Server defines 4 tools with complete input schemas and basic descriptions. All tools are READ_ONLY and follow verb_noun naming (search_*, get_*). However, descriptions are minimal (34-75 chars, below the 50-200 char LLM-optimized baseline), lack context on WHEN to use each tool, and omit dependency hints. Output schemas are not documented, LLMs cannot plan downstream calls. Parameters lack format constraints (e.g., id ranges, name length limits). No error recovery guidance. Schemas are present but sparse; descriptions exist but underutilized.

• Expand tool descriptions to 100-150 chars. Include WHEN to use: 'Search for characters by partial name when you don't know the exact ID. Returns up to 20 matches; use get_character() if you have an ID.'
• Document output schemas in tool definitions. Add 'Returns: {id: number, name: string, status: string, species: string, gender: string, origin: string, location: string, image: string, episode_count: number}' to get_character description.
• Replace example values with format constraints. Change 'Character ID (e.g. 1 for Rick Sanchez)' to 'Character ID (positive integer, 1-826)' based on API limits.
• Add error recovery hints. Wrap fetch calls with try-catch that returns: 'Character ID {id} not found. Try search_characters(name="Rick") to find valid IDs.' instead of generic API error.
• Document pagination for search_characters. Add limit and offset parameters (default limit=20, max=50) and return total_count so LLMs know if more results exist.