MCP server for querying Pokémon data from PokéAPI
This MCP server demonstrates basic functionality but falls well short of production quality. All 5 tools are present with visible definitions in pokemon/poke.py, but critical gaps exist across naming, descriptions, schemas, and error handling. Tool descriptions are minimal (10-60 chars), parameter schemas lack depth, and error responses are bare strings offering no recovery guidance. The server reads from the public PokéAPI but has no structured output formatting, pagination, or input validation.
Create a powerful squad of Pokémon for a tournament.
Get a Pokémon's evolution chain.
Get detailed info about a Pokémon by name.
List a Pokémon's moves (limited).
List popular tournament-ready Pokémon.
create_tournament_squad has empty input schema (no parameters) but no documentation of what controls squad composition (team size, Pokemon pool, strength criteria). LLMs cannot reason about when to call this tool vs alternatives.
No parameter descriptions for 'name' and 'limit' across tools. Descriptions state 'Name of the Pokémon' and 'Maximum number of moves to return' but lack format guidance (case sensitivity, spelling variants, valid range). RUBRIC: 100% of A+ tools have param descriptions; this has 0%.
All tool descriptions are 10-60 characters; RUBRIC baseline is 194 chars average (p10=34, p90=392). Descriptions lack WHAT/WHEN/HOW context. E.g., get_pokemon_info is 38 chars; should explain when to use it vs get_pokemon_evolution_chain, what fields are returned, and how to handle 'not found' cases.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Error responses are bare strings with no recovery guidance (e.g., 'No data found for Pokémon: {name}'). RUBRIC PATTERN: errors must tell the LLM what to do next. Should return structured error with category (retryable, user-fixable, fatal) and suggestions.
No input validation or constraint documentation. The 'limit' parameter in list_pokemon_moves has a default (10) and is clamped to 50 in code, but this constraint is not documented in the parameter schema or description. LLM has no visibility into valid range.
Output schemas not documented. All tools return free-form strings; no structured field definitions. RUBRIC: 100% of A+ tools have documented return types. Agents cannot extract specific fields or chain calls reliably.
No pagination or result limiting documentation. create_tournament_squad always returns 6 Pokemon; list_popular_pokemon always returns 6. If the data source changes or agents need flexibility, there's no way to control batch size or retrieve next/previous batches.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only and safe to retry, but this is not declared. Current spec (2026-07-28) supports these annotations; omitting them loses context for agent planning.
create_tournament_squad is poorly named. It does not accept parameters (team size, Pokemon pool, strategy). The name suggests customizable squad creation, but the tool always returns the same 6 hardcoded Pokemon. Better names: 'get_recommended_tournament_squad' or 'get_default_tournament_team'. Current name misleads LLMs about capability.
Case sensitivity not documented. 'name' parameter is normalized to lowercase in code (name.lower()), but LLMs and users often use mixed case (Charizard, CHARIZARD). Docs should state: 'Name is case-insensitive; e.g., Charizard, charizard, and CHARIZARD all work.'