Five read-only tools with reasonable naming (verb_noun pattern) and moderate descriptions. All tools have input schemas with typed parameters and descriptions. However, descriptions are brief (avg ~80 chars, below the 194-char baseline), lack WHEN-to-use guidance, and omit error recovery hints. Output schemas are not documented. No tool annotations (readOnlyHint, idempotentHint). Parameter descriptions are functional but sparse. Error handling is absent from tool definitions. The server is competent but falls short of production-grade polish.

• Expand tool descriptions to 150 - 250 chars. Include WHEN to use (e.g., 'Use this to find unread messages in a specific conversation') and WHAT it returns (e.g., 'Returns a list of messages with sender, timestamp, and body text').
• Document output schemas for each tool. Specify the fields returned (e.g., get_messages returns [{id, body, sent_at, conversationName, senderName, type}]) so LLMs can plan chained calls.
• Add tool annotations: set readOnlyHint=true and idempotentHint=true on all five tools to signal they are safe to retry and do not modify state.
• Enhance parameter descriptions with bounds and examples. E.g., 'limit: Max results (1 - 100, default 20)' and 'from: Start date (ISO 2025-01-15 or relative 5h/3d/10m)'.
• Add error recovery hints to descriptions. E.g., 'If conversation not found, try get_conversations() to list available conversations' or 'If search returns no results, try broadening the search query or removing filters'.