Server exposes 5 read-only tools with complete JSON schemas and descriptions. All tools are properly annotated with ToolAnnotations (readOnlyHint=True, idempotentHint=True). Naming follows verb_noun pattern (list_questions, search_answers, get_answer, self_check_list, standard_info). Descriptions are present and actionable (avg ~100 chars). Parameters have types and descriptions. Output is structured JSON. However, descriptions are terse (under 150 chars), parameter descriptions lack format/constraint details, and error handling is minimal. No pagination support despite potentially large result sets. Tool composition is clean (single responsibility), but search_answers lacks result limiting guidance.

• Add formal enum constraints to 'cluster' parameter in list_questions and search_answers (e.g., enum: ['ai-service-standard', 'enterprise-knowledge-base', '']) instead of relying on description text.
• Document min/max bounds for 'top_k' in search_answers (e.g., 'integer, 1-50, default 5') and cap results at 50 to prevent context window exhaustion.
• Expand tool descriptions to 150-250 chars, explicitly stating WHEN to use each tool. E.g., 'Use search_answers for keyword-based discovery; use list_questions to browse all titles; use get_answer once you have a slug.'
• Add pagination support to search_answers: return a 'next_cursor' field and accept an optional 'cursor' parameter to enable multi-page result traversal.
• Enhance error messages in self_check_list and standard_info to include recovery hints. E.g., if self-check items not found, suggest 'Try search_answers("self-check")' or list available clusters.