MCP server providing tools to get realtime financial data from global financial markets (NASDAQ, NYSE, DOW, HKEX, Shanghai, Shenzhen, LSE, TSE, NSE India, etc.) with free API access and no key requirement
Single tool with moderate naming clarity but significant gaps in schema documentation, parameter descriptions, and output specification. The tool name 'get_stock_price_global_market' starts with a verb and is reasonably specific, but input parameter descriptions lack formal constraints (enums, ranges, formats). The schema shows basic structure but parameters lack type validation detail. Output is documented in the docstring as JSON samples, but no formal output schema is registered. Error handling is minimal (generic exception catches with plain text returns). The server provides one tool with incomplete definition quality below production baseline.
Get Public Available Stock Symbols from Global Marketplace
Parameter 'market' lacks enum constraint. Description lists examples ('HK', 'CN_MAINLAND', 'US', etc.) but does not declare them as enum values. LLMs will guess valid values and pass invalid markets, causing runtime failures.
Output schema not formally documented. Docstring contains sample JSON output, but no structured output schema is defined in the tool registration. LLMs cannot reliably parse multi-source outputs with varying field names (e.g., 'avg_price' vs 'current', 'percent' vs 'chg').
Parameter 'symbol_list' description includes example values (700, 1024, TSLA, MSFT, GOOG, SHEL, ULVR) but does not specify format constraints, length limits, or validation rules. LLMs may reuse exact examples literally or pass malformed symbols.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 40 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 37 | - | v1 |
Error handling returns plain text strings without recovery guidance. Errors like 'Error communicating with FinanceAgent API' or 'Unexpected error' provide no actionable next steps for the agent. No error categorization (retryable vs fatal).
No input validation before calling downstream FinanceAgent API. Symbol list could be empty, market could be invalid, and no pre-flight checks prevent malformed calls. Validation errors should return early with actionable feedback.
Tool description is 194 characters and describes general capability well, but lacks critical metadata: What format are symbols in per market? What does the tool return on partial failures (some symbols succeed, some fail)? Does it support pagination for large symbol lists?
Return type is 'str' (JSON-serialized). LLMs must parse JSON from string, increasing token cost and error surface. FastMCP should return structured Pydantic models or dicts, allowing the framework to handle serialization.
No timeout or rate-limiting protection. External API calls to morningstar.com, xueqiu.com, etc. could hang indefinitely. Agent would block waiting for response, potentially consuming service quota without bounds.
Prompt mentions multiple data sources (nasdaq.com, morningstar.com, xueqiu.com, etc.) but tool provides no control over which source is used for a given market. LLM cannot select preferred source or fallback if one is blocked.