Server has 4 tools with basic schemas and descriptions in Chinese. All tools are READ_ONLY and properly registered with Zod schemas. However, descriptions are minimal (10-30 chars), lack context on WHEN to use each tool, and do not explain dependencies between tools. Parameters lack detailed constraints (e.g., chat_id format, date format validation). Output schemas are not documented, responses are free-text JSON strings rather than structured objects. No error handling guidance. Tool names are clear (verb_noun pattern) but descriptions fall short of the 50-200 char LLM-optimized baseline. Missing pagination documentation for list_chats and search_messages despite returning potentially large result sets.

• Expand tool descriptions to 50-200 chars. Example: 'get_messages: Retrieve message history from a Telegram chat by ID. Supports date range filtering and pagination. Use this to fetch conversation context; use search_messages to find specific keywords.' This clarifies intent, prerequisites, and when to choose this tool over alternatives.
• Document output schemas explicitly. Instead of returning JSON.stringify(), return structured objects: { messages: [{id, timestamp, sender, text}], total_count, has_more }. Include field types and meanings in tool descriptions.
• Add format constraints to parameters. chat_id: 'Telegram chat ID (numeric string, e.g., 3695601502)'; start_date/end_date: 'ISO 8601 date (YYYY-MM-DD), must be <= today'; limit: 'Number of results (1-100, default 50)'.
• Implement result limits and pagination. Cap list_chats and search_messages at 50 results by default. Return total_count and next_offset to enable agent-driven pagination. Document in descriptions: 'Returns up to 50 results; use offset to fetch more.'
• Add error handling with recovery guidance. Catch database errors and return: 'Chat not found. Call list_chats() to see available chats.' For invalid dates: 'start_date must be <= end_date and <= today (YYYY-MM-DD format).'