Server has 11 tools with complete schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention well (set_, get_, list_, analyze_, discover_, connect_, find_). Descriptions are present but often generic or lack actionable context. Parameter descriptions exist but many lack format/constraint details. No output schemas documented. Error handling is minimal, no recovery guidance or categorization. Missing tool annotations (readOnlyHint, destructiveHint). Several tools lack required parameter descriptions (e.g., 'format' in analyze_customer_health has no description of what 'detailed' vs 'summary' actually returns).

• Document output schemas for all tools. Example: analyze_customer_health should return {health_scores: [{customer_id, score, status, last_updated}], summary: string}. This enables LLMs to chain tools correctly.
• Add tool annotations: mark set_data_source and connect_to_airtable_base with destructiveHint=true; mark all read-only tools with readOnlyHint=true. This helps agents reason about side effects.
• Expand parameter descriptions with format details. For 'format' enum, explain: 'detailed returns full analysis with all metrics; summary returns only health score and status.' Similar for Airtable schema format.
• Add pagination to list_customers and discover_airtable_bases: accept limit (default 20, max 100) and offset parameters; return {items: [...], total: N, has_more: bool}.
• Add error recovery guidance in tool descriptions. Example: 'If customer_id not found, call list_customers() first to see available IDs.' This guides multi-step planning.