This server defines three read-only tools for querying Apache Hadoop Hive with reasonable structure. Tools are clearly named with action verbs (get_, run_) and have descriptions present. However, several quality gaps limit production readiness: (1) Parameter descriptions lack detail about constraints, formats, and valid values. For example, 'The table name' does not explain whether table names are case-sensitive, what characters are allowed, or how to escape special names. (2) Output schemas are completely undocumented, the tools claim to return CSV format, but there is no structured schema definition of what fields or rows to expect, forcing LLMs to parse unstructured text. (3) Error handling is minimal, the run() method throws RuntimeException with a generic message, giving agents no guidance on retry logic, user-fixable errors, or recovery steps. (4) No pagination support visible despite tools potentially returning large result sets. (5) Tool composition is reasonable, hive_get_tables and hive_get_columns are discoverable prerequisites to hive_run_query, but cross-tool chaining IDs (catalog, schema, table) are not explicitly documented in descriptions.

• Add explicit output schemas to all three tools. Define returned CSV columns as a structured object: for hive_get_tables return {catalog?: string, schema?: string, table: string, table_type: string}; for hive_get_columns return {catalog?: string, schema?: string, table: string, column: string, data_type: string, remarks?: string}; for hive_run_query document the result as 'array of objects with keys matching SQL SELECT columns'.
• Enhance parameter descriptions with actionable constraints. Example for 'table' parameter: 'The table name (case-sensitive, alphanumeric and underscore only, max 255 chars, may include schema prefix as schema.table).' For 'sql' parameter: 'A SELECT statement (max 10,000 chars; supported clauses: FROM, INNER JOIN, LEFT JOIN, GROUP BY, ORDER BY, LIMIT (max 1000 rows); identifiers must be double-quoted if they contain special characters or reserved words).'
• Add pagination support to hive_get_tables. Include optional parameters: limit (1-1000, default 100), offset (0+, default 0). Return in response: total_count, result_count, has_more for LLM to manage iteration.
• Document and enforce query result limits explicitly. Modify hive_run_query description: 'Returns up to 1000 rows; use LIMIT and OFFSET to paginate. If result set exceeds 1000 rows, the response includes a note: [truncated, use OFFSET to retrieve remaining rows].' Cap the actual implementation to enforce this.
• Implement structured error responses. Catch SQLException and other exceptions separately; return structured error with retryable flag, error_code, and user-actionable message. Example: {isRetryable: true, errorCode: 'TABLE_NOT_FOUND', message: 'Table not found. Call hive_get_tables() to list available tables.'}

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

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