Server has 5 well-named tools with clear verb-noun patterns (get_*, search_*, list_). All tools have descriptions (avg 180 chars, within 10-1024 baseline). Input schemas are present with type definitions and parameter descriptions. However, output schemas are NOT documented, LLMs cannot predict return structure. Error handling returns error dicts but lacks recovery guidance (e.g., 'Try search_cards() first'). No tool annotations (readOnlyHint, idempotentHint). Parameter descriptions are adequate but some lack format/constraint details (e.g., 'theme' accepts 'slug' but no examples of valid slugs provided). Overall: solid foundation, but missing output documentation and error recovery patterns.

• Document output schemas for all tools. For get_card, specify: {name: string, mana_cost: string, type_line: string, color_identity: array, commander_legality: string, usd_price: number, image_url: string, set: string}. For search_cards, specify: {count: integer, cards: array of card objects}. This enables LLMs to plan downstream operations.
• Enhance error messages with recovery guidance. E.g., when CardNotFoundError: return {error: 'Card "xyz" not found. Try search_cards(query="xyz") to find similar cards.'}. When CommanderNotFoundError: return {error: 'Commander not found. Call list_commander_themes(commander="...") to verify the exact name.'}
• Add readOnlyHint annotation to all 5 tools in the tool definition (via FastMCP metadata or schema). This signals to LLMs that these tools are safe to retry and have no side effects.
• Constrain 'limit' parameter in search_cards: add description 'Maximum number of cards to return (1-100, default 20)' and validate input to reject values outside this range.
• For 'theme' parameter, either provide an enum of valid theme slugs (if finite) or add description: 'Optional theme slug. Call list_commander_themes(commander) first to see available themes for this commander.'