The server has 4 tools with reasonable naming conventions (all start with verbs: list_, analyze_). Descriptions are present and moderately detailed (120-180 chars each), explaining WHAT the tools do and basic context. However, several critical gaps reduce overall quality: (1) Input schemas are minimal, most parameters lack type constraints, enums, or detailed format descriptions; (2) No output schemas are documented anywhere in the visible code; (3) Error handling is not evident in tool definitions; (4) No pagination support for list_ tools despite returning potentially large result sets; (5) Parameter descriptions are sparse. The tools are READ-ONLY and well-scoped (one action each), meeting basic composition patterns, but fall short of production-grade LLM integration standards.

• Add structured output schema documentation for each tool. For example: 'list_projects returns [{"project_id": string, "project_name": string, "description": string, "allowed_dataset_patterns": string[]}]'. This enables LLM chaining and field extraction.
• Declare 'table_type' in list_tables as a JSON Schema enum: {"type": "string", "enum": ["all", "table", "view", "materialized_view"]} instead of free-form description.
• Add pagination to list_datasets and list_tables: introduce 'limit' (1-1000, default 50) and 'offset' (or next_cursor) parameters, with a note: 'Returns paginated results. Limit capped at 1000 to prevent context exhaustion. Use offset/next_cursor to fetch additional pages.'
• Expand 'project' parameter description in list_datasets: 'Project ID to list datasets from. If omitted, defaults to the billing project configured in the server. Required format: alphanumeric with hyphens (e.g., my-project-123).'
• Add error guidance to all tool descriptions: 'On permission denied, ensure the service account has roles/bigquery.dataViewer. On table not found, try list_tables first to confirm the table exists.'
• Add dependency hints to tool descriptions: list_projects states 'Call this first to discover available projects before listing datasets.' list_datasets states 'Requires a project_id from list_projects or configuration.' list_tables states 'Requires a dataset_id from list_datasets.'
• Document analyze_table return fields explicitly: e.g., 'Returns {"table_id": string, "row_count": number, "columns": [{"name": string, "type": string, "null_count": number, "cardinality": number}], "created_time": ISO8601, "modified_time": ISO8601}'.

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

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