Tonghuashun opens its A-share data to AI agents and scripts

What it does Tonghuashun (HiThink) operates this official service for pulling A-share quotes, financial statements, indices, sector heat, and public fund data. It funnels everything through a single API key, offering REST endpoints, a Python SDK, a command-line tool, and hosted MCP servers for tools like Claude. A unified hithink-finance Skill sits on top and lets an AI agent pick whichever transport layer fits the job.

The interesting bit The README treats access mechanics as a first-class product: it is less a raw data dump than a vendor repackaging itself for LLM tool use. The standout piece is a local DuckDB cache—dubbed marketdb—that lets you run SQL against historical prices offline instead of repeatedly calling the remote API.

Key highlights

• Covers real-time snapshots, historical K-lines, financial reports, valuation ratios, limit-up/limit-down pools, and hot lists for A-shares.
• Ships a unified Agent Skill that auto-selects between REST, MCP, CLI, or Python SDK depending on the task and environment.
• Includes a local marketdb backed by DuckDB for incremental sync, SQL queries, and rights-adjusted price calculations.
• Explicitly excludes minute-level, tick, overseas, macro, and news data from the public offering.

Caveats

• Minute K, tick, overseas markets, macro data, and news announcements are explicitly outside the current scope.

Verdict Grab it if you need a legitimate, multi-interface source for China A-share data in quant research or AI agent workflows. Look elsewhere if you require tick-level granularity, overseas markets, or a fully free standalone data engine.