dsh-industry-graph-mcp

In A-share research or strategy conversations, an agent often needs to answer questions like "which industry does this stock belong to, where does it sit in the supply chain, and which concepts does it touch." Guessing from model memory is unreliable and hard to trace. This plugin turns A-share industry chains, Shenwan sectors, and concept boards into a local knowledge graph that the LLM can query in real time, returning structured and traceable answers instead of vague recall.

Seven tools cover the core workflow: look up a stock industry mapping, find comparable peers, view supply-chain structure with a highlighted node, list concept-board constituents, intersect two concepts for cross-selection, fuzzy-search covered tickers, and get dataset statistics. The seed dataset covers 31 Shenwan Level-1 industries, 52 concept boards, 14 industry chains, and 181 unique stocks, intentionally small and precise yet extensible via gen_graph.py to incorporate custom lists.

It is zero-dependency (pure Node ESM, no npm install), fully offline, and requires no API key; all data ships with the repository. It suits A-share strategy analysts, sector comparability researchers, thematic-trackers, and anyone who wants structured financial domain knowledge inside a dsh conversation.