AEGISQUANT
NEW YORK — --:--:--
Auto Dream AegisQuant
01 — RESEARCH TERMINAL

Before you invest, research.

Market data, AI and quant in one place — with full citation of every finding and an explicit level of uncertainty. Forecast, probability, scenario. Never a promise of profit.

02 — SUPERVISED AUTONOMY

Autonomous trading,
PAPER only.

The Hard Risk Kernel decides; the language model never places an order. Without a validated edge, every scan honestly ends in NO_TRADE — and that is a correct result, not a failure.

03 — PRODUCT STATE

20 screens, real data.

See the product Status & gates
What AegisQuant is

One terminal for data, AI and risk

AegisQuant combines live market data, quotes and portfolio monitoring, multi-timeframe interactive charting, a quantitative strategy engine (Strategy Zoo), a 15-agent AI Intelligence system, probabilistic forecasting, risk management and broker connectivity — in one cohesive product for desktop and mobile. The backend runs 24/7 (NATS, PostgreSQL/TimescaleDB, FastAPI), and a separate "Guardian" process watches the market outside trading hours too. Every number carries a source and a freshness marker; no screen pretends to have data it doesn't.

PAPER (simulated) account only Web + installable PWA No App Store / Google Play — beta in preparation Interface: PL · EN · DE · ES · RU
AegisQuant home screen on desktop: portfolio KPIs, equity curve and Auto Trade status
AegisQuant home screen on phone
0
screens on real data
0
AI Intelligence agents
0
Strategy Zoo strategies
the product's language, not a profit promise
Quick tour

11 screens, one click

Switch tabs to see a real screenshot of each key screen — no need to scroll the whole page. A detailed walkthrough of each one is further down.

AegisQuant home screen

Home — 4 portfolio KPIs and a second row (buying power, market, broker, Auto Trade). Equity curve, PAPER positions, latest decisions and news.

Web + PWA

The same app, in your pocket

AegisQuant is a web app installable as a PWA — no App Store, no Google Play. Below are real phone screenshots (an Android wrapper exists, but isn't published to a store yet).

AegisQuant home screen: portfolio KPIs, indices, equity curve, PAPER positions, latest Auto Trade decisions and news
AegisQuant home screen on mobile

Home — everything at a glance

Four headline KPIs (portfolio value, today's result, cash, open risk) and a second row (buying power, market, broker, Auto Trade). Below: indices, the equity curve, PAPER positions, the latest Auto Trade decisions, news and research candidates.

  • Alpaca PAPER account with reconciliation every 30 seconds
  • Auto Trade status and the last scan without switching screens
  • A real screen recording — this is the running app, not a mockup
Research candidates on the Home screen are leads for further analysis — not entry signals.
Chart Lab: 1-minute to 1-week candles, SMA/EMA/RSI/MACD/ATR indicators, market structure (BOS/CHoCH/FVG/order blocks), layers panel
Markets screen: sectors, market breadth, leaders and laggards
Markets
Calendar screen: macro events and earnings
Calendar

Analysis — markets, instruments, Chart Lab, calendar

A market terminal (indices, 1D/5D/1M sectors, S&P 500 breadth, VIX, Treasury yields), an instrument page with a chart, fundamentals and SEC filings, and Chart Lab — a full technical workbench on Alpaca SIP candles with a market-structure heuristic.

  • SMA/EMA/RSI/MACD/ATR/volume indicators, a collapsible 320 px layers panel
  • A macro calendar (Fed/FOMC, BLS, BEA, TreasuryDirect) and earnings
  • News with category, sentiment and importance (rule-based, not an AI model)
Market structure (BOS/CHoCH/FVG) is a heuristic read off the candles — not a trading signal.
AI Intelligence: the Today tab with findings from 15 agents, symbols needing attention and the biggest risks
AI Intelligence on mobile

AI Intelligence — 15 agents, cited research

The Today tab shows in plain language what the AI found: new findings, your symbols with a positive/negative/mixed balance, the biggest risks and live changes. The AI Quality / Audit tab is the full lab view: evaluation metrics, an archive, replay and a hallucination test. Every finding is sourced — no black box.

  • A live scheduler under a daily AI cost budget (a cost cap, not a marketing counter)
  • Aegis AI — a chat with market, quant, forecast and risk tools, in your interface language
  • Fast mode (12–25 s) and full mode (about 60 s, 15 live agents, with visible progress)
AI findings are research, never orders. The language model has no access to the broker.
Forecast: 1-20 session horizons, p10-p90 quantiles, probability of a positive return, BASE/BULL/BEAR scenarios, calibration status

ForecastEngine — a distribution, not a point

The engine aggregates Strategy Zoo signals, AI findings, market regime and historical analogues into a quantile distribution (p10–p90) across 1–20 session horizons: three scenarios (BASE 50%, BULL 25% along the p90 path, BEAR 25% along the p10 path), probability of a positive return and expected volatility. Every number is a real calculation over point-in-time data — never a model's guess.

  • Calibration status is always visible: "uncalibrated" by default, "insufficient data" when the sample is thin
  • A forecast cone drawn directly on the instrument chart
  • Look-ahead protection (point-in-time) — a violation stops the forecast
A forecast is a probability range, not a price prediction. Labeled "UNCALIBRATED" until proven otherwise.
Chart Vision: uploading a chart screenshot, the Image → Symbol → Timeframe → Structure → Quant → AI → Forecast → Risk pipeline
Chart Vision on mobile

Chart Vision — paste a chart screenshot, get context

Upload a chart screenshot — a vision model recognizes the symbol and timeframe and describes in words what it sees, each with an honest confidence value. When symbol confidence clears the threshold, the extraction is reconciled against a real Alpaca quote — never against a fabricated "confirmation".

  • Real image-content validation (PNG/JPEG/WebP, 8 MB cap) — a spoofed file is rejected
  • Manual symbol selection when confidence is low
  • Zero fabricated comparisons: low confidence means no reconciliation, with a clear message
An image is never a price source — only Alpaca SIP is. Pattern and support/resistance detection are planned, not built yet.
Auto Trade: loop status, Off / Scan-only / PAPER-auto modes, the latest decision with its stages
Auto Trade on mobile

Trading — risk, plans, Auto Trade, broker

The Hard Risk Kernel computes position size, stops and 1R–3R targets before any order reaches the broker. AutonomousPaperTrader scans the market every 15 minutes in session, runs through the execution gates and sends a PAPER order only after explicit operator consent — and without a validated Strategy Zoo edge, the result is an honest NO_TRADE.

Risk screen: open risk, daily budget, drawdown, Kill Switch
Risk
Broker screen: PAPER account, positions, orders, fills
Broker
  • Operator limits can only tighten the kernel, never loosen it
  • SHADOW and LIVE are blocked in code — not only in the interface
  • Every decision and every NO_TRADE is logged with a reason
Alerts: price rules, signal sources, alert list

Activity — history and alerts

History is a single timeline of everything that happened: research, NO_TRADE decisions with their reasons, forecasts, AI findings, plans, orders, fills, system events — filterable by category and symbol. The alerts engine combines price, quant, news, SEC, AI, risk, broker and system in one place.

  • Price rules you can add and remove yourself
  • System notifications while the app is in the background
  • Email and push adapters are built; channels are not yet configured
System: health of 12 components, Guardian, scheduler, data sources, gates

System — 24/7, no pretending everything is green

Health of 12 components every 30 seconds, in Trading & risk / Infrastructure / Data & AI groups. Guardian watches the market outside trading hours. Local mode without Docker is shown separately as "local fallback" — honestly, without posing as a full HEALTHY state.

  • A circuit breaker for external data providers
  • AI budget and scheduler jobs visible live
  • Settings: profile, appearance, language, notifications, market data, broker, security
How it works

From data to decision — with a human at the end

Six stages between a market candle and a possible PAPER order. The language model sits at the top of the pipeline — it supplies context and research. It has no write path to settings, trade plans or the broker.

Live market data

Alpaca SIP, Cboe VIX, Treasury yields, SEC EDGAR, the macro calendar. A candle older than 4 days blocks further stages.

Strategy Zoo + AI Intelligence

6 quant strategies tested point-in-time on the S&P 500 (2020–2026) and 15 AI agents producing research findings, never orders.

ForecastEngine

Quant signals, AI findings, market regime and historical analogues aggregate into a probability distribution — quantiles, not a point.

Hard Risk Kernel

Position size, stops, 1R–3R targets, a portfolio heat cap and a Kill Switch — before any order reaches the broker.

PAPER broker

The connector accepts only the Alpaca PAPER environment. LIVE cannot be selected in code — not just disabled in the interface.

AutonomousPaperTrader

Scan → validation → risk → a PAPER order after explicit operator consent. Without a validated edge: an honest NO_TRADE with a stated reason.

Auto Trade

Five modes, one rule: PAPER or nothing

Real money is out of the code's reach — the broker connector accepts only the Alpaca PAPER endpoint, and the LIVE environment cannot be selected. SHADOW (G7) and LIVE (G8) stay blocked until the matching gate is formally passed.

ModeStatusCondition
OFFrunningdefault — no scans, no orders
SCAN_ONLYrunningfull pipeline analysis, never an order
PAPER_AUTOrunningexplicit operator consent to the current settings and risk profile
SHADOWblockedrequires passing G7 (Shadow Trading)
LIVEblockedrequires passing G8 (Limited Live)
NO_TRADE is a fully valid scan outcome. No validated edge, stale data, an active Kill Switch, a stale broker reconciliation, or Scan-only mode — all of these produce a logged NO_TRADE with its reason. The system never forces a position.
Safety

Seven invariants, not seven promises

AegisQuant's safety is not a marketing claim — it is rules written into the code and covered by tests.

Broker: PAPER only

The connector works only with the Alpaca PAPER endpoint. LIVE is impossible in code; SHADOW and LIVE stay blocked in the interface until G7 and G8.

The language model never places an order

An order can only come from a plan approved by the Hard Risk Kernel, submitted exactly once, with a deterministic identifier.

Consent is tied to a settings version

Autonomy requires explicit operator consent to a specific settings and risk-profile version. Any change requires re-confirmation.

Settings can only tighten the kernel

The operator can only narrow the limits (risk per trade ≤ 1%, daily loss ≤ 3%, drawdown ≤ 20%). The AI has no access to them.

Reconciliation and Safe Mode

Any unexplained mismatch with the broker triggers SAFE_MODE. Protection of open positions runs at all times, regardless of Auto Trade's state.

Zero secrets in review material

Every review package is scanned with gitleaks and compared against the real keys/account number. Result: 0 hits, verifiable in a file next to the archive.

FREE and Premium

Start free, unlock the rest later

Exactly what you see in the app itself — no prices, no dates, no guessing. Premium is in preparation.

AegisQuant home screen on the FREE plan — locked menu items with a lock icon

FREE

active now

Exactly the account plan every new user sees today — no card, no commitment.

  • Markets, instruments and basic charts
  • News, calendar and SEC filings
  • A preview of AI Intelligence and Agent Zoo
  • Full forecasts for SPY, QQQ, NVDA and AAPL
Your own PAPER account, portfolio and Auto Trade: in preparation for the FREE plan.
AegisQuant Premium window — the full list of unlocked features

Premium

Premium coming soon

The full AI research terminal: 15 evaluated agents, forecasts across five horizons, Chart Vision, Aegis AI and quant research on real market data.

  • AI Intelligence — findings from 15 agents on market, news, SEC and macro, with sources
  • Aegis AI — ask about the market, stocks, events and strategies, Fast and Full modes
  • Full forecasts — five horizons, p10–p90 distribution, scenarios
  • Chart Vision — a chart screenshot connected to real data, quant, AI and the forecast
  • Advanced Chart Lab and Strategy Zoo — full indicators, SMC, market structure, backtested strategies
Daily AI limit on Premium: 5 quick questions, 1 full research run, 3 Chart Vision analyses. No prices and no payment — decisions live in SUBSCRIPTION_DECISIONS_PENDING.md.
Product status

An honest status, not a marketing launch date

Status as of 2026-09-26. Trading environment: Alpaca PAPER only (simulated account). No part of the product can place an order with real money.

G0 – G5 · Infrastructure, data, Strategy Zoo, AI, Hard Risk Kernel
PASS
G6 · Real PAPER Broker
IN PROGRESS — CI blocked by billing
G7 · Shadow Trading
Plan pending approval, not implemented
G8 · Limited Live
Not started
Web + PWAinstallable, no App Store/Google Play
PL·EN·DE·ES·RUinterface and chat language
30 chaptersuser guide in 5 languages
Betatesters being prepared

This is not investment advice

AegisQuant is a research and decision-support tool — never a financial advisor. The product does not promise profit, does not guarantee an outcome, and does not make personal investment recommendations. The product's language is: forecast, probability, scenario, research, NO_TRADE.

All trading happens exclusively on a PAPER (simulated) Alpaca account — no part of the product has access to real money. Historical results, backtests and probabilistic forecasts do not guarantee future results. Every statistic states its calculation method and calibration status.

The investment decision is made solely by the user, outside this product and at their own responsibility. AegisQuant is a research workbench, not a machine for making money.