NQ · MNQ · CME Globex tick data

Backtest your futures strategy on every tick, by just describing it to your AI assistant.

StratPilot resolves every order on real CME trade prints from Databento, not on guessed OHLC bar paths. Connect it to Claude or ChatGPT, describe your rules in plain English, and get trades, charts, equity curves and a tick-by-tick replay of every trade.

Backtest: 5-min bars, buy pullbacks to the 20 EMA when price is above VWAP and EMA20 > EMA50. Stop 1.5×ATR, target 2R, 9:00–13:30 CT.
stratpilot · run_backtest → 30 trades · 43% win · PF 1.37 · net +$5,095 (NQ×1, after $4.50 RT and 1-tick slippage).
A 5-minute-bar backtester would have resolved 2 trades differently (+$300 instead). Want the trade charts or a robustness sweep?

Illustrative conversation. All results are hypothetical.

What you get

Tick-accurate fills

Market, stop and limit orders, stops and targets are resolved in ts_event order on every CME trade print (Databento GLBX.MDP3). You know whether the stop or the target was hit first. Nothing is guessed.

OHLC-vs-tick audit

Every run re-resolves the same entries the way a bar-only backtester would and shows which trades, and how many dollars, the bar version gets wrong.

Use it from Claude or ChatGPT

StratPilot is an MCP server. Paste one URL into your assistant, sign in, and it can run backtests, sweeps and reports for you in plain conversation.

Trade charts & images in chat

Your assistant can show the chart of any trade inline (entry, stop, target and exit marked), plus the optimizer heatmap.

Two-pane synced replay

Replay a trade tick by tick, with time bars (1/2/5/15m) on the left and tick bars (1000/500/250/100t) on the right on one market clock. Hours of pre-trade context are included.

Trade spreadsheet

A sortable trade list with entry, exit, MAE/MFE, points and net $ for every trade, and a CSV download for your own analysis.

P&L and equity curves

Equity curve, daily P&L, P&L histogram, win rate, profit factor, expectancy, max drawdown and daily Sharpe.

Robustness, not curve-fitting

Grid sweeps over R:R, stop, entry window and cutoff, with a train/test split, walk-forward folds, overfitting flags and an honest verdict.

Premium: small-model add-ons optional

Plain-English rules to a validated strategy, auto-tuning proposed from training data and scored out-of-sample, and plain-language explanations. Metered and budget-capped. The core engine never uses an LLM.

How it works

  1. Create an accountEmail and password. Free.
  2. Connect your assistantPaste https://mcp.stratpilot.dev/mcp and sign in.
  3. Describe your rulesYour assistant translates them to a strategy and runs it.
  4. Review every tradeCharts in chat, plus the full report in your dashboard.

How to connect in Claude / ChatGPT

You need a StratPilot account. There's nothing to install and no API key to copy: you paste the URL and sign in.

MCP server URLhttps://mcp.stratpilot.dev/mcp

Claude (claude.ai / Desktop)

  1. Open Customize → Connectors (or Settings → Connectors), click + Add and choose Add custom connector.
  2. Name it StratPilot, paste the URL above and click Continue. Claude detects the sign-in settings; keep Sign in now and the default OAuth client option.
  3. A StratPilot window opens: log in (or sign up) and click Allow.
  4. In a chat, enable StratPilot under + → Connectors and ask: "Describe the StratPilot strategy schema, then backtest the lincoln_v1 preset."

On Team/Enterprise plans an owner adds the connector once in Organization settings, then each member clicks Connect and signs in.

ChatGPT

  1. Open Plugins, click + and choose Add custom MCP server. On workspace plans that use developer mode, go to Settings → Apps → Create instead.
  2. Name it StratPilot, paste the URL above and choose OAuth authentication.
  3. Sign in to StratPilot in the window that opens and click Allow.
  4. In a chat, pick StratPilot (type @ or use the tools menu) and describe your strategy.

Availability depends on your ChatGPT plan and workspace settings.

Claude Code / other MCP clients

claude mcp add --transport http stratpilot https://mcp.stratpilot.dev/mcp

Then run /mcp to sign in. Any client that supports MCP OAuth (streamable HTTP) works the same way.

Clients without OAuth

Create a personal API token on your account page and send it as Authorization: Bearer <token>, or append ?key=<token> to the URL. You can revoke a token at any time.

Pricing

Free

$0

  • All core MCP tools: backtests, sweeps, charts, replay
  • Private per-user dashboard and reports
  • NQ front month, 21 sessions (2026-09-18 → 2026-10-16)
  • 20 backtests and 3 optimizations per day, 1 job at a time
Sign up

Premium add-ons

Coming soon

  • Plain-English → strategy JSON
  • LLM auto-tune with out-of-sample scoring
  • Plain-language result explanations
  • Higher limits: 200 backtests and 30 optimizations per day, 2 jobs at a time
Billing isn't live yet.

The fine print

Every result is hypothetical. Fills are simulated on trades-only data, with no order book: slippage is a fixed tick count and limit fills are touch or trade-through, with no queue position. The current dataset is about one month of NQ front-month ticks, so treat any strategy as unproven until it's tested on far more data and forward-tested. Nothing here is investment advice. Market data © CME Group, delivered via Databento, shown to logged-in users only.