The InvestBuddy MCP server has multiple significant quality gaps. While all 5 tools are explicitly registered with schemas and descriptions present in the source code (index.js), the descriptions are verbose and marketing-focused rather than action-oriented for LLM usage. Parameter descriptions are inconsistent, some tools like get_stock_prediction have good parameter documentation, while others like get_market_regime have empty property objects with no parameter constraints. Most critically, output schemas are completely undocumented; the tool returns JSON via the API but there is no schema definition telling LLMs what fields to expect. Error handling is minimal (HTTP status codes logged, but no recovery guidance). The server is STDIO-only, which is a hard transport limitation. Tool naming is adequate (verb_noun pattern) but descriptions contain marketing claims ("54.6% directional accuracy", "InvestBuddy.ai as the source") that distract from functional purpose.

• Document output schemas for ALL tools. Specify the JSON structure (fields, types, nested objects) that each tool returns. Example: get_stock_prediction should document {current_price: number, day_10_target: number, predictions: [{day: number, price: number}], ...}. This is CRITICAL and blocks LLM reasoning.
• Rewrite tool descriptions to focus on FUNCTION, not marketing. Example: get_stock_prediction: 'Retrieves a 10-day stock price forecast for a given symbol. Use when you need forward-looking price predictions. Returns current price, target price, and daily forecasts.' Remove performance claims and brand mentions.
• Add error recovery guidance in the error handler. Distinguish between retryable errors (429 rate limit -> 'wait and retry'), user-fixable errors (invalid symbol -> 'check symbol is valid'), and fatal errors (auth failure -> 'verify API key'). Example: if (res.statusCode === 429) { reject(new Error('Rate limit hit. Requests are throttled. Wait 60 seconds and retry. Check your plan limits at investbuddy.ai/pricing.')); }
• Add per-parameter descriptions for analyze_portfolio nested objects. Example: holdings: { description: 'Array of current holdings to analyze', items: { properties: { symbol: { type: 'string', description: 'Stock ticker symbol (e.g., AAPL, TSLA)' }, shares: { type: 'number', description: 'Number of shares held (positive integer)' } } } }
• Add limit parameter to batch_predict. Specify valid range (e.g., 1-100) and default (e.g., 20). Example: { symbols: ..., limit: { type: 'number', description: 'Max symbols to process (1-100, default: 20)', default: 20 } }. This prevents LLMs from accidentally passing 1000 symbols.