Server has 5 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-100 chars, mostly adequate), and input schemas are present with type definitions. However, parameter descriptions are sparse or missing entirely (geo, timeframe lack context), output schemas are undocumented, and error handling provides no recovery guidance. No tool annotations (readOnlyHint/destructiveHint). The proxy_refresh tool is destructive but lacks confirmation or dry-run support. Tool naming is clear (verb_noun pattern), but composition could be tighter, proxy_status and proxy_refresh are operational concerns mixed with data tools.

• Document output schemas for all tools. Add 'outputSchema' field to each tool definition with typed fields (e.g. timeline: array of {date: string, values: object}, regions: array of {region: string, value: number}). This enables LLMs to extract and chain results.
• Expand parameter descriptions with format, constraints, and examples. E.g. geo: 'ISO 3166-1 alpha-2 country code (US, RU, GB) or empty string for worldwide'; timeframe: 'Format: today 1-m|3-m|12-m|5-y (default: today 12-m)'.
• Add tool annotations. Mark compare_keywords, get_related_queries, get_interest_by_region with readOnlyHint: true. Mark proxy_refresh with destructiveHint: true and idempotentHint: false to signal blocking behavior.
• Implement structured error responses. Replace generic 'Error: {message}' with categorized errors: {error: 'invalid_keyword', message: 'Keyword too long (max 100 chars)', suggestion: 'Try a shorter keyword'}. Include retryable flag.
• Add confirmation pattern to proxy_refresh. Return {requiresConfirmation: true, action: 'refresh_proxies'} on first call; require a confirm_action parameter on retry. Prevents accidental blocking calls.