NQ BACKTEST API

Your AI can backtest futures now.

Describe a strategy in plain language. Your AI composes it, this API runs it against 15+ years of NQ 1-minute history in under a second, and every run lands in your private vault. Works in Claude, Cursor, and any MCP-capable client — no code, no data wrangling, no Python environment.

you > Backtest a 30-minute opening range breakout on NQ since 2022 — long only, entries before noon, 60 tick stop, 120 tick target, one trade a day.
AI > Done — 580 trades over 1,217 sessions. Net +$7,180, profit factor 1.06, win rate 35.7%, max drawdown −$12,932. The edge is paper-thin after costs: nearly all the profit is 2023 plus early 2024. Want me to try an 80-tick stop, or add shorts?

That answer is the product: honest numbers, costs and slippage included, explained. A backtest that tells you a strategy loses money is the tool working.

Install: paste one block

Claude Desktop → claude_desktop_config.json (Cursor and others take the same server entry):


  

  
Then just talk to it: “what's my best strategy by Sharpe this month?”,
  “re-run my ORB with a 40-tick stop and compare”.


  
What your AI gets

  

    
A real grammar, not a menu
Indicators, session stats, opening
      ranges, time filters, crosses and logic compose freely into strategies nobody pre-programmed.
      The free /spec endpoint teaches the grammar; validation errors teach
      the fixes.

    
Fills you can defend
Signal at bar close, entry next open,
      slippage against you, stop checked before target inside a bar, gaps fill at the open.
      Deterministic: same spec, same answer, stamped with the data version.

    
The vault
Every run auto-saved: spec, metrics, equity curve.
      List, compare, re-run with one change, tag the keepers. Identical re-runs are free.

    
15+ years, validated
Every session since June 2010 checked
      against the CME calendar — gaps, halts and holiday sessions accounted for and documented.
      Raw bars never leave the server.

  


  
Pricing

  

    
Free
$0

      
5 backtests a month. Enough to find out your idea doesn't survive costs.

      Get a free key

    
Basic
$15/mo

      
200 credits a month, full vault, full history.

      

    
Pro
$30/mo

      
1,000 credits, 1-minute-bar backtests, exports. For system builders.

      

  

  
A backtest costs 1 credit (2 on 1-minute bars). Re-running an identical spec is free.
  Cancel any time from the payment receipt; your key drops back to the free tier.


  
Start free, right now

  
No card. Your key appears here once - save it, we only keep a hash.

  

    

      
      
    

    

  


  
For developers and agents

  
Underneath it's a plain HTTPS API — no AI required. The grammar and execution rules are
  public and machine-readable:

  

    
    

  


  
Questions traders ask

  
Is this NQ only? Today, yes — NQ (E-mini Nasdaq-100), continuous
  back-adjusted, intraday. ES, gold and crude are next on the same engine.

  
Which AI does it work with? Anything speaking MCP: Claude, Cursor, Windsurf,
  Cline and friends — or call the REST API directly from Python.

  
Do you give me the price data? No. You get computed statistics — metrics,
  a downsampled equity curve, monthly tables. That keeps exchange licensing clean and the price low.

  
Are the results real? They're deterministic, cost-inclusive, and conservative
  (stop assumed before target when a bar spans both). The methodology is published at
  /spec so your AI — or you — can audit every rule.