Server has three well-named tools with complete input schemas and descriptions. Naming follows verb_noun pattern (get_*, ping_*). All parameters have type definitions and descriptions. However, output schemas are not documented, LLMs cannot predict response structure. Error handling is minimal (generic error codes, no recovery guidance). Parameter descriptions are adequate but lack format/constraint details. No tool annotations (readOnlyHint, idempotentHint). Missing documentation of what fields are returned and how to chain tools.

• Document output schemas for all three tools. Specify what fields are returned (e.g., get_my_ip returns {ip: string, country: string, ...}). This enables LLMs to plan chaining and extract correct fields.
• Add tool annotations to capabilities. Include readOnlyHint: true for all three tools (they are read-only). This signals to LLMs that tools are safe to retry.
• Enhance error messages with recovery guidance. Instead of 'Invalid IP', return 'Invalid IP address. Expected IPv4 format (e.g., 192.168.1.1).' or 'IP not found. Try a different IP or use get_my_ip() to get your current IP.'
• Add format/constraint details to parameter descriptions. For 'areas', clarify: 'Comma-separated region codes or "all". Common regions: beijing, tokyo, california, london. Use "all" to test all 25 regions (slower).'
• Cap result limits and document them. If ping_test returns many results, cap at 50 and state in description: 'Returns up to 50 ping results per area. Use specific areas parameter to reduce output.'