Finance

China A-Share Market Data via AI Agents (Tushare)

Use an AI agent to pull China A-share daily bars, index data, daily fundamentals and intraday data from Tushare in natural language — no interface application or maintenance. Learn what data you can fetch and how to do it in one chain.

China A-Share Market Data via AI Agents (Tushare)

Tool names in this guide are for reference only and may have changed or been renamed. Please use the latest tools in the tool marketplace when implementing.

Key takeaways

  • Fetch A-share daily, index, fundamentals and intraday data in natural language — no API to build
  • Let the agent compute indicators or charts right after fetching
  • This guide covers data access only — no forecasts or investment advice

Tools used

Why pulling A-share data is such a hassle

Getting first-party A-share data into your code usually means applying for a data interface, reading the docs, figuring out each endpoint's fields and parameters, then writing scripts to pull, store and align it — one endpoint for daily bars, another for indices, another for valuation metrics. There are calling conventions to respect, field names differ, and maintenance isn't trivial. By the time the data finally lands, most of the effort is spent and you're still far from analysis.

For quants, research desks and individual investors, the real goal is to compute metrics, compare and chart — not to wrestle with endpoint fields. Pulling exactly the data you need in natural language removes that friction.

Note: this guide only covers what data you can fetch and how. It contains no market predictions or investment advice.

The workflow in four steps

Turn "confirm the ticker → fetch → organize" into one reproducible loop, driving the Tushare toolset from a single agent in one instruction:

  1. Tell the agent the tickers and period you care about.
  2. Confirm the ticker, then fetch its daily bars.
  3. Fetch an index as a benchmark, plus daily fundamentals.
  4. (Intraday) fetch intraday bars, and have the agent organize everything into a table or basic metrics.

How to do each step

The whole chain runs through natural-language instructions in any MCP-capable agent — no applying for interfaces, no fetch scripts.

Step 1: Confirm the ticker

Tell the agent which tickers and which period, using a placeholder ticker and date range as in the examples below.

Step 2: Confirm the ticker and fetch daily bars

Say: use ae to confirm this ticker, then fetch its daily bars. It first uses a stock-basics tool (such as ae_tushare_stock_basic) to verify the code and name, then a daily-bars tool (such as ae_tushare_daily) to fetch the daily series.

Step 3: Fetch an index benchmark and fundamentals

Have the agent use an index-daily tool (such as ae_tushare_index_daily) to pull an index series as a benchmark, then a daily-fundamentals tool (such as ae_tushare_daily_basic) for valuation-type metrics. That gives you the stock, the benchmark and valuation together.

Step 4: Fetch intraday and organize

For a finer intraday view, use an intraday-data tool (such as ae_tushare_rt_min). Finally, have the agent organize it all into a table or compute a few basic metrics for your follow-up analysis.

Aspect Apply for an interface + write scripts Fetch in natural language via the agent
Ramp-up Read docs, map fields, write code State what you need in one sentence
Multi-endpoint Pull each endpoint separately Fetched in one chain
Maintenance Track field/convention changes Handled on the platform side
After fetching Align it yourself Agent organizes into a table

How to get started

  1. Sign up for AgentEarth — new users get 1,000 free credits.
  2. In agents like WorkBuddy or Trae, search for "AgentEarth" and connect in-app; or paste a one-line MCP config into Claude Code / Cursor / VSCode.
  3. Just say use ae to fetch this ticker's daily bars, plus an index as a benchmark, and the agent picks the tools and runs them.

FAQ

Which periods are supported? There are daily-bar tools (such as ae_tushare_daily and ae_tushare_index_daily), and intraday-data tools (such as ae_tushare_rt_min) for the trading session.

Can I fetch an index as a benchmark at the same time? Yes. In the same chain that fetches a stock's daily bars, use an index-daily tool to pull an index series for comparison.

What about data latency? It varies by tool type — some are real-time/near-real-time, others are end-of-day; refer to each tool's description in the tool marketplace. This guide gives no specific latency figures.

Why fetch the data through an agent instead of calling the interface directly? Calling the raw interface means applying for access, reading docs, mapping each endpoint's fields and writing and maintaining pull scripts — and repeating that for daily bars, indices, fundamentals and intraday separately. Fetching in natural language lets you chain those endpoints in one request and get the results back already organized, so you spend your time on analysis rather than on plumbing. You still decide what to fetch; the agent just handles the retrieval and assembly.


Sign up for credits · Browse 2,400+ tools · Integration guide