Spring MCP Server exposes 3 basic customer lookup tools with minimal documentation. Tool names lack action verbs (getAllCustomer, getNumberOfCustomer, getCustomer use camelCase instead of verb_noun convention). Descriptions are present but generic (10-60 chars, below the 50-200 char LLM-optimized baseline). Only getCustomer has a parameter with type info (userId: int), but lacks parameter description. No output schemas documented. No error handling guidance. No pagination support despite tools returning lists. No tool annotations (readOnlyHint, etc.). The codebase shows Spring AI @Tool annotations are used, but the MCP protocol transport is unknown, cannot verify HTTP/SSE/STDIO compliance.

• Rename tools to verb_noun convention: getAllCustomer → list_customers, getNumberOfCustomer → count_customers, getCustomer → get_customer. This matches LLM expectations and production baselines.
• Add parameter descriptions: userId should document 'The unique customer identifier (integer, 1 - 999999). Required.' Include valid range and format.
• Document output schemas for each tool. E.g., list_customers returns [{id: int, name: string, email: string, created_at: ISO8601}]. This enables LLM planning and field extraction.
• Add pagination to list_customers: accept limit (1 - 100, default 20) and offset parameters; return total_count and has_more. Cap results at 50 items per call.
• Add error handling: return structured errors like {error: 'customer_not_found', message: 'Customer 999 not found. Try list_customers() to see available IDs.', recovery: 'list_customers'} instead of raw exceptions.