wolfjkd/tradex-hub

This page was translated automatically. Technical details may differ from the original. View the original

68 stars 23 forks Quality 71 Trend 28

A-share AI Financial Intelligence Decision Platform · 129 MCP tools · eltdx Tongdaxin protocol + akshare + Tonghuashun + Dongcai + local data · SmartRouter multi-source downgrade routing · Free zero-authentication

Overview

AI Financial Intelligence Decision Platform AKShare Packaging · eltdx Tongdaxin Protocol · astock_signals Signal Module · Quantitative Computing Engine · SmartRouter Full Routing · Local MCP Server · 129 tools Provided for AI Agents (WorkBuddy / Claude Code / Cursor). (129 tools): - : Centrally obtain data via SmartRouter (market/finance/valuation/industry/news/macro/daily limit/Dragon and tiger list/five-tier order book/order/intraday/theme/short-term indicators, etc.), including 31 tools for eLTX Tongdaxin protocol - (8 tools): 6 technical indicator calculations, 2 performance indicator calculations - (16 tools): 3 trading signal generation, 2 multi-factor analysis, 2 conditional stock selection, 5 system diagnostics, 3 comprehensive analysis, and 1 technical analysis engine : - : 78 data types, 102 data sources registered (37 types), divided into three tiers by ban risk - : L1 tool collects data uniformly via SmartRouter.route(), automatic health scoring/downgrade/fault isolation - : The first main source for market data, 31 tools covering five order books/call auctions/individual trades/F10/intraday/candlestick/panoramic profiles/short-term indicators/themes/categorized markets, etc.; Added resident connection manager eltdx_stream.py (cursor incremental polling enables quasi-real-time five-level market sizes). 3.x is the Rust override kernel (native wheel, cp310-abi3) - First tier (eltdx/Tencent/local vipdoc, no IP banning) + Second tier (Tonghuashun/Sina/CNINFO/Cailian Press, low risk) + Third tier (Dongcai push2/push2ex/slist, unique data only + rate limiting and anti-blocking) - :p ython -m tradex.dashboard (port 8765), visualizing data source health/routing/tool distribution; MCP tool get_data_source_dashboard can be queried in Agent dialogue : Global direct connection (clearing proxy environment variables when importing, domestic data sources do not use proxies) + new data sources (Tonghuashun 4 interface / Dongcai slist section ownership / Dongcai traffic limit and anti-ban / real-time number of rising and falling companies / industry gains / Tongdaxin local data) + 2 local data MCP tools, total of 127→129 tools.

Overview (machine translated)

AI Financial Intelligence Decision Platform

Provides AI Agents with a unified A-share data + quantitative computing + decision support interface

Project Introduction

TradeX-hub is a locally running financial data middle platform that packages A-share data scattered across multiple sources (Tongdaxin, Tencent Finance, Eastmoney, Tonghuashun, Sina, CNINFO, Cailian Press, AKShare, etc.) into 168 MCP tools + 58 REST endpoints (baseline after local source access on 2026-10-13), providing AI Agents (WorkBuddy / Claude Code / Cursor / Dify / LangChain) and front-end Kanban (TradeX) for one-stop calls.

Core Positioning: Allows AI Agents to not worry about where data comes from, how to reverse crawl, or how to downgrade — just tune a unified interface, tradex-hub automatically selects the optimal data source, performs automatic health checks, and automatically isolates faults.

Core Competencies

Ability	Description
Unified Access	A set of interfaces serves both the MCP client (AI Agent) and HTTP client (Kanban/Frontend), sharing the same business logic layer
Multi-Source Fusion	80 data types, 103 data source instances, divided into three tiers by ban risk, automatically downgraded
Quantitative Calculation	Technical indicators (MACD/KDJ/RSI/BOLL), performance indicators, multi-factor analysis, conditional stock selection
Decision Support	Trading signal generation, comprehensive diagnostics (individual stocks/markets/technology), Dragon and Tiger Rankings, Limit-Up Revelations
Observability	Prometheus metrics, monitoring dashboard, data source health table, endpoint P95 performance, slow query logs
Access-Friendly	Triple-layer token bucket rate limiting + dual access log writing + TypeScript/Python dual-language SDK

Three-layer capability model

Dual-protocol coexistence architecture

tradex-hub exposes two protocol interfaces simultaneously, sharing the same business logic, with callers selectable as needed:

Protocol	Applicable Scenarios	Interface Form
MCP	AI Agent Conversational Invocation (Tool Description Understandable by LLM)	JSON-RPC over HTTP, 168 Tools
REST	Kanban / External System / Custom Scripts	Standard HTTP + JSON, 58 endpoints, parcel response `{code, data, msg}`

The data returned by both protocols is exactly the same (only serialization differences: MCP returns JSON strings, REST returns packet dicts), and contracts are strictly aligned.

This page was translated automatically. Technical details may differ from the original.

Install

This server does not publish a one-line install command.

Open the repository installation guide New to MCP? Start with the mcpskill.org guide →

Configuration

{
  "mcpServers": {
    "tradex": {
      "command": "python",
      "args": ["-m", "tradex", "--http"],
      "env": {
        "TRADER_HUB_MODE": "http"
      }
    }
  }
}