Severe quality gaps across all 46 tools. All tools lack input parameter descriptions, parameters are defined with only type and generic descriptions like 'Query parameters for filtering' or 'The X data to create', providing no actionable guidance to LLMs. No output schemas are documented. Descriptions are minimal (10-30 chars) and do not explain WHEN to use each tool or what distinguishes similar tools (e.g., get_access_invitations vs get_access_invitation). No error handling guidance. Parameter schemas use generic 'object' types without field-level constraints, enums, or validation rules. No pagination guidance despite 46 list-style tools. Credentials are injected via environment variables (correct), but no per-tool permission declarations. Tools lack idempotency hints and confirmation patterns for destructive operations.

• Add detailed parameter descriptions to all tools. For 'data' params in POST/PATCH tools, document required fields, valid enums, and constraints. Example: 'data: JSON object with fields {name (string, 1-255 chars), status (enum: active|inactive), priority (int, 1-5)}.'
• Document output schemas for all tools. Specify returned fields, types, and which IDs/references are available for chaining. Example: 'Returns {id, name, status, created_at, owner_id}.'
• Add pagination support to all list tools. Accept limit (1-100, default 20) and offset/cursor parameters. Return total_count and next_cursor. Document in descriptions.
• Add confirmation/dry-run patterns to destructive tools (delete_*). Require explicit confirmation or offer a dry-run mode that shows what would be deleted without executing.
• Declare required permissions for each tool (e.g., @mcp.tool(permissions=['read:submissions']) for get_submission). Enable least-privilege agent configuration and audit logging.