A Python library to interact with BitMEX's public API. The library can be used to fetch market data, make trades, or create third-party clients.
Python objects are created dynamically using Bravado to auto-generate
a BitMEX adapter object from BitMEX's Swagger JSON.
For applications requiring higher limits, try the Websocket API. For information on rate limiting, see the BitMEX documentation.
$ pip install bitmex
This is an introduction on how to get started with BitMEX. First, make sure the BitMEX library is installed.
The next thing you need to do is import the library and get an instance of the client:
import bitmex
client = bitmex.bitmex()
Bravado creates a function for each API endpoint exposed by BitMEX.
You can view the list of endpoints here. Endpoints are accessed like
regular class attributes: for example, client.Quote or client.Trade.
Each one has a set of methods that mirror the corresponding REST methods. They can be listed via dir(),
and are called like regular methods. Call result() to make the request and get results. For example:
client.Quote.Quote_get(symbol='XBTUSD').result()
The result() function returns a 2-tuple of the form (body, response). The parameters supported by each method are
listed in the API explorer documentation.
BitMEX has an environment for testing programs and strategies at Testnet. Trading is completely free and is identical to the live market.
The client connects to testnet by default. To connect to the live environment, create a client with the parameter
test=False:
import bitmex
client = bitmex.bitmex(test=False, api_key=<API_KEY>, api_secret=<API_SECRET>)
Get the most recent bid/ask price for an instrument:
>>> import bitmex
>>> client = bitmex.bitmex()
>>> result = client.Quote.Quote_get(symbol="XBTUSD", reverse=True, count=1).result()
>>> result[0][0]['bidPrice']
13489.5
>>> result[0][0]['bidSize']
4950
>>> result[0][0]['askPrice']
13495.5
The API supports fetching full data for one or multiple instruments.
>>> client.Instrument.Instrument_get(filter=json.dumps({'symbol': 'XBTJPY'})).result()
([{'symbol': 'XBTJPY', 'rootSymbol': 'XBJ', 'state': 'Open', 'typ': 'FFWCSX', 'listing': datetime.datetime(2017, 10, 29, 12, 0, ...
Any time-based parameters accept Python datetime objects. All timestamps returned from BitMEX are UTC.
>>> client = bitmex.bitmex()
>>> client.Quote.Quote_get(symbol="XBTUSD", startTime=datetime.datetime(2018, 1, 1)).result()
([{'timestamp': datetime.datetime(2018, 1, 1, 0, 0, 11, 701000, tzinfo=tzutc()), 'symbol': 'XBTUSD', ...
>>> client.Quote.Quote_get(symbol="XBTUSD", endTime=datetime.datetime.utcnow()).result()
([{'timestamp': datetime.datetime(2018, 1, 11, 12, 14, 9, 392000, tzinfo=tzutc()), 'symbol': 'XBTUSD', ...
Private endpoints require authentication. Clients authenticate with an API key. Keys can be generated here.
To get an authenticated client instance:
>>> client = bitmex.bitmex(api_key=<YOUR API KEY>, api_secret=<YOUR API SECRET>)
If you try to access a private endpoint with an unauthenticated client, an HTTPUnauthorized error is raised. Calls
to private endpoints work the same as regular ones:
client.Position.Position_get(filter=json.dumps({'symbol': 'XBTUSD'})).result()
An order can be placed through the Order_new() function. See
the API explorer for required and optional parameters.
client.Order.Order_new(symbol='XBTUSD', orderQty=10, price=12345.0).result()
Instruments can be shorted by specifying a negative order quantity:
client.Order.Order_new(symbol='XBTUSD', orderQty=-10, price=12345.0).result()
Orders can be amended by providing the original order ID. Quantity or price can be amended:
>>> client.Order.Order_new(symbol='XBTUSD', orderQty=10, price=1389.0).result()
({'orderID': '688e6956-95df-4b12-a411-7df0edc82135', ...
>>> client.Order.Order_amend(orderID='688e6956-95df-4b12-a411-7df0edc82135', price=13890.0).result()
An order can be canceled given the order ID:
client.Order.Order_cancel(orderID='').result()
You can cancel all open orders:
client.Order.Order_cancelAll().result()
Results can be filtered by most attributes. Use a regular python dict in the form {'key': 'value'}. The endpoint
requires JSON - use json.dumps to convert your python dict into a properly-formatted string. For example:
client.Instrument.Instrument_get(filter=json.dumps({'rootSymbol': 'XBT'})).result()
The BitMEX API returns a maximum of 500 results per request. The default number of results is 100. Use the parameter
count to increase this:
client.Instrument.Instrument_get(count=250).result()
To get more than 500 results, use the parameter start to specify the first result. Results are zero-indexed:
first_page = client.Instrument.Instrument_get(count=500).result()
second_page = client.Instrument.Instrument_get(count=500, start=500).result()
By default, results are returned oldest first. To get newest rows first, use reverse=True:
newest_instruments = client.Instrument.Instrument_get(reverse=True).result()
You have hit your rate limit. For each request, BitMEX returns these headers:
'X-RateLimit-Limit': '150',
'X-RateLimit-Remaining': '147',
'X-RateLimit-Reset': '1516119556'
To view the headers for a response, use the headers attribute of the response object:
>>> result = client.Instrument.Instrument_get(symbol='XBTUSD').result()
>>> result[1].headers
{'X-RateLimit-Limit': '150', 'Content-Type': 'application/json; charset=utf-8', ...
Use these headers to determine your current limit and remaining requests. At the UNIX timestamp designated by
X-Ratelimit-Reset, you will have enough requests left to retry your current request. If you have not exceeded your
limit, this value is always the current timestamp.
If you are limited, you will receive a 429 response and an additional header, Retry-After, that indicates the number
of seconds you should sleep before retrying.
The trading engine is under heavy load. To help improve responsiveness during high-load periods, the BitMEX trading engine will begin load-shedding when requests reach a critical queue depth. When this happens, you will quickly receive a 503 status code with the message "The system is currently overloaded. Please try again later." The request will not have reached the engine, and you should retry after at least 500 milliseconds.