MCP server for financial data via yfinance
Finance MCP server has clear, action-verb naming (get_*, calculate_*, compare_*) and reasonable descriptions for all 5 tools. Input schemas are present and properly typed. However, several issues prevent a higher score: (1) parameter descriptions lack specificity about constraints and formats (e.g., 'ticker' param has no guidance on valid formats, case sensitivity is shown in code but not documented); (2) output schemas are not explicitly documented, responses are inferred from code but not declared to the LLM; (3) error handling returns generic exception messages rather than actionable recovery guidance; (4) no input validation documentation for edge cases (e.g., invalid ticker symbols, out-of-range count parameter); (5) some parameters have weak descriptions (e.g., 'ratios' array lacks clarity on what happens if an unsupported ratio is requested). The server follows basic tool patterns but lacks the polish and completeness expected of production-grade tools.
Compute valuation ratios for a stock from Yahoo Finance data.
Side-by-side comparison of 2-5 stocks.
Pull financial statements for a stock ticker.
Fetch recent news headlines for a stock.
Fetch the latest price data for a stock ticker.
Output schemas not documented. Tool descriptions do not specify what fields or structure the LLM should expect in responses. This forces the LLM to infer output shape from trial and error.
Parameter descriptions lack constraint details. 'ticker' parameter (all tools) does not document: case sensitivity (code shows .upper() but LLM doesn't see it), valid format, length limits, or what happens if ticker is invalid. 'count' in get_news says 'max 50' in description but code enforces 1-50, inconsistent.
Error responses are generic. All tools catch exceptions and return {'error': str(e), 'ticker': ticker}. This tells the LLM nothing actionable: no distinction between retryable vs fatal errors, no guidance on what went wrong (invalid ticker? API rate limit? network timeout?), no recovery path.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 44 | - | v1 |
calculate_ratio parameter 'ratios' array description is unclear. It says 'If omitted, computes all' but does not explain what happens if an unsupported ratio is included in the list. Code shows unsupported ratios return {label: r, value: None, note: 'unsupported ratio'} but this is not documented, LLM won't know this is a non-fatal partial success vs a failure.
compare_tickers validation constraint (2-5 tickers) is documented in code error path but not in parameter description. LLM must infer the constraint from error handling rather than reading the schema.
No documentation of response structure. For example, get_stock_price returns fields like 'current_price', 'previous_close', 'market_cap' but the description does not list these or explain which might be null. Agents cannot plan downstream use of the response without trial.