ViewCue MCPUI clues for WebMCP

Make agent activity visible where it happens. Point to the right surface, explain the context, and keep the host application in control.

Open the Coretax demo

Read the quick start

Point to the right thing

Attach a clue to stable semantic targets instead of guessing from arbitrary selectors or raw page structure.

Explain what it means

Surface plain-language context for labels, terminology, validation, choices, and results.

Stay out of the driver’s seat

WebMCP tools still belong to the host application. ViewCue adds context without replacing clicks, typing, submissions, or business logic.

One small layer, many useful moments

ViewCue can show a single clue next to a billing code, explain why a field matters, point out the next step, or optionally reflect a tool result. A lifecycle indicator is one example, not the product boundary.

Designed for the page already in front of the person.

Framework-agnostic TypeScript and DOM primitives. No tour library. No remote model. No action interception.

What the integration looks like

Declare a stable target, then call showClue() from the application’s WebMCP tool or UI lifecycle.

runtime.showClue({ targetId, message })

Built around WebMCP

WebMCP gives a page a structured tool surface that an agent can discover and call. ViewCue makes that interaction legible in the same page, while the application continues to own the actual action.

Start with the tutorial →