Server has 6 tools with complete input schemas and descriptions. Naming follows verb_noun convention (compare_documents, get_balance, list_comparisons, export_comparison_*). Descriptions are present and actionable (avg ~80 chars). However, parameter descriptions lack detail on constraints, formats, and dependencies. Output schemas are not documented, LLMs cannot predict response structure. Error handling is present in code (timeout guidance, rate-limit hints) but not surfaced in tool descriptions. No tool annotations (readOnlyHint, idempotentHint). Missing pagination guidance for list_comparisons despite accepting limit parameter.

• Document output schema for each tool. Example for compare_documents: 'Returns {comparison_id: string, status: "completed"|"pending"|"failed", created_at: ISO8601, diff_summary: {added: number, removed: number, modified: number}, ...}'
• Add constraint details to parameter descriptions. Example: 'limit: Maximum number of comparisons to return (1 - 100, default 20)' and 'locale: Language code (en, de, fr, es, etc.)'
• Add readOnlyHint: true to all 6 tool schemas to signal they are safe to retry and have no side effects.
• Enhance get_balance description: 'Returns current API credits and rate-limit status. Call before compare_documents to verify sufficient credits. Returns {credits_remaining: number, rate_limit_resets_at: ISO8601, ...}'
• Add pagination to list_comparisons: accept 'offset' or 'cursor' parameter, return 'total_count' and 'next_cursor' in response, document default limit (20) and max (100).