aurastream

One API for crypto, FX, rates, macro, positioning and alternative data, from 47 sources, in one response format.

The official Python client for AuraStream by Unbiased Alpha. Pull Binance candles and funding, ECB and BIS statistics, FRED vintages, CFTC positioning and SEC filings with the same call, and get a clean pandas DataFrame back.

pip install "aurastream[pandas]"

The package installs as aurastream and imports as aurastream_client.

Free plan: 500 calls a month, no card. Get a key →

Quickstart

from aurastream_client import AuraStream

client = AuraStream()   # reads AURASTREAM_API_KEY from the environment

# Bitcoin daily candles from Binance, with no provider account needed
btc = client.get_frame("BTCUSDT", "2025-01-01", "2025-06-30", source="binance")

# Several coins' closes in one wide DataFrame
prices = client.get_panel(["BTCUSDT", "ETHUSDT", "SOLUSDT"], "2024-01-01", "2025-01-01", source="binance")
prices.pct_change().corr()

# The ECB's daily EUR/USD reference rate
eurusd = client.get_frame("ECB,EXR,1.0/D.USD.EUR.SP00.A", "2024-01-01", "2025-01-01", source="sdmx", db="ECB")

Why AuraStream

• One endpoint, one schema. Stop writing a new adapter for every vendor. Every source returns {ticker: {timestamp: {field: value}}}.
• 47 sources. Binance, Bybit, Kraken, Deribit, Hyperliquid, dYdX, FRED, US Treasury, ECB/BIS/OECD (SDMX), World Bank, CFTC, SEC EDGAR, USDA WASDE, GDELT, TSA, and more, plus brokers and vendors you already have an account with (Oanda, Alpaca, Polygon...). Full list.
• Point-in-time where it matters. FRED series come with every revision, and CFTC data is stamped by release date, so your backtests don't peek.
• Bring your own keys. For sources that need an account (Oanda, Alpaca, Polygon, FRED…), your credentials pass through per request and are never stored.

Examples

Every notebook opens in Google Colab and runs on the free plan.

Usage

Raw JSON

data = client.timeseries(["BTCUSDT", "ETHUSDT"], "2025-01-01", "2025-01-31", "1h", source="bybit")

Source-specific parameters

Pass any extra field as a keyword argument, and it goes straight into the request body:

# Binance USD-M perpetual funding rates
funding = client.get_frame("BTCUSDT", "2025-01-01", "2025-06-30", "8h",
                           source="binance", instrument="futures", bars="funding")

# CFTC Traders in Financial Futures report for euro futures
cot = client.get_frame("6E", "2020-01-01", "2025-01-01", "1w", source="cftc", report="tff")

Provider credentials

Pass credentials= explicitly, or set environment variables (the same names aurastream-mcp uses):

cpi = client.get_frame("CPIAUCSL", "2020-01-01", "2025-01-01", "1m",
                       source="fred", credentials={"key": "YOUR_FRED_KEY"})

Large requests

The API stops a request after about 30 seconds. Split long histories into several date ranges, and long ticker lists with batch_size:

coins = ["BTCUSDT", "ETHUSDT", "SOLUSDT", "XRPUSDT", "BNBUSDT", "DOGEUSDT", "ADAUSDT", "LINKUSDT"]
data = client.timeseries(coins, "2025-01-01", "2025-06-30", "1h", source="binance", batch_size=4)

Each batch counts as one call against your allowance.

Errors and retries

The client retries rate limiting (429 Too Many Requests) and server errors with exponential backoff, three times by default. It never retries bad requests, auth errors, an exhausted monthly allowance or a timeout (504: split the request instead):

from aurastream_client import QuotaExceededError, BadRequestError

try:
    client.timeseries("XAU_USD", "2025-01-01", "2025-02-01", source="oanda")
except BadRequestError as e:
    print(e.message)          # e.g. missing credentials
except QuotaExceededError:
    print("Monthly allowance used: upgrade at aurastream.unbiased-alpha.com")

Also available

• MCP server for Claude and ChatGPT: pip install aurastream-mcp (repo)
• OpenBB extension: pip install openbb-aurastream (repo)
• REST docs: aurastream.unbiased-alpha.com/docs

Plans

Compare plans →

License

MIT © Unbiased Alpha GmbH