Vibe code
your
trading
strategies.
Describe your idea. The agent turns it into research, code, visualization, and tests across granular market data.
Get $10 in free monthly agent credits, no credit card required.
I'll check what the archive holds for ES first, then write the strategy.
June and the first half of July are complete in the archive. I'll fit on June and keep July 1 to 17 untouched as the held-out stretch.
from atlantive import BUY, NEW_YORK, SELL, Bars, RollingMinMax, Session, Strategy # A quarter of a point, in the engine's prices (nanos). TICK = 250_000_000
The surface is a broad plateau rather than a spike, so I keep 24 and 48 ticks and run them once over the held-out days.
Fitted on June and run once on July 1 to 17, both on ES one-second bars at $2.10 a contract. The sweep is a plateau rather than a spike, so 24/48 was not picked off a peak.
Thirteen held-out sessions is a short test. Run it on the next month before trusting the size of the edge.
Held-out run, comparison
How Atlantive works
Describe the idea in plain English
Fills that never look ahead
Charts drawn from recorded data
A strategy is a short Python file
Prop firm rules, before real money
Narrow mornings lose. Under 6 points of opening range, version 1 took 4 trades in June, won 1 and lost $604. Version 2 skips those mornings and changes nothing else.
On June, in sample, version 2 has a Sharpe of 4.10 against 2.94, on 18 trades against 22.
The order book is not open to runs yet, and this is a demonstration of it. At 13:29:04.318, 66 seconds after price first reached it. Resting from the entry, it joins the queue behind 64 contracts in 23 orders. The first visit traded 11 of them, the second 26, and 14 more were cancelled ahead of it. The third reached it. A fill on touch would have sold at 13:27:58, with contracts that belonged to the orders in front.
- Ahead of the orderAhead
- 0
- Traded at the levelTraded
- 50
- Cancelled aheadCancelled
- 14
July 15 opened in a 10.50 point range and broke down at 10:10. The short sold 7259.75 and covered at 7247.50 at 14:34, +$608.30 after costs. The surface is the June sweep: a plateau from 24 to 40 ticks of stop, which is why 24 and 48 were kept rather than the single best cell.
One parameter and three lines: the width of the morning's range, and a return before any entry when it is under 6 points. The stop, the target and the exit at 16:00 are as they were. The Pine Script carries the same parameters, cash and fees, and fills at the next bar's open, as TradingView does.
//@version=6
// OpeningRangeBreakout, exported from Atlantive.
// Chart simulation: decisions use completed bars; market orders fill at the next bar open.
// Native fills, fees, data gaps and session boundaries can differ. Re-test on the target symbol.
// Use standard time-based candles and the strategy's timeframe, not synthetic chart prices.
// Pine v6 trims old trade records at its order limit; the strategy keeps calculating.
// Margin is explicitly disabled to retain the native signal simulation's sizing.
strategy("OpeningRangeBreakout", overlay=true, initial_capital=100000.0, commission_type=strategy.commission.cash_per_contract, commission_value=2.1, default_qty_type=strategy.fixed, default_qty_value=1, calc_on_every_tick=false, process_orders_on_close=false, slippage=0, margin_long=0, margin_short=0)
Skip a morning whose opening range is under 6 points: a narrow range breaks out on noise.
8 unchanged lines99class Params:1010stop_ticks: int = 241111target_ticks: int = 4812+min_range_ticks: int = 241213clip: int = 113141415class State:25 unchanged lines4041if close >= self.stop or close <= self.target or minute >= 955:4142orders.ioc(side=BUY, qty=self.p.clip)4243elif 585 <= minute < 955 and self.clock.prev_day != self.traded_day:44+# A narrow morning breaks out on noise: leave it alone.45+width = self.opening.max() - self.opening.min()46+if width < self.p.min_range_ticks * tick:47+return4348# The first five-minute close outside the range, once a day.4449if close > self.opening.max():4550orders.ioc(side=BUY, qty=self.p.clip)8 unchanged lines
| Market | Parameters | Return | Buy & hold | Sharpe | Max drawdown | Trades |
|---|---|---|---|---|---|---|
| esu6-glbx-ohlcv-1s (2026-06-01 to 2026-07-17) | Defaults | +4.26% | -5.76% | 4.06 | -1.38% | 28 |
On its own trades, June 1 to July 17, the account passes on day 32, never nearer than $874.90 to the trailing floor. Across 10,000 resampled orderings 87% pass, and 99% of the failures are the max drawdown. June is where the rule was found, so read July as the test.
- This run
- Passed
- Day 32, Fri 2026-07-03
- Pass rate
- 87.1%
- of 10,000 attempts
- EV per attempt
- +2,536.49
- p5 +2,460.07 to p95 +2,600.22
- Days to pass
- 32
- median, p95 62
Profit +3,178.50 against the 3,000.00 target after 20 trading days, the best day 24% of it.
Show the day on the run's priceThis run: its own 20 of 28 trades in the order they traded, from 2026-06-02, on trading days that turn at 17:00 Chicago time. Watched at trade closes only: the run recorded no hourly extremes of its equity, so a loss inside a trade that recovered by its close is not seen. The pass rate is over 10,000 attempts resampled from them in blocks of 5 trades.
Rules
- Profit target+3,000.00+3,178.50 ReachedAt the end of day 32, Fri 2026-07-03 Passed by 87.1% of attempts, median day 32
- Daily loss limit1,000.00-479.20 HeldWorst day, day 9, Wed 2026-06-10 Ends 0.6% of failed attempts, median day 20
- Max drawdown2,000.001,125.10 Heldtrailing, end of day Closest to the floor on day 21, 874.90 above it Ends 99.4% of failed attempts, median day 25
- Minimum trading days2 days20 days Met
- Consistency50%24% Withinthe best day's share of the profit Best day +758.30, day 11
Account, day by day
Attempts
What ends the attempts that fail
- Max drawdown99.4%median day 25
- Daily loss limit0.6%median day 20
Agent
Describe the idea in plain English. The agent writes the strategy, runs the backtest, reads the result and comes back with a sharper version. It has a real Python interpreter for the research behind it, and every step is logged in a session you can audit.
Backtesting
A run reads the exchange's own one-second bars, decides only on bars that have closed and fills on the next second's prices: no look ahead and no invented prices. The panel shows the order book, where your order waits in the recorded queue until the tape reaches it. That is not open to runs yet.
Charts
Candles with your trades and your strategy's own lines printed on them, equity and drawdown, all drawn from recorded data that never redraws itself. When the view you need does not exist, describe it and the agent builds it, up to a full 3D surface of Sharpe across every stop and every target.
Strategies
A strategy is a short Python file you can open, edit, diff and version, and every run keeps the version it measured. The same file exports to Pine Script when you want it on TradingView, with the same parameters, cash and fees.
Validation
Before a strategy touches money it faces the rules prop firms apply to the traders they fund: daily loss cap, max drawdown, profit target. Its own trades are played through them day by day, then resampled ten thousand times to see how often it would pass. A sweep keeps the plateau that survives everywhere, never the peak that got lucky once.
Describe the idea in plain English. The agent writes the strategy, runs the backtest, reads the result and comes back with a sharper version. It has a real Python interpreter for the research behind it, and every step is logged in a session you can audit.
A run reads the exchange's own one-second bars, decides only on bars that have closed and fills on the next second's prices: no look ahead and no invented prices. The panel shows the order book, where your order waits in the recorded queue until the tape reaches it. That is not open to runs yet.
Candles with your trades and your strategy's own lines printed on them, equity and drawdown, all drawn from recorded data that never redraws itself. When the view you need does not exist, describe it and the agent builds it, up to a full 3D surface of Sharpe across every stop and every target.
A strategy is a short Python file you can open, edit, diff and version, and every run keeps the version it measured. The same file exports to Pine Script when you want it on TradingView, with the same parameters, cash and fees.
Before a strategy touches money it faces the rules prop firms apply to the traders they fund: daily loss cap, max drawdown, profit target. Its own trades are played through them day by day, then resampled ten thousand times to see how often it would pass. A sweep keeps the plateau that survives everywhere, never the peak that got lucky once.
Agent
Describe the idea in plain English. The agent writes the strategy, runs the backtest, reads the result and comes back with a sharper version. It has a real Python interpreter for the research behind it, and every step is logged in a session you can audit.
Backtesting
A run reads the exchange's own one-second bars, decides only on bars that have closed and fills on the next second's prices: no look ahead and no invented prices. The panel shows the order book, where your order waits in the recorded queue until the tape reaches it. That is not open to runs yet.
Charts
Candles with your trades and your strategy's own lines printed on them, equity and drawdown, all drawn from recorded data that never redraws itself. When the view you need does not exist, describe it and the agent builds it, up to a full 3D surface of Sharpe across every stop and every target.
Strategies
A strategy is a short Python file you can open, edit, diff and version, and every run keeps the version it measured. The same file exports to Pine Script when you want it on TradingView, with the same parameters, cash and fees.
Validation
Before a strategy touches money it faces the rules prop firms apply to the traders they fund: daily loss cap, max drawdown, profit target. Its own trades are played through them day by day, then resampled ten thousand times to see how often it would pass. A sweep keeps the plateau that survives everywhere, never the peak that got lucky once.
Write it in Python.
Export it as Pine Script.
You write 38 lines of Python. TradingView gets the 114 lines of Pine Script they compile to, the same every time.
A closed vocabulary
Five components and three order verbs, in plain Python.
Checked before it runs
A mistake comes back with its line, and nothing starts.
Params become inputs
Tune them on TradingView, starting from your strategy's values.
No AI in the export
A compiler writes the Pine, the same way every time.
Years of futures data, built in.

Sixteen years of CME Globex, already in the archive. 3,799,825,966 one-second bars since June 2010. Nothing to buy, connect, or import.

And every order in the book. A month of the full CME order book, rebuilt event by event, and not open to runs yet.
One of them, as the exchange stamped it
- Trade
- ESU6, a buyer takes 2 at 7,482.50
- Against
- an order of 11, placed 99.9 ms before
- Recorded
- 234 µs after the exchange stamped it
Questions traders ask first.
What is Atlantive?
Atlantive is research software for futures traders. You describe a trading idea in plain English, and an AI agent turns it into a Python strategy, backtests it on historical futures data and reports what it measured: the trades, the equity curve, the drawdown, and whether the result holds up. Anyone can sign up for free.
Do I need to know how to code?
No. You describe the idea in your own words and the Atlantive agent writes the code. What it writes is still an ordinary, short Python file that you can open, read, edit and version, so nothing about how the strategy trades is hidden from you.
Which markets and data does Atlantive cover?
Atlantive covers futures on CME Globex, the electronic market of CME Group: equity index, interest rate, currency, energy, metal, agricultural and crypto futures, such as ES, NQ, ZN, 6E, CL, GC and BTC. Strategies run on price bars from one second to one month, with history back to 2010. Stocks, ETFs and spot currencies are not covered.
How does Atlantive keep a backtest honest?
In Atlantive an order is sent at a bar's close and filled on the next bar's prices, so no decision sees data it could not have had, and commission is charged per contract on every side and reported with the result. The agent then raises the costs, checks the result on data it was not tuned on, and sweeps the parameters to keep a plateau that holds across neighbouring settings rather than a peak that got lucky once.
Can I use an Atlantive strategy on TradingView?
Yes. A strategy written in Atlantive compiles to Pine Script, TradingView's language, with one click. The compiler translates the file directly rather than asking an AI to rewrite it, and it refuses, by name, anything a chart cannot reproduce instead of approximating it.
Can Atlantive test a strategy against prop firm rules?
Yes. Atlantive simulates a finished backtest against a prop firm's challenge rules, or against rules you set yourself. It resamples the run's own trades into thousands of attempts, judges each one day by day, and reports how likely the strategy is to pass, fail or run out of time, and which rule stops it first. Before you buy an account, compare futures prop firms on what each costs to reach funded.
Does Atlantive trade for me or give investment advice?
No. Atlantive is research software: it never places orders or connects to a brokerage account, and nothing it produces is investment advice or a recommendation. A backtest measures how rules would have behaved on past prices, it is not a forecast, and trading futures carries a risk of loss. The Risk disclosure says more.
Who owns the strategies I build?
You do. Your strategies, code, files, backtests and conversations remain yours, and Atlantive does not sell them or use them to train models. The Privacy policy names the providers that process them to produce the agent's answers.
How much does Atlantive cost?
Atlantive is free during the beta, and new accounts receive $10 in monthly credits for the agent. Anyone can sign up to create a workspace, or sign in to an existing account. Credits cannot be bought, and no payment details are asked for.