A React-based e-commerce demo application (shoe store) that exposes shopping cart and product filtering functionality as MCP tools
This is a demo/placeholder MCP server with critical deficiencies. Tools are defined in a minified JavaScript bundle (docs/assets/index-DX1opX6j.js), making source code inspection extremely difficult. The package.json and TypeScript setup suggest this is a Vite-based frontend application (React + Zod), not an actual MCP server implementation. No visible MCP server code, transport layer, or proper tool registration is present. Tool definitions appear to be inferred from bundle output rather than explicit MCP tool registration. Descriptions exist but are minimal (10-40 chars). Input schemas are partially visible but lack parameter descriptions entirely. No error handling, security considerations, or output schemas are documented. The server appears to be a proof-of-concept web app rather than a production-grade MCP implementation.
Add an item to the shopping cart
Display the current shopping cart items to user
Filter products by brand
Filter products by category
Filter products by gender (e.g., MEN, WOMEN, KIDS)
Get all available store items
Get available filter options for products
No parameter descriptions in any schema. Tools like add_to_cart accept 'id' but do not describe what kind of ID, how to obtain it, or valid range.
display_cart and purchase_cart have no input schema defined (empty object {}), making it impossible for LLMs to understand what parameters are valid or required.
Tool descriptions are too brief (10-40 chars) and lack context. 'Display the current shopping cart items to user' does not explain when to call it vs other tools, what structure is returned, or what it does internally.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Purchase the current shopping cart items
No output schemas documented anywhere. LLMs cannot plan downstream calls or extract relevant fields from responses. What fields does get_all_store_items return? Is there pagination?
purchase_cart is an irreversible operation (WRITE/IRREVERSIBLE risk) but has no description explaining it modifies state, no confirmation mechanism, no error guidance on what to do if payment fails.
No error handling documented. What error does add_to_cart return if the item ID is invalid? How should the LLM recover?
Filter tools (filter_by_brand, filter_by_gender, filter_by_category) are marked WRITE operations but are actually queries. They should be READ_ONLY, and their descriptions should clarify they filter the current product list.
No tool composition guidance. After get_all_store_items, the LLM cannot determine: should I filter first, then add to cart? Or add to cart, then display? Responses should include IDs and references for chaining.