Grounded
MCP tools for AI agent grounding

Ground an AI agent at runtime with the skincare core.

This static shell shows how an AI agent calls MCP tools for graph context, PubMed evidence, routine ingredients, product ranking, safety notes, and an explicit confidence contract.

Scenario acne + hyperpigmentation Evidence loading Products loading Safety loading Confidence loading

Runtime confidence contract

Confidence is carried through the response envelope instead of becoming a free-form paragraph.

Fetching the runtime payload from /api/scenario/acne-hyperpigmentation

Graph evidence

Graph-routing context and caution signals that shape the skincare recommendation before evidence ranking.

Loading graph evidence…

Evidence-based ingredients

PubMed-backed ingredient table with timing and pregnancy-risk context.

Loading ingredient plan…

Evidence

PubMed-backed abstracts with provenance and an explicit confidence signal.

Loading evidence…

Morning routine

Products matched for the AM slot with ingredient rationale and live product links.

Loading morning product shortlist…

Evening routine

Products matched for the PM slot with the same runtime data path.

Loading evening product shortlist…

Safety / cautions

Signals to watch before showing the recommendation as an action item.

Loading safety notes…