Open source UI for watching AI agents in clinical operations.
The queue surface that almost every AI-native medtech company needs. The one almost none have shipped.
import { SupervisionQueue } from '@wiserframe/supervision-queue';
export function OpsConsole() {
return (
<SupervisionQueue
feed={agentFeed}
onTakeover={routeToHuman}
onEscalate={pageSupervisor}
audit={auditChain}
a11y="AA"
/>
);
}What it is
A small, focused UI component for the operations leader who watches live AI agents work. Live agent queue with priority routing. Per-call status, confidence, and language detection. Escalation routing. Takeover affordance. Audit chain.
Built in React, with TypeScript types. Designed for use inside a host product — your dashboard, your console, your hospital ops surface. Apache-licensed.
Designed around real-world patterns: the call-center supervisor watching a voice agent take an inbound patient call, the prior-auth ops lead watching a back-office agent fetch records, the patient-access manager watching the AI agent triage incoming inquiries.
Why this is open source
The supervisor surface is the one almost everyone needs and almost no one has shipped well. Here's why we're putting ours in the open.
Almost every AI-native medtech company is shipping an AI agent and bolting on a basic admin view. The supervisor surface — the place where ops staff watch the agent, route escalations, and intervene when something goes off the rails — deserves better than a CRUD table with a status column.
Trust requires interrogability; interrogability requires the supervisor surface; the supervisor surface shouldn't depend on whose AI you happen to be using.
It will be. The supervision queue is the first component to ship as a real, installable artifact. It's the proving ground for the library distribution pattern.
Who it's for
Building voice AI agents, AI scribes, autonomous back-office agents, or any AI deployment where a non-technical operations leader needs to watch the AI work in production.
With experience in clinical UI, accessibility, or real-time UI patterns. The codebase is small and the contribution surface is well-scoped.
Studying clinical AI deployment. The component is documented enough to use as a reference in coursework or research write-ups; cite the GitHub repo.
What's in the project
Why we built it
Sitting alongside our medtech client work, we needed a supervisor surface that could be reused across deployments and across host products. The constraints — real-time updates, live transcript handling, fast operator intervention, clean audit trail, accessibility from the first frame — all turn out to be the same constraints, whether the AI is a voice agent or a back-office automation.
Rather than build it three times privately, we built it once, properly, and put it where it can be useful to the rest of the field. The version that we ship to clients is the version that's in the open-source repo, plus product-specific integration.
If you build on top of it, we'd love to know. If you find something wrong, the issue tracker is open. If you ship something better, fork it and tell us what you did.
How it connects to the rest of WiserFrame
This is one component of the Wiserframe Med Component Library. The full library — 230+ patterns across 20 clinical categories — is a design exemplar today, on the path to becoming an installable component package. The supervision queue is the first piece of that package to ship as real, usable code.
Closing CTA
If you're shipping an AI agent in healthcare operations and the supervisor surface is the part you haven't figured out yet — book the $1,000 Diagnostic Working Session. Two hours together plus follow-up and a written output, including a redesigned screen against your actual product if useful.
Or just open an issue on the GitHub repo — we read them all.