The server provides three well-named, appropriately scoped read-only SQL tools with clear descriptions and basic JSON Schema support. However, significant gaps exist in parameter documentation, output schema specification, and error handling guidance. Parameter descriptions are present but often minimal (e.g., 'The catalog name' for optional catalog parameter). No output schema documentation exists, LLMs cannot determine what fields to expect from CSV results. Error handling is minimal; exceptions are wrapped in generic RuntimeException without recovery guidance. The tools follow the verb_noun naming pattern correctly and avoid overly generic names, but the implementation lacks the rigor expected of production-grade database connectors.

• Add explicit output schema documentation to each tool. Example for get_tables: 'Returns CSV with columns: TABLE_CATALOG, TABLE_SCHEMA, TABLE_NAME, TABLE_TYPE. First row contains headers.' Specify which columns are always present vs conditional.
• Expand parameter descriptions to 50-100 chars each with actionable guidance. E.g., catalog: 'The catalog/database name. Optional; if omitted, defaults to the connection catalog. Leave empty for single-catalog sources.' instead of just 'The catalog name'.
• Add pagination to get_tables and get_columns. Include optional 'limit' (1-1000, default 100) and 'offset' (0+, default 0) parameters. Document: 'Results are limited to prevent context window exhaustion. Use offset to fetch subsequent pages.'
• Enhance error messages with recovery guidance. Instead of throwing 'ERROR: Table not found', catch SQLException and return: 'Table "xyz" not found. Call sybaseiq_get_tables to list available tables.' This guides the LLM's next step.
• For run_query tool, replace the clause list with prose: 'Executes SQL SELECT statements using Sybase IQ dialect. Supported: SELECT, FROM, WHERE, INNER/LEFT JOIN, GROUP BY, HAVING, ORDER BY, LIMIT, OFFSET. Identifiers must be quoted with backticks. Transactions not supported, each call is auto-committed.'
• Add tool annotations to schemas (if using MCP 2026-07-28+): include readOnlyHint: true for all three tools to signal they are safe for agent exploration without side effects.
• Document timeout and result size limits in descriptions. E.g., 'Queries timeout after 30 seconds. Results are capped at 10,000 rows to prevent context window exhaustion. Use LIMIT and OFFSET to fetch pages.' This sets expectations and prevents silent failures.

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

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