> Source: https://ashareapi.com/en/developers/  ·  Markdown version for LLMs / AI agents

Developers

# Two lines to a DataFrame
 The official SDKs wrap the HTTP endpoints into one-line method calls — no URL building, no retry handling, no response parsing.
 [Read the docs](/en/docs)[See pricing](/en/pricing)

 Install SHELL

 ```
pip install ashareapi
```
 ```
npm install ashareapi
```

 Up and running in 30 seconds

```
from ashareapi import AShareAPI

cli = AShareAPI() # free endpoints need no key
df = cli.kline("sh600519", "day", 30)
print(df.tail())
```

 Up and running in 30 seconds

```
import { AShareAPI } from 'ashareapi'

const cli = new AShareAPI() // free endpoints need no key
const rows = await cli.kline('sh600519', 'day', 30)
console.log(rows.slice(-5))
```

 Paid endpoints (financials / capital flow / dragon-tiger list / factor screening / backtesting) just need your key — everything else stays the same.

## Why use the SDK instead of raw requests
 Responses are DataFrames
 Most endpoints return tables; the SDK turns them into a pandas DataFrame. No pandas? It falls back to a list of dicts.

 Code formats handled for you
 sh600667 / 600667.SH / 600667 all work — the SDK normalises them.

 Clear error classes
 Auth / rate limit / upstream failure / empty result are distinct — empty is not failure, and each tells you what to do.

 Zero deps, full types
 The Node build uses only native fetch (no third-party deps) and ships TypeScript types with editor autocomplete.

 Fixed conventions
 K-line adjustment is fixed server-side — no double-adjusting.

 Nothing for you to maintain
 Multi-source switching, caching and monitoring live on the server; a failed fetch rotates to another source.

## What you can build with it
 The SDK does one thing: turn data into a structure you can use directly. The rest is up to you — these are the common cases.
 Notebooks / local research
 After `pip install ashareapi`, one variable is one table — plot it, compute factors, write it up.

 Dashboards / internal tools
 Call it server-side and feed the DataFrame or JSON into your frontend or BI stack.

 Strategy scripts / backtesting
 Data fetching stays inside the SDK; your strategy code only handles logic.

 Batch scanning
 Pull capital flow or K-lines for a whole batch at once — good for scanning a pool.

## Connecting to QMT / Ptrade
 Both terminals ship an older built-in Python (QMT 3.6 · Ptrade 3.5, some brokers 3.11) and enforce broker library whitelists and sandboxes, so third-party packages generally will not install. The workable pattern is "fetch outside, read files inside" — your strategy only reads local files and makes no network calls.
 ① Fetch outside (normal Python)
 Use the SDK to pull data and write local CSVs — this step runs outside the terminal.

 ② Read files inside the strategy
 QMT / Ptrade bundle pandas, so read the CSV directly — no third-party installs, no network calls.

 ③ Swap symbols by editing one list
 Data layer and strategy layer stay separate; strategy code does not change with the data source.

 Fetch outside

```
# fetch.py —— 在终端【之外】用普通 Python 跑（3.9+）
from ashareapi import AShareAPI

cli = AShareAPI("ct-你的KEY")
codes = ["sh600519", "sz000001"]

for code in codes:
 df = cli.kline(code, "day", 250) # 已返回 DataFrame
 df.to_csv("%s.csv" % code, index=False)
```

 Read inside strategy

```
# strategy.py —— 放进 QMT / Ptrade 的策略文件里（只读本地文件）
# ⚠️ 列名以 SDK 实际返回为准：date / open / last / high / low / volume / amount
import pandas as pd

def initialize(context):
 # 终端内置 pandas ⇒ 直接读 CSV，不装第三方包、不联网
 context.data = pd.read_csv("sh600519.csv")
 context.data["ma20"] = context.data["last"].rolling(20).mean()

def handle_data(context, bar):
 row = context.data[context.data.date == bar.time]
 if len(row) == 0:
 return
 ma20 = row.iloc[0]["ma20"]
 if bar.close > ma20:
 order_target_percent("600519.SS", 1.0)
```

 Why not run the SDK inside the terminal: the built-in version is below this package’s 3.9 requirement, and whitelists / sandboxes block installs. File exchange sidesteps both while keeping the strategy side dependency-free.

## Common methods
 The full list and parameters live in the docs; these are the ones you reach for first.
 |
| | Method | What it does

| | `quote(code)` | Realtime quote snapshot

| | `kline(code, period, count)` | K-lines (day / week / month / season / year + intraday)

| | `klineFull(code)` | Full history of daily bars in one call

| | `minute(code, days)` | Intraday (today / last 5 days)

| | `fund(code)` | Capital flow + dragon-tiger list + block trades + margin

| | `lhb(type)` | Dragon-tiger boards (institutions / hot money / active seats)

| | `finance(code)` | The three financial statements

| | `technical(code)` | MA / MACD / KDJ / RSI / BOLL

| | `screen(expr, preset)` | Factor screening

| | `backtest(opts)` | Quant backtesting

 [See the full endpoint list](/en/endpoints)

## Related
 [Docs / Quick start

 Auth, error codes, response shape](/en/docs)[Endpoint list

 Parameters and fields for every endpoint](/en/endpoints)[MCP

 Let AI Agents call it directly](/en/mcp)