TDX Stock Data

Problem

TDX stock data is spread across quotes, K-lines, financial statements, filings, research reports, and macro series. When engineers wire these sources directly, they must manage tokens, parameter schemas, retries, and permission boundaries. This skill wraps the TDX MCP service as an agent-callable tool set, so conversational or automated workflows can query individual securities, screen conditions, and retrieve research material without manually navigating dashboards.

How It Works

It exposes 10 tools through one native HTTP MCP server, tdx-finance_qclaw, driven by mcporter. The main flow is:

• Initialization: after loading, it obtains the latest token, then checks and refreshes the MCP server configuration, avoiding stale or plaintext tokens.
• Routing: user intent maps to tools such as tdx_quotes, tdx_kline, tdx_api_data, tdx_screener, and wenda_report_query.
• Schema-aware calls: parameters are confirmed before invocation; for example, tdx_screener uses message for natural-language screening conditions rather than query.

Use cases include real-time quotes for A-shares, HK stocks, US stocks, and indices; K-lines; financial statements; shareholder changes; dragon-tiger list data; broker research ratings; company announcements; macro data; and natural-language screening such as limit-up or MACD crossover.

Boundaries And Cautions

The skill depends on TDX authorization and mcporter configuration. If calls return 401 or 403, re-run the token initialization flow. Returned data may be delayed, so outputs should include timestamps and should not be treated as investment advice. When a request is outside the current tool coverage or permission scope, the agent should state the limitation instead of guessing parameters or retrying blindly.