Automate your trading with AI.

6 of the 19 firms we track, including Apex, TradeDay and Alpha Futures, prohibit automation. Topstep allows it through its API Access add-on but requires orders from your own computer, not a VPS or remote server, and its Live Funded accounts can’t use the API.

Every firm’s policy is shown with its sources and the date it was last reviewed. We block deployments a policy prohibits and ask you to acknowledge anything conditional or unclear. Always confirm with your firm.

Today: Paper, TradersPost, PickMyTrade and any custom webhook that accepts JSON, and deploying takes minutes once the account is connected. Through the bridges you can reach Tradovate-based prop accounts. They can also reach TopstepX, but Topstep’s rule that orders come from your own device makes a bridge there an open question, so we ask you to confirm with Topstep first. Bridges don’t report fills back, so those positions are labeled Estimated. The connection test checks the webhook URL’s format only: TradersPost has no dry run, so nothing is sent.

Native Tradovate and ProjectX connections are built from the brokers’ API documentation but not yet tested against live accounts. They’re off unless the server sets ENABLE_NATIVE_BROKERS=1, and Tradovate’s personal API keys reportedly can’t reach prop evaluation accounts.

Yes. Backtests, Charts and the numbers on this page come from 1-minute CME history via Databento: 21,938,051 bars across ES, NQ, YM, RTY, CL, GC, since 2016 (RTY since 2017). Micros trade on the full-size series. A self-installed copy starts with synthetic sample data until you set your own Databento key and run npm run data:databento.

Backtest results are hypothetical: how a strategy’s rules would have traded the past, with commission and slippage, not what they will do. The live chart is TradingView’s. Pass odds are simulated from the backtest’s own trading days, and the deploy monitor shows a paper deployment on the simulated feed; both are labeled that way. A live Databento feed for deployments is built but not yet tested against the live service; paper deployments can run on a simulated feed that replays history, always labeled Simulated feed.

The risk guard checks every entry against your limits (max contracts, daily loss, trailing drawdown, trading window and flatten time) and blocks any entry whose worst-case stop-out would cross one.

If a fill or a fast market trips a limit anyway, the guard flattens the position and locks the deployment until the next session starts at 18:00 ET. Exits are never blocked.

Today the runner runs on the computer where QuantReplay is installed: a separate process that keeps each bot going and checks every order before it leaves. Topstep requires orders from your own computer, so a Topstep deployment must always run there.

No. They are a simulation: your backtest’s trading days, resampled through the evaluation’s rules thousands of times. They show how past results interact with the rules, not what will happen next, and a small sample makes them unreliable.

Yes. A bot is a plain Python file, a qy.Strategy class, plus a small JSON manifest. You can read, edit and copy every line.

There are no user accounts yet. Bots are stored as files in the app’s data folder on the machine where QuantReplay runs, and we don’t publish, sell or trade them. When you chat with Quinn, your messages and the relevant code are sent to Anthropic’s API to generate replies. The privacy policy (draft) has the details.

Paid plans aren’t on sale yet. The refund policy will be published in the terms before they are.