Live trading
Promote a trained model to a capital deployment on your broker. Paper and live share one pipeline; live capital is quality-gated and invite-phased until production ready.
Promote to capital
Choose a trained model and create a capital deployment. Link a broker account; timeframe and universe come from your experiment.
From an experiment overview, select a model (e.g. a specific checkpoint), pick a broker account, and create a deployment. Timeframe and symbols are read from the experiment so training and capital execution stay aligned. Live capital requires paper-before-live and fitness gates—not a separate toy stack.
Agent runs in the cloud
The platform runs your agent on a schedule aligned to your timeframe. It fetches data, runs inference, and places market orders via your broker—no local setup.
Deployments run in the cloud on a schedule (e.g. once per day for daily bars, or at intraday cadence during market hours). The runner fetches the latest candles, runs your model, and places market orders via your broker. You don’t need to host or maintain the execution environment.
One broker for data and execution
Training and capital deployments both use the same broker. Data and execution stay consistent; no fragmented pipelines.
Experiments are configured with a broker account for OHLCV data. The same broker account is used when the agent places orders on a capital deployment. One connection, one source of truth for market data and execution—paper and live differ by credentials and environment, not by product path.