.