Server has well-structured tool definitions with complete schemas, clear descriptions, and proper parameter typing. All 6 tools include input schemas with type constraints and descriptions. Tool annotations (readOnlyHint, destructiveHint, idempotentHint) are correctly applied. However, descriptions could be more concise (many exceed 200 chars), and output schemas are not explicitly documented. Error handling guidance is absent from tool descriptions. Naming follows verb_noun convention consistently (edgar_search, edgar_get, fred_search, fred_get, market_get).

• Document output schemas for all tools. Add a 'Returns' section to each tool description specifying field names, types, and structure. Example: 'Returns: {"cik": string, "company_name": string, "ticker": string}'.
• Condense tool descriptions to 50-200 characters. Current descriptions are verbose. Example: 'edgar_get_filings' could be 'Fetch SEC filings for a company by CIK. Filter by form type (10-K, 10-Q, 8-K, DEF 14A). Returns filing dates and URLs.' (120 chars vs current 280).
• Add error handling guidance to descriptions. Example: 'If CIK is invalid, returns error. Use edgar_search_company to find the correct CIK.' This guides LLM recovery without requiring separate error documentation.
• Clarify pagination and result limits. For fred_search_series and fred_get_series, state: 'Returns up to {limit} results. If more exist, use offset parameter to paginate.' (if offset is supported) or 'Results are capped at {limit}; use multiple calls with different queries to retrieve more.'
• Add data freshness and idempotency notes. For market_get_quote, add: 'Non-idempotent, stock prices change during market hours. Retry may return different values.' For market_get_overview, add: 'Idempotent, fundamentals update quarterly.'