Spring MCP Server exposes 4 tools (2 unique, duplicated across webflux/webmvc) with minimal descriptions and no visible input schemas in the MCP protocol layer. Tools use Spring AI @Tool annotations but lack formal JSON Schema definitions, parameter constraints, and output documentation. Descriptions are extremely brief (13-32 chars), below the 50-200 char baseline for LLM-optimized tools. No error handling guidance, no pagination support for list operations, and no tool annotations (readOnlyHint, etc.). The server appears to be a Spring AI wrapper rather than a true MCP implementation with proper schema exposure.

• Expand tool descriptions to 50-200 chars. Example: 'getPersonById: Retrieve a single person record by their unique ID. Returns name, age, and nationality. Use when you have a specific person ID; for searches by name or nationality, use getPersonsByNationality.' This guides LLM selection and sets expectations.
• Add formal JSON Schema to tool inputs. Define 'id' as {type: 'integer', minimum: 1, description: 'Unique person identifier'} and 'nationality' as {type: 'string', minLength: 2, maxLength: 50, description: 'ISO 3166-1 country name or code'}. Enums for nationality (USA, France, Germany, etc.) prevent hallucinated values.
• Document output schema for Person object: {type: 'object', properties: {id: {type: 'integer'}, name: {type: 'string'}, age: {type: 'integer'}, nationality: {type: 'string'}}, required: ['id', 'name', 'nationality']}. Clarify null handling for optional fields.
• Add pagination to getPersonsByNationality: introduce 'limit' (1-100, default 20) and 'offset' (default 0) parameters. Return {total: number, results: Person[], hasMore: boolean}. Document in description: 'Returns up to 20 results per call; use offset to fetch additional pages.'
• Consolidate duplicate tool definitions. Remove one webflux/webmvc pair and expose tools via a single canonical MCP server endpoint. If both reactive and blocking variants are needed, use a single tool with async/await internally, not duplicate registrations.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry