This stock analysis MCP server provides 6 tools with complete input schemas and descriptions in Korean. Naming follows verb_noun convention appropriately (get_*, analyze_*). Descriptions are substantive (150-250 chars average) and explain both what each tool does and when to use it. However, OUTPUT SCHEMAS ARE COMPLETELY UNDOCUMENTED, the server returns JSON but does not publish what fields agents should expect. Error handling is minimal (generic try-catch with 'error' field only). No parameter descriptions for numeric constraints like 'days' min/max are enforced beyond JSON Schema. Tool composition is reasonable for a domain-specific server, though some tools could be split (e.g., analyze_chart_pattern returns both patterns AND 60-candle OHLCV data, mixing concerns). The server is STDIO-only, which hard-caps protocol readiness at 50.

• Document the output schema for each tool. For example, get_price_data should declare: 'Returns {"ticker": string, "prices": [{"date": ISO8601, "open": number, "high": number, "low": number, "close": number, "volume": number}], "latest_price": number, "change_pct": number}'. This enables agents to extract and chain data correctly.
• Add structured error responses with error codes and recovery hints. Example: Instead of {"error": "ticker not found"}, return {"error_code": "ticker_not_found", "error_message": "KRX ticker '999999' does not exist.", "recovery": "Try search_tickers() to find the correct code. Popular tickers: 005930 (Samsung), 000660 (SK Hynix)."}.
• Add natural-language descriptions of constraints to parameter descriptions. E.g., for 'days' in get_news_sentiment, update description to: 'Number of past days to search (1 - 30, default 7). Use 1 - 7 for recent momentum, 7 - 30 for trend analysis.'
• Split analyze_chart_pattern into two tools: 1) analyze_chart_pattern (returns only pattern name, confidence, signal) and 2) get_ohlcv_history (returns 60-candle raw data). Agents can then choose whether they need patterns alone or the underlying candles.
• Provide English translations of tool descriptions (keep Korean as primary but add English version) for international discoverability. Example: Tool description field could use '|' separator: 'Korean description | English description'.
• Implement per-item error reporting for multi-step operations. If get_news_sentiment fetches 10 articles and parsing 2 fails, return {"articles": [success items], "failures": [{"article_id": ..., "reason": ...}]} rather than a complete failure.