Server has 4 well-defined tools with complete input schemas and descriptions. All tools follow verb_noun naming (calculate_*). Descriptions are substantive (150-250 chars, within baseline 194 avg). Input parameters have types and descriptions. However, output schemas are not formally documented, responses are markdown strings rather than structured objects. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint, idempotentHint). Parameter descriptions could be more prescriptive about constraints and formats.

• Define formal output schemas for each tool. Instead of markdown strings, return structured JSON: {result: number, steps?: string[], unit?: string, breakdown?: object}. This enables LLMs to extract and chain results reliably.
• Add tool annotations to inputSchema: readOnlyHint: true for all tools. This signals to clients that these tools are safe to call without user confirmation.
• Enhance error messages with recovery guidance. E.g., 'Unable to fetch exchange rates (API timeout). Try again in 30 seconds, or use a cached rate if available.'
• Add min/max constraints to numeric parameters. E.g., annualRate: z.number().min(0).max(100).describe('Annual interest rate (0 - 100%)'), years: z.number().min(1).max(50).
• Validate currency codes against ISO 4217 enum or a whitelist. Replace the regex check with: from: z.enum(['USD', 'EUR', 'JPY', 'THB', ...]) to prevent invalid codes from reaching the API.