Skip to main content

AI Strategy Builder

The Strategy Builder turns a plain-English description of a trading idea into working Python strategy code — validated, backtested against real historical data, and ready to activate on a paper or live account. No programming required, though you can read and edit every line it writes.

The Strategy Builder: ARIA chat, generated code, backtest configuration

Open it from the sidebar: Strategy Builder (available on every paid plan).

How it works​

Describe the idea → AI writes Python → validated in the engine's sandbox → backtest → refine → activate
  1. Describe — tell ARIA what you want in the chat: "Go long on a 5-minute opening-range breakout on MNQ, 20-tick stop, 40-tick target, flat by 4 PM." The more specific you are about entries, exits, and sessions, the better the first draft.
  2. Generate — the AI writes a complete strategy against Atomik's strategy contract: the same BaseStrategy interface the live engine runs. The code appears in the editor with a Valid badge once it passes validation.
  3. Backtest — pick a symbol, timeframe, and date range, and run it against real historical futures data. You get win rate, profit factor, drawdown, equity curve, and the full trade log.
  4. Refine — ask for changes in the chat ("add a volume filter", "move the stop to 30 ticks") and re-test. Asking questions about your strategy is always free.
  5. Activate — when you're happy, activate it on an account. It runs on Atomik's engine against live market data — start on paper.

Everything you build is saved under My Strategies, and every backtest is kept in Backtest History so you can compare runs.

AI credits​

AI work is billed in credits, priced by what each action actually costs:

ActionCredits
Build a new strategy3
Adjust an existing strategy1
Ask a question about your strategy0 — always free
PlanAI credits / monthBacktests / month
Starter7530
Pro300200
Elite600500

Credits reset at the start of each billing month, with a daily ceiling so one busy day can't consume the whole month. Pro and Elite use a premium AI model for longer, more intricate strategies. Your remaining credits are always visible in the Builder's top bar.

Symbols and data​

What you can backtest depends on your plan:

PlanSymbolsMax range per backtest
StarterMNQ, MES, MGC, ES, NQ90 days
Pro+ YM, RTY, CL, GC, SI, ZB1 year
Elite+ YM, RTY, CL, GC, SI, ZB5 years*

*Bounded by stored history — Atomik warehouses roughly two years of data for the major contracts. The Builder shows each symbol's exact stored range under the date pickers.

Timeframes: 1s, 1m, 2m, 3m, 5m, 10m, 15m, 30m, 1h, 2h, 4h, 1d. Backtest the same timeframe your strategy's code declares.

Write your own Python instead​

The Builder is one door; your own code is another. Choose Engine Strategy in the dashboard's Create flow, or use the MCP connector to have Claude write strategies into your account. All three paths produce the same thing: a validated strategy that backtests and activates identically.

Honest expectations​

The Builder makes it fast to test ideas — it does not make ideas good. Most strategy ideas fail honest backtesting; that's the point of testing before risking money. Read How Backtesting Works to understand exactly what the simulator does and doesn't model, and treat a good backtest as permission to try it on paper, not proof it will make money live.

Next steps​