Pokemon Trading Card Game MCP server for searching cards, tracking prices, and suggesting decks
Four tools with basic schemas and descriptions, but significant gaps in parameter documentation and output structure. All tools have descriptions (10-100 chars, mostly adequate), and input schemas use Zod with type definitions. However, parameter descriptions are sparse or missing context. No documented output schemas. Error handling is minimal (generic try-catch). Tool naming follows verb_noun convention (search_, get_, suggest_) but lacks clarity on composition and chaining. The suggest_deck tool is particularly weak, it returns a 60-card deck recommendation but the output format is undocumented, making it impossible for an LLM to extract and use the results downstream.
Get full details for a specific card by its ID.
Get current market prices for a Pokemon TCG card from TCGPlayer and Cardmarket. Search by name to find pricing for all matching cards.
Search for Pokemon TCG cards by name, type, subtype, legality, HP, and retreat cost.
Suggest a 60-card Pokemon TCG deck based on a strategy, type, or featured Pokemon. Returns a categorized list of recommended cards with rationale.
Output schemas undocumented. search_cards, get_card_details, and get_card_prices return Card objects with complex nested structures (attacks, prices, legalities), but the LLM has no schema to understand what fields are available or how to extract them. suggest_deck returns a 60-card deck but the structure is completely opaque.
suggest_deck description is vague and lacks actionable detail. 'Suggest a 60-card Pokemon TCG deck based on a strategy, type, or featured Pokemon. Returns a categorized list of recommended cards with rationale.', does not explain what 'categorized list' means, what fields are in the response, or how the LLM should use the output. No guidance on deck legality validation or budget constraints.
Parameter descriptions lack format/constraint guidance. search_cards 'hp' param says 'HP filter e.g. [100 TO 200] or [* TO 100]' but does not explain the syntax or validate input. retreatCost says 'Converted retreat cost e.g. 0 for free retreat' but does not clarify if it accepts ranges or single values. No min/max bounds on pageSize (default 10, max 250 stated but not enforced in schema).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Error handling is generic and non-actionable. All tools catch errors with 'Error searching cards: {message}' or similar. No recovery guidance (e.g., 'Try a simpler query' or 'Check card ID format'). No categorization of retryable vs. fatal errors. LLM cannot determine next steps on failure.
Tool composition broken. search_cards returns Card objects with 'id' field, but get_card_details expects 'cardId' parameter. No documentation of this mapping. LLM must infer the relationship or fail to chain tools. suggest_deck returns card recommendations but no IDs or references to feed into get_card_prices or get_card_details.