A Pokémon information server using Model Context Protocol
MCPoke has 12 tools with consistent naming (verb_noun pattern) and basic schemas. However, descriptions are generic (avg ~60 chars, well below the 194-char baseline), parameter descriptions are minimal or absent, and output schemas are not documented. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint). The server demonstrates functional structure but falls short of production-grade clarity for LLM reasoning.
Download sprites for a range of Pokémon IDs
Check if sprites for a Pokémon are cached locally
Clear the sprite cache
Compare multiple Pokémon side by side
Download and cache all sprites for a Pokémon
Get information about a Pokémon ability by name or ID
Get information about a Pokémon move by name or ID
Descriptions are generic and under 100 chars; lack context for LLM tool selection. E.g., 'Get detailed information about a Pokémon by name or ID' does not explain when to use get_pokemon vs search_pokemon or what fields are returned.
Parameter descriptions are missing or trivial. 'name_or_id' appears in 7 tools but is never described (format, valid range, examples). LLMs cannot infer whether to pass 'pikachu', '25', or 'PIKACHU'.
Output schemas are not documented. Tools return processed data but LLMs cannot see what fields to expect (e.g., does get_pokemon return stats, abilities, moves, or all three?). This forces agents to guess and wastes tokens on trial-and-error.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | <=2025-11-25 | v2 |
Get detailed information about a Pokémon by name or ID
Get statistics about cached sprites
Get information about a Pokémon type by name or ID
Calculate type effectiveness between attacking and defending types
Search for Pokémon by partial name match
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes. clear_sprite_cache is destructive but unmarked; download_sprites and bulk_download_sprites are writes but unmarked.
Error messages lack recovery guidance. E.g., 'Ability "X" not found' does not suggest calling search_ability or listing available abilities. Agents cannot self-correct without extra discovery calls.