# PRSIM

> A merchant-governed adaptive commerce prototype built with WebMCP.

PRSIM lets a shopper's AI companion operate the product page already open in the browser. Explicit purchase context is matched to the closest merchant-authored shopping scenario. The website then renders approved layouts, copy, imagery, customer evidence, product facts and commerce controls.

The model does not generate arbitrary interface, product claims, prices or policies. PRSIM owns rendering and state; the merchant owns the available experiences; the shopper retains product choice and purchase confirmation.

## Core interaction model

1. The shopper expresses a need such as a destination, budget, deadline, practical constraint, aesthetic preference or gift context.
2. `prepare_shopping_experience` resolves the nearest authored scenario and updates the visible storefront.
3. The broad entry tool is replaced by focused follow-up capabilities.
4. Later questions can refine one part of the page without regenerating the complete experience.

## Focused experience capabilities

- Change the current hero while preserving product truth.
- Add, remove, replace or prioritize an authored content block.
- Show customer evidence with its source and prototype qualification.
- Answer questions from structured product facts.
- Verify airline fit and estimate delivery.
- Select a color or optional bundle.
- Navigate to relevant information, reset the experience or open payment review.

## Commerce capabilities

PRSIM exposes catalogue, product, variant, policy, cart and order actions comparable to a Shopify MCP storefront. It also includes a local UCP-like checkout lifecycle covering checkout creation, reading, updating, completion and cancellation.

These commerce integrations are local prototype implementations, not connections to a live Shopify store, real inventory or production payment processing.

## Merchant system

The merchant studio demonstrates the content model behind the storefront:

- Brand and product truth
- Reusable layouts
- Approved images and customer evidence
- Purpose-based content blocks
- Assemblies binding layouts, copy and assets
- Authored shopping scenarios and coverage analytics

The prototype contains 21 commercial scenario seeds, four bag formats, four colorways and 198 registered visual assets. A separate make-it-yourself experience is an authored non-commercial easter egg.

## Privacy and decision support

The prototype uses only purchase context explicitly present in the current conversation. The active need is session-local and can be reset. Demonstration character names are storytelling devices, not hidden user profiles.

PRSIM is inspired by the Elaboration Likelihood Model by Richard E. Petty and John T. Cacioppo. Its goal is relevance-first decision support: reduce irrelevant cognitive work while keeping product truth, limitations, alternatives and shopper control accessible.

## Project links

- Live application: https://prsim.yafa.sh/#preview
- Source code and technical documentation: https://github.com/Hodisy/Prsim
- OpenAI WebMCP documentation: https://learn.chatgpt.com/docs/webmcp
- Research reference: https://doi.org/10.1016/S0065-2601(08)60214-2