Server has 4 tools with complete input schemas and basic descriptions. All tools follow verb_noun naming (calcom_add/update/delete/list_appointment). Descriptions are present but generic (avg 150 chars, below production baseline of 194). Parameter descriptions are minimal and lack format constraints. No output schema documentation. No error handling guidance. Rate limiting is implemented server-side but not exposed to LLM. Missing idempotency hints for write operations.

• Document output schemas for all tools. Example for calcom_add_appointment: 'Returns {id: number, eventTypeId: number, startTime: string, endTime: string, status: string, attendees: [{name, email}], notes?: string}'. This enables agents to chain calls (e.g., pass returned booking ID to update_appointment).
• Add format constraints to parameter descriptions. E.g., 'startTime: ISO 8601 string (YYYY-MM-DDTHH:mm:ss.sssZ), must be in UTC, must be in the future'. Replace vague 'ISO format' with explicit rules.
• Implement dry-run or confirmation for destructive operations. Add optional 'confirm: boolean' parameter to delete_appointment and update_appointment. Return a preview of changes before committing.
• Enhance error messages with recovery guidance. E.g., instead of throwing 'Rate limit exceeded', return {error: 'Rate limit exceeded. Please retry in 5 seconds or reduce request frequency.', retryable: true, retryAfterSeconds: 5}.
• Add pagination to list_appointments. Include optional 'limit' (default 20, max 100) and 'offset' (default 0) parameters. Return {appointments: [...], total: number, hasMore: boolean} to enable agents to iterate through large result sets.