The SEC EDGAR MCP server provides 21 tools with reasonable coverage of SEC filing and financial data retrieval. All tools have descriptions and most have properly typed input schemas. However, several issues prevent a higher score: (1) Output schemas are largely undocumented, the source shows tool definitions but no explicit documentation of return structures, forcing LLMs to infer results; (2) Error handling is minimal, no recovery guidance, categorization, or actionable error messages visible; (3) Several tools accept overloaded 'identifier' parameters (ticker OR CIK) without clear guidance on resolution priority or fallback behavior; (4) Parameter descriptions lack detail on constraints, formats, and edge cases (e.g., what constitutes a valid XBRL concept name? How are dates formatted?); (5) Some tool names could be more precise (e.g., 'analyze_insider_sentiment' is vague about what 'sentiment' means, bullish vs. bearish signal? Volume analysis?). The server correctly uses structured input schemas with types and descriptions, and naming generally follows verb_noun convention. Tools are well-scoped (one responsibility each). However, the lack of visible output documentation and weak error handling place this in the 'good but incomplete' category.

• Document output schemas for all tools. For each tool, add an explicit schema section describing the structure, field names, types, and semantics of returned data. Example: 'Returns: {"cik": "string (10-digit zero-padded)", "ticker": "string (1-5 uppercase letters)", "company_name": "string", "sec_url": "string (URL)"}'. This enables LLMs to parse results reliably.
• Add comprehensive error handling with recovery guidance. Wrap tool implementations in try-catch blocks and return structured error responses: {"error": true, "code": "TICKER_NOT_FOUND", "message": "Ticker 'XYZ' not found in SEC records.", "recovery": "Try search_companies(query='XYZ') to find similar companies.", "alternatives": ["XYZM", "XYZU"]}. This lets LLMs self-correct without wasting calls.
• Clarify the 'identifier' parameter resolution logic. Add a parameter description: 'Pass a ticker symbol (e.g., "AAPL") or a 10-digit CIK (e.g., "0000320193"). Ticker lookup is attempted first; if not found, the value is treated as a CIK.' Alternatively, split into separate parameters: ticker (string, 1-5 uppercase) and cik (string, 10-digit). This removes ambiguity.
• Rename 'analyze_insider_sentiment' to 'analyze_insider_trading_sentiment' and clarify the output. Update description: 'Analyzes insider trading activity over the specified period to infer sentiment: bullish (net buying), bearish (net selling), or mixed. Returns buy/sell ratios, volume trends, and confidence score.' Include concrete output fields in schema.
• Add min/max constraints to numeric parameters. For example, 'days: integer, range 1 - 365, default 30' or 'limit: integer, range 1 - 100, default 50'. This prevents LLMs from passing unreasonable values like days=0, limit=999999, or offset=-100.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry