MCP Server for Shopify API, enabling interaction with store data through GraphQL API.
Server has 10 tools with consistent naming (verb_noun pattern) and basic descriptions. All tools have input schemas with Zod validation. However, descriptions are generic (avg ~40 chars, well below the 194-char baseline), lack context on WHEN to use each tool, and omit error recovery guidance. Parameter descriptions are minimal. Output schemas are not documented. No tool annotations (readOnlyHint/destructiveHint). Error handling exists but returns generic messages without actionable recovery steps. Composition is sound (each tool does one thing), but descriptions fail to guide LLM selection between similar tools (e.g., get-orders vs get-order, get-products vs get-product-details).
Complete a draft order
Create a basic discount code
Get collections from the shop
Get shopify customers with pagination support
Get a single order by ID
Get orders with advanced filtering and sorting
Get detailed information about a specific product
Descriptions are too short (avg 35 chars vs 194-char baseline). 'Get shopify customers with pagination support' lacks context on WHEN to call this vs other customer tools, what fields are returned, or when pagination is needed.
Output schemas are not documented. LLMs cannot plan downstream calls or extract required fields (e.g., customer IDs for tag-customer, order IDs for get-order). Responses are returned as text blobs without structured field definitions.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish safe read-only tools from destructive ones (tag-customer, create-discount, complete-draft-order). This violates current MCP spec (2026-07-28) which expects tool annotations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Get Shopify products with support for pagination and search
Get shop details
Add tags to a customer
Error messages are generic ('Failed to retrieve customers', 'Failed to tag customer'). No recovery guidance. LLMs cannot self-correct or know whether to retry, ask the user, or abort.
Parameter descriptions are minimal. 'Next page cursor' does not explain format, how to obtain it, or when pagination is exhausted. 'Filter orders using query syntax' lacks examples of valid query syntax.