FastMCP server for Kai - Economist data backbone. Exposes tools for SEC filings, news streams, and financial calculations used by specialist investment analysis agents (Charlie, Delta, Gamma).
The server defines 4 tools with explicit input schemas and descriptions. However, significant gaps undermine quality: (1) descriptions lack LLM-optimized guidance on WHEN to use each tool vs. similar tools; (2) parameter descriptions are generic and often miss constraints/ranges; (3) no output schemas are documented, agents cannot predict response structure; (4) error handling is absent, no recovery guidance or classification; (5) tool naming follows verb_noun convention but descriptions do not answer the three core questions (what, when, why). The descriptions average ~150 chars, which is below the ideal 50-200 char range for LLM optimization. All tools are READ_ONLY, which is positive, but the lack of actionable error context and missing output schema documentation are critical gaps for production use.
Calculate risk metrics including volatility, Sharpe ratio, and moving averages. Used by the Valuation Specialist (Gamma) for risk assessment.
Calculate comprehensive valuation metrics including PE ratio, PEG ratio, and optionally DCF valuation. Essential for the Valuation Specialist (Gamma).
Fetch SEC filing data (10-K, 10-Q, 8-K) for a company ticker. Returns recent filings with metadata and document URLs. Used by the Fundamental Analysis Specialist (Charlie) for deep business research.
Fetch recent news headlines for a stock ticker. Returns articles with titles, URLs, publication dates, and sources. Used by the Sentiment Analysis Specialist (Delta) to gauge market narratives.
No output schemas documented. Agents cannot predict what fields will be returned, forcing them to reason about response structure and increasing hallucination risk.
Error handling completely absent. No guidance on retryability, recovery steps, or error classification. If a tool fails (network timeout, invalid ticker, insufficient data), the agent receives no actionable recovery path.
Parameter descriptions lack specificity and constraints. Examples: 'days' param in get_news_stream has a range (1-30) but no explanation of what happens if the ticker has no news in that window. 'count' in fetch_filing_data lacks guidance on what fields are returned or how to interpret multiple filings.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Tool descriptions do not answer 'WHEN to use', they describe what the tool does but not when to invoke it instead of an alternative or whether prerequisites must be met first. E.g., calculate_valuation_metrics assumes price, eps, and growth_rate are known; no guidance on obtaining them if the agent lacks this data.
No pagination or result limits guidance in fetch_filing_data and get_news_stream descriptions. If a company has hundreds of filings or thousands of news articles, the tool may return unbounded results, exhausting context window.
calculate_risk_metrics accepts optional 'returns' parameter but does not explain the relationship to 'prices' or what happens if both are provided (or neither). Undocumented parameter relationships cause silent misuse.
Tool descriptions lack domain context for selection. A new user or agent cannot distinguish when to call fetch_filing_data vs get_news_stream vs calculate_valuation_metrics. Descriptions should explain the analysis phase each tool supports (fundamental vs sentiment vs valuation).