RemoteAC exposes 7 tools with Korean descriptions and minimal parameter documentation. All tools are registered via [McpServerTool] attributes with names and descriptions present. However, critical gaps undermine usability: (1) Input schemas are either empty {} or minimally documented, tools like room_get_status and ac_get_status accept no parameters but lack output schema documentation; (2) Parameter descriptions exist but are sparse and lack type constraints (e.g., ac_set_op_mode accepts 'operationMode' string with no enum constraint in schema, only in description); (3) No error handling guidance, SetTemperature clamps silently, SetOperationMode returns error text but no structured recovery path; (4) Output schemas are inferred from return types (RoomStatus, AirconStatus, ControlResult) but not explicitly documented in tool definitions. Tool names follow verb_noun pattern (get_, set_, turn_) which is good, but descriptions are in Korean, limiting LLM comprehension in English-first environments.

• Add explicit output schemas to all tools. Document RoomStatus fields (temperature, humidity, etc.), AirconStatus fields (isRun, opMode, temperature, fanSpeed), and ControlResult.Result as structured JSON Schema.
• Convert parameter constraints to JSON Schema enums. For ac_set_op_mode, change schema from {"operationMode":{"type":"string"}} to {"operationMode":{"type":"string","enum":["Auto","Cool","Dry","Heat","Fan"]}}.
• Translate all descriptions to English. Rewrite tool and parameter descriptions in English to ensure compatibility with English-first LLM agents and broader adoption.
• Add numeric constraints to ac_set_temperature schema: {"temperature":{"type":"integer","minimum":17,"maximum":30,"description":"Target temperature in Celsius (17 - 30°C)"}}.
• Implement structured error responses. Return JSON objects with 'error' and 'suggestion' fields instead of plain text, e.g., {"error":"Invalid operation mode","suggestion":"Valid modes: Auto, Cool, Dry, Heat, Fan"}.