Wukong Quant Financial Agent

Problem

Quant research workflows often scatter single-stock views, limit-up ladders, macro readings, sector heat, and earnings signals across multiple feeds. Engineers and analysts may repeatedly assemble API calls, parse inconsistent fields, and infer data freshness before drawing conclusions. wukong-quant consolidates these actions into MCP tool calls so a conversation can directly request market data, analysis artifacts, and watchlist operations.

How It Works

• Deep analysis: after a user supplies a stock symbol, the skill calls mcp_wukong_quant_deep_analysis; the server waits for completion, usually 60-180 seconds, and returns investment advice, risk level, technicals, chip analysis, quant cross-checks, and horizon forecasts for full-text output.
• Market and theme analytics: supported features include the daily limit-up ladder, limit-up narrative, AI market summary, global market brief, hot-stock review, macro analysis, sector/concept review, and earnings signals; most reads come from caches or trading-day refreshes rather than triggering an LLM.
• Scoring and personalization: the skill can query index score rankings for hs300, a500, and other supported universes, inspect single-stock six-dimension scores, and manage watchlists; personalization endpoints require X-User-Token in request headers and return 401 when missing or expired.
• Boundaries: the skill depends on the wukong-quant MCP server and an internal key; analytics differ by freshness, source, and refresh timing, so post-market narratives, trading-day updates, and data-provider differences should be noted in the conclusion.