This tool turns a plain-English trading strategy description into real Python code, runs it against real NSE 1-minute stock data in a sandboxed environment, and shows you the results. No coding needed — just describe your strategy like you would to a person.
Describe entry, exit, position sizing, and which stocks — as much or as little detail as you want. Anything you don't specify gets a sensible default (see below), and the result always tells you exactly what was assumed.
| Setting | Default | How to override |
|---|---|---|
| Capital | ₹1,00,000 | "capital 50000" or "2 lakh" |
| Brokerage/cost | ₹0/trade | "cost 20/trade" or "brokerage 50 per trade" |
| Target (non-daily) | 1% | "target 2%" or "take profit 1.5%" |
| Stop-loss (non-daily) | 0.5% | "stop loss 1%" or "sl 0.5%" |
| Target (daily/swing) | 15% | same as above |
| Stop-loss (daily/swing) | 30% | same as above |
| Leverage | 1x (always) | not user-configurable |
| Dataset window | Full available history | "past 1 year", "last 6 months", "past 90 days" |
| Stock universe | Random 5 stocks | name NIFTY50/100/500 or a ticker |
Every result message ends with an "Assumed:" line stating exactly which defaults were applied — nothing is hidden.
Important: if your message already explains a setting in words (even without the exact keyword), the system tries to follow your written explanation instead of silently overriding it with a default — describe it as precisely as you can for the closest match.
Re-sending the exact same description reuses the exact same generated code every time — fully deterministic, identical results on every run. Saying "create new strategy" or "regenerate" forces a fresh AI-written implementation, which can differ slightly in logic and therefore give different (still individually valid) results. Only do this if you deliberately want to try a different implementation of the same idea.
You're currently getting 5× the standard daily usage budget, while the site is free during testing.
Usage is metered in credits, not a flat per-message count — each symbol actually backtested costs 1 credit (that's the real sandbox/compute cost per stock), and an unusually long or complex strategy description costs a little more (extra Gemini codegen cost). A single-stock test barely touches your budget; a NIFTY500 sweep uses a lot more of it, proportional to what it actually costs to run. Your free-trial budget is 1500 credits/day — once billing starts, the standard plan will be 300 credits/day. Check your current usage any time from the account menu at the bottom of the sidebar.