Risk-first agent design: what to configure before going live
Drawdown limits, leverage caps, and approval modes that keep automated trading under control.
The most durable agents are built risk-first. Market logic matters, but survivability matters more. On Habi, risk configuration is not an afterthought—it is the gate every trade plan passes through before execution.
Drawdown and exposure limits
Set a maximum drawdown threshold that pauses the agent when equity deteriorates beyond your tolerance. Pair it with exposure limits so the agent cannot concentrate too heavily in a single market, asset, or correlated basket.
These limits should reflect how you trade manually. If you never run more than 3x leverage on BTC perps, encode that explicitly rather than relying on the agent to infer conservative defaults.
Approval modes that match your style
Paper mode is for discovery. Approval mode is for supervised automation. Live mode is for trusted playbooks that have already survived real market conditions.
Most traders live in approval mode for weeks: the agent does the monitoring and planning, you keep final say on entries and exits. That balance preserves speed without giving up control.
Receipts for every decision
Every risk check, plan step, and execution attempt leaves a receipt in Habi. When something does not fill or gets rejected, you can trace exactly which rule fired and why.
Treat those receipts as feedback. Tighten rules that are too loose, loosen rules that block good trades too often, and iterate until the agent's behavior feels like an extension of your own process.