Alpaca API python client
alpaca-trade-api-python is a python library for the Alpaca Commission Free Trading API.
It allows rapid trading algo development easily, with support for
both REST and streaming data interfaces. For details of each API behavior,
please see the online API document.
Note that this package supports only python version 3.6 and above, due to the async/await and websockets module dependency.
We support python>=3.6. If you want to work with python 3.6, please note that these package dropped support for python <3.7 for the following versions:
pandas >= 1.2.0 numpy >= 1.20.0 scipy >= 1.6.0
The solution - manually install these package before installing alpaca-trade-api. e.g:
pip install pandas==1.1.5 numpy==1.19.4 scipy==1.5.4
Also note that we do not limit the version of the websockets library, but we advice using
Installing using pip
$ pip3 install alpaca-trade-api
To use this package you first need to obtain an API key. Go here to signup
These services are provided by Alpaca:
The free services are limited, please check the docs to see the differences between paid/free services.
Alpaca Environment Variables
The Alpaca SDK will check the environment for a number of variables that can be used rather than hard-coding these into your scripts.
Alternatively you could pass the credentials directly to the SDK instances.
|APCA_API_KEY_ID=<key_id>||Your API Key|
|APCA_API_SECRET_KEY=<secret_key>||Your API Secret Key|
|APCA_API_BASE_URL=url||https://api.alpaca.markets (for live)||Specify the URL for API calls, Default is live, you must specify
https://paper-api.alpaca.markets to switch to paper endpoint!
|APCA_API_DATA_URL=url||https://data.alpaca.markets||Endpoint for data API|
|APCA_RETRY_MAX=3||3||The number of subsequent API calls to retry on timeouts|
|APCA_RETRY_WAIT=3||3||seconds to wait between each retry attempt|
|APCA_RETRY_CODES=429,504||429,504||comma-separated HTTP status code for which retry is attempted|
|DATA_PROXY_WS||When using the alpaca-proxy-agent you need to set this environment variable as described|
Working with Data
You could get one of these historic data types:
First thing to understand is the new data polling mechanism. You could query up to 10000 items, and the API is using a pagination mechanism to provide you with the data.
You now have 2 options:
- Working with data as it is received with a generator. (meaning it's faster but you need to process each item alone)
- Wait for the entire data to be received, and then work with it as a list or dataframe. We provide you with both options to choose from.
option 1: wait for the data
from alpaca_trade_api.rest import REST api = REST() api.get_bars("AAPL", TimeFrame.Hour, "2021-02-08", "2021-02-08", limit=10, adjustment='raw').df open high low close volume timestamp 2021-02-08 09:00:00+00:00 136.6800 137.15 136.450 136.9600 38707 2021-02-08 10:00:00+00:00 136.9600 137.08 136.800 136.8300 22334 2021-02-08 11:00:00+00:00 136.8300 136.97 136.710 136.9100 19546 2021-02-08 12:00:00+00:00 136.9000 136.97 136.050 136.2200 483167 2021-02-08 13:00:00+00:00 136.2200 136.25 136.010 136.1700 307755 2021-02-08 14:00:00+00:00 136.1525 136.35 134.920 135.6900 12159693 2021-02-08 15:00:00+00:00 135.6800 136.68 135.595 136.6100 8076122 2021-02-08 16:00:00+00:00 136.6800 136.91 135.040 136.2350 8484923 2021-02-08 17:00:00+00:00 136.2400 136.54 135.030 135.9745 4610247 2021-02-08 18:00:00+00:00 135.9799 136.30 135.800 135.9330 3375300
option 2: iterate over bars
def process_bar(bar): # process bar print(bar) bar_iter = api.get_bars_iter("AAPL", TimeFrame.Hour, "2021-02-08", "2021-02-08", limit=10, adjustment='raw') for bar in bar_iter: process_bar(bar)
option 1: wait for the data
from alpaca_trade_api.rest import REST api = REST() api.get_quotes("AAPL", "2021-02-08", "2021-02-08", limit=10).df ask_exchange ask_price ask_size bid_exchange bid_price bid_size conditions timestamp 2021-02-08 09:02:07.697204555+00:00 Q 136.80 1 P 136.52 1 [R] 2021-02-08 09:02:07.706401536+00:00 Q 136.80 1 P 136.56 2 [R] 2021-02-08 09:02:07.837365238+00:00 P 136.81 1 P 136.56 2 [R] 2021-02-08 09:02:07.838885705+00:00 Q 136.79 1 P 136.56 2 [R] 2021-02-08 09:02:30.946732544+00:00 P 136.64 1 P 136.50 1 [R] 2021-02-08 09:02:32.558048512+00:00 P 136.64 1 P 136.37 1 [R] 2021-02-08 09:02:32.794415360+00:00 Q 136.69 1 P 136.37 1 [R] 2021-02-08 09:02:32.795173632+00:00 P 136.62 1 P 136.37 1 [R] 2021-02-08 09:02:33.969686528+00:00 Q 136.69 1 P 136.37 1 [R] 2021-02-08 09:02:34.069692672+00:00 P 136.55 1 P 136.37 1 [R]
option 2: iterate over quotes
def process_quote(quote): # process quote print(quote) quote_iter = api.get_quotes_iter("AAPL", "2021-02-08", "2021-02-08", limit=10) for quote in quote_iter: process_quote(quote)
option 1: wait for the data
from alpaca_trade_api.rest import REST api = REST() api.get_trades("AAPL", "2021-02-08", "2021-02-08", limit=10).df exchange price size conditions id tape timestamp 2021-02-08 09:00:11.764828160+00:00 P 136.68 1 [@, T, I] 46 C 2021-02-08 09:00:13.885322240+00:00 P 136.75 35 [@, T, I] 49 C 2021-02-08 09:00:13.885322240+00:00 P 136.75 10 [@, T, I] 48 C 2021-02-08 09:00:13.885322240+00:00 P 136.68 28 [@, T, I] 47 C 2021-02-08 09:00:17.024569856+00:00 P 136.61 16 [@, T, I] 50 C 2021-02-08 09:00:17.810107904+00:00 P 136.66 1 [@, T, I] 51 C 2021-02-08 09:00:19.932405248+00:00 P 136.68 25 [@, T, I] 55 C 2021-02-08 09:00:19.932405248+00:00 P 136.75 18 [@, T, I] 56 C 2021-02-08 09:00:19.932405248+00:00 P 136.68 11 [@, T, I] 54 C 2021-02-08 09:00:19.932405248+00:00 P 136.67 10 [@, T, I] 53 C
option 2: iterate over trades
def process_trade(trade): # process trade print(trade) trades_iter = api.get_trades_iter("AAPL", "2021-02-08", "2021-02-08", limit=10) for trade in trades_iter: process_trade(trade)
Live Stream Data
There are 2 streams available as described here.
The free plan is using the
iex stream, while the paid subscription is using the
You could subscribe to bars, trades or quotes and trade updates as well.
Under the example folder you could find different code samples to achieve different goals. Let's see the basic example
We present a new Streamer class under
alpaca_trade_api.stream for API V2.
async def trade_callback(t): print('trade', t) async def quote_callback(q): print('quote', q) # Initiate Class Instance stream = Stream(<ALPACA_API_KEY>, <ALPACA_SECRET_KEY>, base_url=URL('https://paper-api.alpaca.markets'), data_feed='iex') # <- replace to SIP if you have PRO subscription # subscribing to event stream.subscribe_trades(trade_callback, 'AAPL') stream.subscribe_quotes(quote_callback, 'IBM') stream.run()
Acount & Portfolio Management
The HTTP API document is located at https://docs.alpaca.markets/
API Version now defaults to 'v2', however, if you still have a 'v1' account, you may need to specify api_version='v1' to properly use the API until you migrate.
The Alpaca API requires API key ID and secret key, which you can obtain from the
web console after you sign in. You can pass
secret_key to the initializers of
StreamConn as arguments, or set up environment variables as
REST class is the entry point for the API request. The instance of this
class provides all REST API calls such as account, orders, positions,
Each returned object is wrapped by a subclass of the
Entity class (or a list of it).
This helper class provides property access (the "dot notation") to the
json object, backed by the original object stored in the
It also converts certain types to the appropriate python object.
import alpaca_trade_api as tradeapi api = tradeapi.REST() account = api.get_account() account.status => 'ACTIVE'
Entity class also converts the timestamp string field to a pandas.Timestamp
_raw property returns the original raw primitive data unmarshaled
from the response JSON text.
Please note that the API is throttled, currently 200 requests per minute, per account. If your client exceeds this number, a 429 Too many requests status will be returned and this library will retry according to the retry environment variables as configured.
If the retries are exceeded, or other API error is returned,
alpaca_trade_api.rest.APIError is raised.
You can access the following information through this object.
- the API error code:
- the API error message:
- the original request object:
- the original response objecgt:
- the HTTP status code:
API REST Methods
|Rest Method||End Point||Result|
|list_orders(status=None, limit=None, after=None, until=None, direction=None, nested=None)||
|submit_order(symbol, qty=None, side="buy", type="market", time_in_force="day", limit_price=None, stop_price=None, client_order_id=None, order_class=None, take_profit=None, stop_loss=None, trail_price=None, trail_percent=None, notional=None)||
|get_barset(symbols, timeframe, limit, start=None, end=None, after=None, until=None)||
|get_aggs(symbol, timespan, multiplier, _from, to)||
|get_portfolio_history(date_start=None, date_end=None, period=None, timeframe=None, extended_hours=None)||
||PortfolioHistory entity. PortfolioHistory.df can be used to get the results as a dataframe|
Please see the
examples/ folder for some example scripts that make use of this API
Below is an example of submitting a bracket order.
api.submit_order( symbol='SPY', side='buy', type='market', qty='100', time_in_force='day', order_class='bracket', take_profit=dict( limit_price='305.0', ), stop_loss=dict( stop_price='295.5', limit_price='295.5', ) )
For simple orders with
time_in_force='day', you can pass a fractional amount (
qty) or a
notional amount (but not both). For instace, if the current market price for SPY is $300, the following calls are equivalent:
api.submit_order( symbol='SPY', qty=1.5, # fractional shares side='buy', type='market', time_in_force='day', )
api.submit_order( symbol='SPY', notional=450, # notional value of 1.5 shares of SPY at $300 side='buy', type='market', time_in_force='day', )
get_barset() (Deprecated. use
import pandas as pd NY = 'America/New_York' start=pd.Timestamp('2020-08-01', tz=NY).isoformat() end=pd.Timestamp('2020-08-30', tz=NY).isoformat() print(api.get_barset(['AAPL', 'GOOG'], 'day', start=start, end=end).df) # Minute data example start=pd.Timestamp('2020-08-28 9:30', tz=NY).isoformat() end=pd.Timestamp('2020-08-28 16:00', tz=NY).isoformat() print(api.get_barset(['AAPL', 'GOOG'], 'minute', start=start, end=end).df)
please note that if you are using limit, it is calculated from the end date. and if end date is not specified, "now" is used.
Take that under consideration when using start date with a limit.
Websocket exceptions may occur during execution.
It will usually happen during the
consume() method, which basically is the
exceptions during the consume method may occur due to:
- server disconnections
- error while handling the response data
We handle the first issue by reconnecting the websocket every time there's a disconnection.
The second issue, is usually a user's code issue. To help you find it, we added a flag to the
StreamConn object called
debug. It is set to False by default, but you can turn it on to get a more
verbose logs when this exception happens.
Turn it on like so
You should define a logger in your app in order to make sure you get all the messages from the different components.
It will help you debug, and make sure you don't miss issues when they occur.
The simplest way to define a logger, if you have no experience with the python logger - will be something like this:
import logging logging.basicConfig(format='%(asctime)s %(message)s', level=logging.INFO)
Websocket best practices
Under the examples folder you could find several examples to do the following:
- Different subscriptions(channels) usage with the alpaca streams
- pause / resume connection
- change subscriptions/channels of existing connection
- ws disconnections handler (make sure we reconnect when the internal mechanism fails)
Running Multiple Strategies
There's a way to execute more than one algorithm at once.
The websocket connection is limited to 1 connection per account.
For that exact purpose this was created
The steps to execute this are:
- Run the Alpaca Proxy Agent as described in the project's README
- Define this env variable:
DATA_PROXY_WSto be the address of the proxy agent. (e.g:
- If you are using the Alpaca data stream, make sure you you initiate the StreamConn object with the container's url, like so: data_url='http://127.0.0.1:8765'
- execute your algorithm. it will connect to the servers through the proxy agent allowing you to execute multiple strategies
Raw Data vs Entity Data
By default the data returned from the api or streamed via StreamConn is wrapped with an Entity object for ease of use.
Some users may prefer working with raw python objects (lists, dicts, ...).
You have 2 options to get the raw data:
- Each Entity object as a
_rawproperty that extract the raw data from the object.
- If you only want to work with raw data, and avoid casting to Entity (which may take more time, casting back and forth)
you could pass
Rest()object or the
Support and Contribution
For technical issues particular to this module, please report the issue on this GitHub repository. Any API issues can be reported through Alpaca's customer support.
New features, as well as bug fixes, by sending a pull request is always welcomed.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size alpaca_trade_api-1.2.3-py3-none-any.whl (40.4 kB)||File type Wheel||Python version py3||Upload date||Hashes View|
|Filename, size alpaca-trade-api-1.2.3.tar.gz (32.2 kB)||File type Source||Python version None||Upload date||Hashes View|
Hashes for alpaca_trade_api-1.2.3-py3-none-any.whl