This Salesforce Marketing MCP server provides three read-only data discovery and query tools with generally clear naming and adequate descriptions. Tool names follow verb_noun pattern (get_tables, get_columns, run_query) and accurately reflect their actions. Descriptions are present and reasonably detailed (ranging 194-298 chars), exceeding the 20-char minimum. Input schemas are properly defined with JSON Schema types and field descriptions. However, there are significant gaps: output schemas are not formally documented (responses mention 'CSV format' but don't define structured output fields), parameter descriptions could be more prescriptive about constraints, error handling lacks recovery guidance, and there is no discussion of limits/pagination for potentially large result sets. The tools are read-only and well-composed (each does one thing), reducing risk. Overall: acceptable for a data discovery toolkit, but falls short of production-grade quality due to missing output documentation and error recovery patterns.

• Document the output schema for each tool. For CSV responses, define the expected column names, types, and structure. Example: salesforce_get_tables returns [TABLE_CAT, TABLE_SCHEM, TABLE_NAME, TABLE_TYPE]; salesforce_run_query returns columns matching the SELECT statement with their inferred data types.
• Implement structured error responses with actionable guidance. Categorize errors as 'retryable' (connection timeout, temporary lock) vs. 'user-fixable' (invalid table name, malformed SQL). Example: 'Table "invalid_table" not found. Call salesforce_get_tables to see available tables.'
• Add explicit result limits and pagination. For run_query, document a maximum result row limit (e.g., 1000 rows) and add optional 'limit' and 'offset' parameters. Update description to state the limit clearly.
• Enhance parameter descriptions with constraint clarity. For 'catalog' and 'schema', state: 'Optional. If omitted, the current catalog/schema is used. Provide as empty string or omit to use defaults.' For 'sql', add: 'Must be a SELECT statement. Supported clauses: FROM, INNER JOIN, LEFT JOIN, GROUP BY, ORDER BY, LIMIT, OFFSET. Identifiers must be quoted with double quotes.'
• Add tool annotations in McpSchema.Tool. Mark all three tools with idempotentHint=true to signal safe retries. Consider readOnlyHint=true for additional clarity.
• Include example column names and data types in descriptions. Example for salesforce_get_columns: 'Returns columns: COLUMN_NAME (string), TYPE_NAME (string), REMARKS (string), TABLE_NAME (string).' This helps LLMs understand output structure without parsing CSV.

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

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