Server has 4 tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling. Tool names follow verb_noun convention (search_, discover_, analyze_), which is good. However, descriptions lack actionable context for LLM selection, many parameters are underdocumented, and output structures are not formally specified. No error recovery guidance. Schemas present but incomplete, several parameters lack descriptions or constraints. This is typical of community servers landing in the D range.

• Document output schemas for all tools. For search_facebook_ads, specify: {ads: [{id, creation_time, spend, impressions, ...}], total_ads: int, brand: string}. Use JSON Schema or TypedDict in docstrings.
• Add parameter constraints and enums. Replace 'ad_type: str' with enum ['ALL', 'POLITICAL_AND_ISSUE_ADS', 'HOUSING_ADS', ...]. Replace 'performance_metrics: array' with enum of valid metric names and document default selection.
• Expand tool descriptions to 100-200 chars explaining WHEN to use each tool. E.g. 'search_facebook_ads: Find active ads from a specific brand in a country. Use this to analyze competitor creative strategy and spending patterns. Returns up to 50 ads by default.'
• Add error recovery guidance. When API fails, return: {'error': 'Facebook API rate limited. Retry in 60 seconds or check token validity.', 'retryable': true, 'next_step': 'Wait and retry, or verify FACEBOOK_ACCESS_TOKEN is set.'}
• Move Facebook token to environment variable only (FACEBOOK_ACCESS_TOKEN). Remove --facebook-token CLI arg. Document in README that token must be set before server starts, never passed at runtime.