How to choose between Tushare and AKShare: fees, the free tier, and when AKShare breaks
A free scrape, or a points-gated dataset. Neither is a commercial data licence.
Short answer: neither is simply better. To get personal research running, use AKShare. It is free and needs no account. The cost is that an interface can fail when a source site changes its pages. For a maintained dataset, or for financial statements and index constituents, use Tushare Pro: register, take a token, and access follows your points. The free 120-point tier returns non-adjusted daily bars only; those other interfaces start at 2,000 points. Neither library allows commercial use.
How to choose: If you are just starting and want one pipeline to run, use AKShare. If you only research China A-share daily bars, AKShare is enough; if you want a more stable source, Tushare's free 120-point tier can return non-adjusted daily bars. If you need financial statements, daily indicators, or index constituents, start at Tushare's 2,000-point tier. If you are building a product or serving data to other people, neither library allows that: find a source with a commercial licence. If the market is mainly US stocks, use the comparison of yfinance and EODHD APIs.
Who it is for: Someone doing personal or academic research who needs China A-share data in Python, and who is choosing between a free library with no account and a points-gated dataset that Tushare maintains. It also fits someone who will use both: AKShare to get the download, the database, and a backtest running, then Tushare to check the series that matter.
Who it is not for: Someone building a product or offering the data to other people. Tushare's data service agreement is a personal, non-commercial licence. AKShare's code is MIT, but its data notice limits the data to academic research and rules out commercial use. This page is not the four-library comparison of yfinance, AKShare, OpenBB, and EODHD, not the how-to-choose tool order, and not the site FAQ.
The same comparison in Chinese is Tushare and AKShare (Chinese). The order of data, then a backtest, then paper trading, then an agent is a different article: how to choose AI quant tools. yfinance, AKShare, OpenBB, and EODHD are also a different article: how to choose a data source. None of those is a language version of this page.
This site has no commercial relationship with Tushare or AKShare. The facts below were checked on 21 Oct 2026 against Tushare's documentation and data service agreement, and against AKShare's GitHub repository. Prices and point rules change. Use Tushare's official page. Catalog entries: Tushare and AKShare.
01 · One table
| Tushare Pro | AKShare | |
|---|---|---|
| How you get data | Register, then call the API with a token | pip install. No account |
| Price | Access follows points. The free tier is non-adjusted daily bars only | Free |
| Where the data comes from | Collected, cleaned, and maintained by the Tushare community | Fetched live from Eastmoney, Sina Finance, the exchanges, and other public sites |
| Stability | The community maintains the data. It does not depend on scraping a third-party page | If a source site changes its pages, the interface can fail. Upgrade often |
| Call limits | Points set calls per minute and the daily total | No published fixed cap. Calling too fast can be limited by the source site |
| Commercial use | No. The agreement is a personal, non-commercial licence | The code is MIT. The data is for academic research only and is not for commercial use |
| AI tools | An official MCP server and Skills | An offline interface search, ak.search |
02 · Tushare now means Tushare Pro
Tushare is a Python package for financial data. The service in use today is Tushare Pro (tushare.pro): register, take a token, then call the API with the Python SDK or over HTTP. The older site that needed no account, tushare.org, returned 404 on 21 Oct 2026. Snippets in old tutorials often no longer apply.
The package on PyPI is named tushare. It is BSD-licensed. On 21 Oct 2026 the latest release was 1.4.29. The official docs require a version greater than 1.2.10. You set the token, then initialize with pro_api().
03 · What Tushare charges: how points work
Tushare Pro uses points to decide which APIs you can call and how fast. The platform points note says registration gives 100 points and completing the profile adds 20. Referrals, code contributions, and introductory posts can add points too. You can also pay to join the official WeChat group and receive points, or sponsor the community for a points permission. Points last one year. Calling data does not spend them. There is no refund after a purchase.
This table is taken from the official points-and-frequency table (checked 21 Oct 2026; Tushare's page wins if it changes):
| Points | Per minute | Daily cap | What you can call | Published price |
|---|---|---|---|---|
| 120 | 50 | 8,000 | Non-adjusted daily bars only | 0 |
| 2,000+ | 200 | 100,000 per API | Whatever each API's point rule allows | 200 yuan a year |
| 5,000+ | 500 | No daily cap on regular data | Whatever each API's point rule allows | 500 yuan a year |
| 10,000+ | 500 | No daily cap on regular data. Special data: 300 per minute | Adds special data such as earnings forecasts and chip distribution | 1,000 yuan a year |
| 15,000+ | 500 | No total cap on special data either | The special-data permission | 1,500 yuan a year |
In short: the free 120 points are only enough for daily bars. On the official permission list, daily bars need 120 points. Weekly bars, monthly bars, adjusted prices, daily indicators (price-to-earnings and turnover, among others), the income statement, the balance sheet, the cash-flow statement, financial indicators, index daily bars, and index constituents all start at 2,000 points. Fundamental research usually means that tier.
A separate group of APIs ignores points and needs its own permission, including minute bars, news, and announcements. Each has its own price, on the same official table.
04 · Why AKShare is free, and what that costs
AKShare is an open-source Python library. The code is MIT-licensed. Install it and call it. No account. Its project overview says the data comes from public sources: when you call it, it fetches Eastmoney, Sina Finance, the exchanges, and similar sites, and returns a table. The README points readers to the author's community and a public WeChat account.
The official docs name three costs:
- An interface can fail without notice. The project says target sites often change their page format, so the interfaces have to be updated with them. Keep AKShare on the latest release. Some interfaces can also be removed for reasons outside the project's control.
- Calling too fast gets limited. The Q&A for timeouts says to lower the request rate or change IP. Pace a bulk download yourself.
- The data is not for commercial use. The code is MIT. The special notice says the interfaces and the data are for academic research only and must not be used for any commercial purpose.
A worked download of daily bars into a database, with a daily incremental update, is AKShare into SQLite. The same walkthrough in Chinese is AKShare walkthrough (Chinese). That walkthrough is not this comparison.
05 · Common errors
- AKShare says an interface does not exist (has no attribute): the Q&A says to confirm Python is 3.11 or newer, then upgrade AKShare. A renamed or removed interface looks the same. Check the changelog.
- AKShare times out or returns nothing: slow the requests and try again later. If it still fails, search that interface name in the GitHub issues. The source site has often changed its pages.
- import akshare fails: the Python is usually too old, or it is not 64-bit. The README requires 64-bit Python 3.11 or newer.
- Tushare says you have no permission: your points are below that API's threshold. Check the permission list above. Minute bars and news need a separate permission.
- Tushare says the token is invalid: copy it again from the account-and-token section. If you refreshed the token, the old one no longer works.
06 · Using both
A common setup is to get the pipeline running on AKShare first (download, store, and backtest), then move the series you will keep onto Tushare, or check them against Tushare.
Watch adjusted prices. The two libraries can adjust from different formulas, so the same stock on the same day can show a different forward-adjusted price. No official side-by-side of those formulas was found, so this page does not say which one is more accurate. Before you store the data, spot-check a few stocks that have paid a dividend or a bonus issue, and compare the two numbers. Mixing the series needs extra care.
07 · Pitfalls
- Do not use a shared token. Tushare's data service agreement says you must not provide the account to anyone else by transfer, rental, lending, sharing, or sale. Selling an account on a marketplace is a violation named in that agreement. The service may be suspended or terminated.
- Read the licence before you build a product. Tushare grants a personal, non-commercial licence, for non-commercial use only. AKShare's data is for academic research only and is not for commercial use. A service for other people needs data with a commercial licence.
- Check the rows before they enter a database. Look for missing trading days, duplicate rows, and how a halted day is stored. Do not backtest on dirty data.
08 · MCP and Skills
- Tushare: the official Tushare MCP is configured from the MCP Server section of the personal center, so Claude Code, Cursor, Trae, and similar tools can call the data. There are also official Skills. Calls still follow your point permission.
- AKShare: the project ships an offline interface search,
ak.search, which can match a plain-language description to an interface. The README says that is useful for programs driven by a language model. AKShare MCP projects elsewhere are usually third-party. Check where they come from before you use one.
A longer note on giving a research agent a market-data source is market data and an LLM API. The same article in Chinese is connecting a data source and an LLM API. That note is not this comparison.
09 · How to choose, by situation
- You are starting, and you want one pipeline to run: AKShare.
- You only research China A-share daily bars: AKShare is enough. For a more stable source, Tushare's free 120-point tier can return non-adjusted daily bars.
- You need financial statements, daily indicators, or index constituents: Tushare from the 2,000-point tier.
- You are building a product or a service for other people: neither library. Find data with a commercial licence.
- The market is mainly US stocks: see yfinance and EODHD APIs in how to choose a data source. Putting EODHD daily bars on the same SQLite table as an A-share pipeline is the EODHD walkthrough. That walkthrough is not this comparison.
Frequently asked questions
How to choose between Tushare and AKShare?
If you are just starting and want one pipeline to run, use AKShare. If you only research China A-share daily bars, AKShare is enough; if you want a more stable source, Tushare's free 120-point tier can return non-adjusted daily bars. If you need financial statements, daily indicators, or index constituents, start at Tushare's 2,000-point tier. If you are building a product or serving data to other people, neither library allows that: find a source with a commercial licence. If the market is mainly US stocks, use the comparison of yfinance and EODHD APIs.
Who it is for?
Someone doing personal or academic research who needs China A-share data in Python, and who is choosing between a free library with no account and a points-gated dataset that Tushare maintains. It also fits someone who will use both: AKShare to get the download, the database, and a backtest running, then Tushare to check the series that matter.
Who it is not for?
Someone building a product or offering the data to other people. Tushare's data service agreement is a personal, non-commercial licence. AKShare's code is MIT, but its data notice limits the data to academic research and rules out commercial use. This page is not the four-library comparison of yfinance, AKShare, OpenBB, and EODHD, not the how-to-choose tool order, and not the site FAQ.
Which is better, Tushare or AKShare?
It depends on the job. To get a personal research pipeline running, use AKShare: it is free and needs no account. For a maintained dataset, or for financial statements and index constituents, use Tushare Pro: register, take a token, and access follows your points. Many people use both. AKShare gets the pipeline working. Tushare checks the series that matter.
Does Tushare charge, and what is free?
Registration gives 100 points and completing the profile adds 20, so 120 points is enough to call the API for free. On the official points-and-frequency table, 120 points can return only non-adjusted daily bars, at 50 calls a minute and 8,000 calls a day. Weekly bars, adjusted prices, daily indicators, and financial statements start at 2,000 points. The published price for that tier is 200 yuan a year. Use Tushare's official table if the numbers change.
How do I get a Tushare token?
Sign in at tushare.pro, open the personal center from the top right, and copy the token from the account-and-token section. The token is the only credential for the API. If someone else has used it, refresh it on the same page. The old token stops working.
Can I use a shared Tushare token?
No. The data service agreement says you must not provide the account to anyone else by transfer, rental, lending, sharing, or sale. A violation can suspend or end the service.
Is AKShare free, and what keeps it running?
Yes. The code is MIT-licensed and needs no account. It reads public sites such as Eastmoney, Sina Finance, and the exchanges when you call it. The README points readers to the author's community and a public WeChat account. The project's own notice says the data is for academic research only and must not be used commercially.
What if AKShare fails to return data?
Upgrade to the latest release and confirm you are on 64-bit Python 3.11 or newer. A missing interface is usually a source site that changed its pages while an old release did not. For timeouts, the project's own Q&A says to call less often or change IP. If that fails, search the interface name in AKShare's GitHub issues.
Can I use Tushare or AKShare commercially?
Not directly. Tushare's data service agreement is a personal, non-commercial licence. AKShare's code is MIT, but the data notice says the interfaces and the data are for academic research only and must not be used for any commercial purpose. A product or a service for other people needs a source that grants commercial use.
This is not investment advice. Data can be wrong or late. Check it before you research with it. Tushare's point rules and prices follow its official page.