The AI trading community building in the open

Coming soon — run your own AI-powered fund.

Your money deserves more than a chat window.

Be Your Own Fund Manager.

Reuse your Claude or ChatGPT subscription — no extra AI bill.

Terminal

 building alphafund… in development  AlphaFund is an app where AI agents help you build and run trading strategies.  follow on X for updates

AlphaFund · concept preview in development

One app to trade them all

AlphaFund brings your exchanges, brokers and data providers together, with more added every release.

Autonomy levels in AI trading

LEVEL 0 / 5

Levels measure who makes the judgment calls, not whether orders fire automatically.

Manual Trading

▍

AIYOU

Levels adapted from the SAE J3016 levels of driving automation, the scale used for self-driving cars.

One ecosystem, from data and ML models to AI trading

AlphaFund runs on the PFund ecosystem: free, open-source Python packages you can also use on their own.

strategy.py

import pfund as pf

class MyStrategy(pf.Strategy):
    def trade(self, X):
        fast = X["close"].rolling(10).mean().iloc[-1]
        slow = X["close"].rolling(30).mean().iloc[-1]
        if fast > slow and not self.position:
            return [self.create_order(side=1, quantity=1)]
        return []

engine = pf.BacktestEngine()  # pf.TradeEngine() to go live
strategy = engine.add_strategy(MyStrategy())
strategy.add_data("ibkr", "AAPL_USD_STK", resolutions=["1d"])
engine.run()

strategy.py

import pfund as pf
import polars as pl

class SMACross(pf.Strategy):
    def backtest(self, df: pl.DataFrame) -> pl.DataFrame:
        fast = df["close"].rolling_mean(10)
        slow = df["close"].rolling_mean(30)
        return (
            df.create_signal(buy_condition=fast > slow, sell_condition=fast < slow)
            .open_position(order_quantity=1)
            .close_position(stop_loss=0.05, take_profit=0.1)
            .backtest()
        )

engine = pf.BacktestEngine(mode="vectorized", data_range="1y")
strategy = engine.add_strategy(SMACross())
strategy.add_data("bybit", "BTC_USDT_PERP", resolutions=["1d"])
engine.run()

Website GitHub

Write once, trade anywhere

Backtest and trade live with the same code, on any venue pfund connects to.

Backtest to live, no rewrite

Swap one line to take a backtested strategy live.

Vectorized backtesting in one chain

Define signals, entries and exits on a pandas or Polars dataframe, then backtest the whole history at once. No per-bar loop.

Terminal

$ pip install "pfund[core,torch]"

download.py

import pfeed as pe

bybit = pe.Bybit()
feed = bybit.market_feed

result = feed.download(
    product="BTC_USDT_PERP",  # or BTC_USDT_PERPETUAL in full
    resolution="1minute",  # '1tick'/'1t' or '2second'/'2s' etc.
    start_date="2026-01-01",
    end_date="2026-01-01",
    io=pe.DuckLakeIO(),  # store it in your local data lakehouse
)
df = result.data.collect()

stream.py

import pfeed as pe

bybit = pe.Bybit()
feed = bybit.market_feed

feed.stream(
    product="BTC_USDT_CRYPTO",
    resolution="1tick",  # or just 'tick' for tick data
    # only DuckLake supports storing streamed data for now, set it to None to skip storing
    io=pe.DuckLakeIO(),
    callback=lambda msg: print(msg),
)

Website GitHub

Any provider, your own data lakehouse

Download, stream and store market data from different providers through one API.

Backtest-ready data, from any provider

Every provider's data comes out cleaned, with the same columns, so it's ready for backtesting. Switching provider means changing one line, not rewriting your pipeline.

Live data, stored as it streams

Stream real-time market data through the same interface, handle each message with your own callback, and store everything in your local data lakehouse as it arrives.

Terminal

$ pip install "pfeed[bybit,ray]"

bybit: your data source (a crypto exchange), swap in the one you need
ray: parallel processing (recommended)