MCP server exposing Yahoo Finance API tools for stock quotes, historical data, options chains, and financial analysis
Yahoo Finance 2 MCP server has decent structure with 12 well-named tools and visible schema definitions, but falls short of production grade due to incomplete input schema visibility, generic parameter descriptions, and missing output schema documentation. All tool names follow verb_noun pattern (quote, search, chart, etc.) which is excellent. However, only a sample of src/mcp/server.ts is provided, limiting verification of full schema completeness. Tools are READ_ONLY (good for safety), but descriptions lack depth about WHEN to use each tool and dependencies between them. Parameter descriptions mention 'optional' frequently but lack specifics on constraints, ranges, or formats. No evidence of error handling guidance or recovery paths.
Fetch chart-ready historical price, dividend, split, and earnings data for a symbol.
Get financial statement time series data such as financials, balance sheet, or cash flow data.
Get historical price, dividend, or split records for a symbol.
Get Yahoo Finance analyst insights, reports, significant developments, and event data for a symbol.
Get options chain data including calls, puts, strikes, expirations, and quote data.
Get real-time or near real-time quote data for one or more Yahoo Finance symbols.
Get quote data for a single symbol through yahoo-finance2's debounced quote combiner.
Parameter descriptions are generic and lack specificity. Examples: 'Optional quote options such as fields' does not explain WHAT fields are available, what values they accept, or consequences of including/excluding them. 'Optional query options passed to yahooFinance.quoteSummary()' assumes LLM knows the underlying API.
No output schema documentation visible. Tools declare what they return (e.g. 'Get real-time quote data') but do not document the structure of the response object, field types, or which fields are always present vs optional. LLMs cannot reason about downstream tool chaining or data extraction without this.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2025-06-18+ | v2 |
Get quote summary data modules such as price, summaryDetail, assetProfile, and secFilings.
Get related and similar stock recommendations for one or more symbols.
Run a predefined Yahoo Finance stock screener such as day_gainers, day_losers, or most_actives.
Search Yahoo Finance instruments, companies, and related news by symbol, name, or keyword.
Get Yahoo Finance trending symbols for a region such as US, GB, or DE.
Parameter type constraints and formats not fully specified in descriptions. For example, 'region' in trendingSymbols accepts 'US', 'GB', 'DE' but description does not enumerate valid values. 'period1' in chart is marked 'Must include' but no format (timestamp? ISO date? epoch?) is documented. LLMs hallucinate formats when constraints are implicit.
Dependency and composition hints missing. Example: 'quoteCombine uses debounced quote combiner' is mentioned but never explains WHEN an agent should prefer quoteCombine over quote, or what the debouncing trade-off means (latency vs throughput). Tools returning symbol or company names do not indicate whether those names can be passed to other tools.
No visible error handling guidance or recovery paths. If a symbol is invalid, if a region code is unknown, or if the API rate limit is hit, no documentation explains what error the LLM will see or what to do next (retry? switch to different symbol? check format?).
Pagination not documented. Tools like 'search', 'screener', and 'trendingSymbols' may return variable-length lists. No description explains whether results are paginated, how to request the next page (offset, cursor, limit), or what the maximum result count is. LLMs can inadvertently request thousands of records.
Incomplete source visibility prevents full schema verification. Only a truncated excerpt of src/mcp/server.ts is shown (interface and first few lines of looseObjectSchema). Cannot verify that all 12 tools are explicitly registered, that input schemas include all required fields, or that validation is present. Per Hard Scoring Rules: if full schemas are not visible, schema score is capped.