ClickHouse MCP server has 4 well-named, read-only tools with clear descriptions and basic input schemas. All tools follow verb_noun naming (clickhouse_list_*, clickhouse_describe_*, clickhouse_query). Descriptions are concise (30-80 chars) and explain purpose. Input schemas present with type definitions and descriptions for all parameters. However, output schemas are not documented, LLMs cannot predict result structure. Error handling is absent from descriptions. No pagination guidance for potentially large result sets. Parameter descriptions lack format/constraint details (e.g., max_rows minimum=1 is in schema but not description text). Missing idempotency hints and recovery guidance.

• Document output schemas for all tools. Example: 'Returns array of {database: string, tables_count: integer}' for clickhouse_list_databases.
• Add explicit pagination guidance to clickhouse_query description: 'Results limited to max_rows (default 1000). For large datasets, use LIMIT clause in SQL or increase max_rows up to [server limit].'
• Enhance clickhouse_query description to state: 'Only SELECT statements allowed. Attempts to execute INSERT, UPDATE, DELETE, or DDL will fail with error message indicating the restriction.'
• Add recovery hints: 'If query fails with syntax error, call clickhouse_describe_table to inspect schema before retrying.'
• Explicitly mark all tools as read-only and idempotent in descriptions: 'This tool is read-only and safe to retry. No data is modified.'
• Document parameter constraints in descriptions, not just schema. E.g., 'max_rows must be between 1 and [limit]; defaults to 1000.'