lit-forge MCP provides 19 financial and market analysis tools with generally adequate schemas and descriptions, but exhibits significant gaps in parameter documentation, output schema clarity, and error handling guidance. Tool naming is consistent and appropriate (verb-noun patterns like get_, simulate_, calculate_, plan_). Descriptions are present for all tools and range from moderate to good length (e.g., 'NISA (つみたてNISA) 積立シミュレーション' is 32 chars; 'Get market thermometer' is descriptive). However, the server has critical weaknesses: (1) Many parameters lack descriptions entirely or are minimally documented. For example, in simulate_nisa, 'annualGrowthRate' has a description but critical context about expected range (0-1 vs percentage) is implicit. (2) Output schemas are not explicitly documented in the source code, handlers return data without declared response structures. (3) Error handling is not visible in tool definitions, no indication of what errors might be thrown or recovery guidance. (4) Some tools accept opaque array parameters (e.g., 'indicators' in get_market_snapshot, 'series' in get_us_macro_latest) without documenting valid values or enums. The positive: tool names are action-verbs (get_, simulate_, calculate_, plan_), input parameters are typed in JSON Schema, and the domain is well-scoped (financial planning and market data).

• Add explicit JSON Schema output type definitions for each tool. Document expected fields, types, and example values for responses. Use a TypeScript-to-JSON-Schema generator (e.g., ts-json-schema-generator) to keep definitions in sync with runtime types.
• Convert string parameters that accept a fixed set of values to enum types. For 'period' in get_market_thermometer, declare enum: ['1d', '1w', '1m', '3m', 'ytd', '1y']. For 'country' in get_economic_events_today, declare enum: ['US', 'JP', 'EU', ...] with explicit list.
• Add bounds and constraints to all numeric parameters. Example: For 'monthlyAmount' in simulate_nisa, update description to 'Monthly savings amount in yen (1 to 10,000,000). Example: 100000 for ¥100k/month.' Include min, max, or pattern constraints in the JSON Schema.
• Document what each boolean flag does. For 'includeHistory' in get_quote, state: 'If true, response includes daily OHLC data for the past 252 trading days. Increases response size and latency by ~500ms.' This helps LLMs decide whether to request it.
• Add error recovery guidance to tool descriptions. Example for get_quote: 'Returns error if symbol is invalid. If you received an error, try search_symbols() with a partial company name, or confirm the exchange code (NASDAQ, NYSE, etc.)'
• Declare result limits in descriptions. For tools returning lists (get_geopolitical_calendar, get_central_bank_speakers), state: 'Returns up to 100 upcoming events. Use startDate/endDate to narrow results.' Add pagination fields (nextCursor, hasMore) to responses.