Automation that never sleeps, second-guesses, or panics.
Strategies travel from research to your NinjaTrader 8 instance through an encrypted ZeroMQ bridge — a low-latency local order path, full event-sourced audit trail, and risk gating that fires before the order ever leaves your machine.
- localruns in your NT8
- 6 gatespre-trade risk checks
- auditedevent-sourced log
Sage Trading System v3.2
vol budget · daily loss
encrypted bridge
bracket order placed
Three reasons most retail automation fails.
If your bot has lost you money in a way you couldn't reconstruct, you've hit one of these. QuantFlow closes them as defaults.
- 01the latency problem
Cloud round-trips lose edge.
Most "automation" tools route signal → cloud → broker → exchange. By the time the order reaches the matching engine, the setup has moved. QuantFlow runs the strategy locally, gates risk locally, and dispatches via ZeroMQ — a low-latency local path from signal to fill.
localsignal → fill, on your box - 02the blowup problem
Strategies that don't gate risk.
Most algos go straight from signal to order. One regime flip, one news shock, one bad assumption, and your account is gone overnight. QuantFlow checks vol budget, daily loss, and regime fit before every order — six gates in front of every fill.
6pre-trade risk gates - 03the audit problem
You can't replay what wasn't logged.
When an algo loses money you need to know exactly why — the signal, the gate decisions, the broker response. QuantFlow is event-sourced: every signal, gate, fill, and adjustment is journaled. Replay any trade exactly as it happened.
100%events journaled
Five stages. Low-latency local execution.
Every order travels the same path. The latency budget is published — if any hop drifts, the platform alerts you before edge degrades.
- step 00strategy
Signal generation
Strategy code runs locally on your machine — Python, NinjaScript, or our DSL. Signals are produced with full access to the indicator stack and the regime model.
budget≤ 8ms - step 01risk gate
Pre-trade gating
Six gates fire in parallel: vol budget, daily loss, position cap, regime fit, exposure correlation, and drawdown stop. Any rejection halts the order before it leaves.
budget≤ 6ms - step 02zeromq bridge
Encrypted dispatch
Authenticated ZeroMQ socket between the strategy host and your NT8 instance. AES-256, mTLS, signed payloads. No cloud hop. Your strategies stay yours.
budget≤ 14ms - step 03ninjatrader 8
Bracket order placed
NT8 receives the directive, places the bracket (entry + stop + target), and acknowledges. We never bypass NT8's order layer — this is automation that respects your existing infrastructure.
budget≤ 38ms - step 04fill + journal
Fill back to ledger
Fill events stream back through the same socket. The trade card lands in the journal, the strategy P&L updates in TradingOps, and the audit log gains one more event.
budget≤ 10ms
49 strategies. 10 families. One pipeline.
Every strategy ships with its gauntlet result (DSR · PBO · WFE · SPA), its regime-fit profile, and its source. None of them are signals — they are full strategies you can run, audit, fork, or turn off. Metrics shown are historical gauntlet results, not a forward guarantee.
- DOM7
Order-book imbalance, sweep detection, iceberg pings.
- Orderflow6
Footprint reversals, absorption, delta divergence.
- Momentum5
Breakout continuation, gap-and-go, range expansion.
- Mean reversion5
VAL/VAH fades, range fades, regime-aware fades.
- Volatility5
Vol-rising chases, vol-crush premium structures, IV/RV gaps.
- Session4
London open, RTH open, ETH overnight, close auctions.
- Macro4
FOMC, NFP, CPI, options-expiry-aware playbooks.
- Auction4
Composite POC, naked POC, single-print, value migration.
- Swing5
Daily/weekly setups, multi-day holds, trend-fit gating.
- Regime4
Regime-switching meta-strategies that route between the above.
- gauntlet ✓regime · meta-strategy
Sage Trading System v3.2
The flagship. Routes between three regime-fit children based on the live HMM state. Composite-z-aware sizing. Historical DSR / PBO / MAR.
- DSR
- 0.71
- PBO
- 0.18
- MAR
- 1.94
- gauntlet ✓DOM · iceberg-aware
QuantFlow OB-Sweep
Detects sweep + refresh order-book footprints. Latency-budgeted, regime-gated, and trades ES / NQ exclusively because that is where the edge actually lives.
- DSR
- 0.64
- PBO
- 0.22
- MAR
- 1.51
- gauntlet ✓auction · mean reversion
VAL Fade · Rotation
Fades VAL touches when the regime classifier reads rotation. Refuses to trade in trend·long or regime·shift. Tight stops, modest targets.
- DSR
- 0.58
- PBO
- 0.26
- MAR
- 1.34
Six checks in front of every order.
Each gate runs in parallel. Any single rejection halts the order before it leaves the host. Logged, replayable, configurable per strategy.
- gate 01
Vol budget
checks · Realized volatility against your daily vol cap.
halts · Above 1.4× rolling 20d realized.
vol_gate · σ=1.18× · pass - gate 02
Daily loss limit
checks · Realized + open P&L versus your account-level daily stop.
halts · Within 90% of the configured floor.
loss_gate · −1.7R / −2.0R · pass - gate 03
Position cap
checks · Per-symbol contract count and aggregate margin used.
halts · Above your per-strategy contract ceiling.
pos_gate · 3 / 5 ES · pass - gate 04
Regime fit
checks · Live HMM regime versus the strategy's regime-fit profile.
halts · Strategy disabled in the active regime.
regime_gate · trend·long ✓ pass - gate 05
Exposure correlation
checks · Correlation of new order to existing open exposure.
halts · ρ > 0.85 against an open trade in the same direction.
corr_gate · ρ=0.41 · pass - gate 06
Drawdown stop
checks · Account drawdown from rolling 30d high-water mark.
halts · DD breaches the -8% configured ceiling.
dd_gate · -3.2% / -8% · pass
Replay any trade. Exactly as it happened.
QuantFlow is event-sourced. Every signal, gate decision, dispatch, fill, and journal write is journaled with millisecond timestamps. Below is the actual event sequence for trade 4287 — entry, fill, ledger write.
- 14:32:17.842signalstrategy=sage_v3.2 · setup=POC bounce · regime=trend·long · score=92
- 14:32:17.851vol_gateσ=1.18× · cap=1.40× · pass
- 14:32:17.852loss_gaterealized=−1.7R · cap=−2.0R · pass
- 14:32:17.853pos_gateopen=3 · cap=5 ES · pass
- 14:32:17.854regime_gatelive=trend·long · profile=ok · pass
- 14:32:17.855corr_gateρ=0.41 · cap=0.85 · pass
- 14:32:17.856dd_gatedd=−3.2% · cap=−8% · pass
- 14:32:17.860dispatchbracket entry=5,812.25 · stop=5,805.50 · target=5,830.00 · qty=3
- 14:32:17.874zeromqencrypted · seq=4287 · ack in 14ms
- 14:32:17.918nt8.placedorder_id=8FF3-A2 · status=working
- 14:32:18.062nt8.filledfill=5,812.25 · slip=0.00 · qty=3
- 14:32:18.071journal.writecard_id=trade_4287 · tags=5 · sage annotation queued
Questions people actually ask.
If yours isn't here, email sage@nexural.io and we'll add it.
- Because most retail futures traders already run it, and the platform’s order layer is robust, well-documented, and widely supported by prop firms. We built QuantFlow to integrate with infrastructure you already trust, not to replace it.
- No. The strategy runs on your host. The ZeroMQ bridge transports order directives — not source code, not strategy state, not positions. We do not have the keys to your strategies.
- Median around 50ms host-to-fill on a healthy network with NT8 on the same machine. p99 budget is 200ms. Cross-host (different machine) typically adds ~5-12ms depending on LAN. We publish live quartiles in TradingOps.
- Yes, three ways: native NinjaScript, our Python SDK, or our DSL for declarative strategies. All three plug into the same risk gate and audit pipeline. The 49 production strategies are published as fork-able templates.
- QuantFlow detects the lost socket within one heartbeat (250ms default). It does not retry the order to a degraded NT8 — that creates double-fill risk. Instead it halts the strategy, alerts TradingOps, and waits for human ack. Open positions are managed by NT8’s native bracket logic until the link recovers.
- Yes — and several of our members run QuantFlow on funded accounts. The risk gates are configurable per-strategy, so you can mirror your prop firm’s daily loss and trailing drawdown rules. Confirm with your firm before deploying; we cannot speak for their compliance policies.
- Yes. QuantFlow is a separate runtime; the Swing Desk feeds it candidates if both are subscribed but is not required. You can run QuantFlow with your own signal source.
Where this fits in the system.
Every surface compounds the others. The Swing Desk feeds the journal. The journal grounds Nexural AI. Nexural AI cites the indicators. The indicators feed QuantFlow.
- indicators
Institutional-grade indicators
The signal layer feeding QuantFlow. Volume profile, QPulse, regime classifiers — the same indicators the strategies key off of.
→ /indicators - journal
Event-sourced journal
Every signal, gate decision, fill, and adjustment is journaled here. Replay any QuantFlow trade exactly as it happened.
→ /journal - nexural-ai
Nexural AI copilot
Ask why a strategy fired, what gate blocked an order, or how a regime shift changed the playbook. Grounded in your live tape.
→ /nexural-ai
Quiet automation. Routed under your account, on your machine.
QuantFlow runs in your NinjaTrader 8 instance and writes every decision to an audit log you own. No mystery code. No third-party execution. If the gates don’t do what we say they do, full refund inside the 7-day money-back window.