Titan Fortress sits between your signal source and your broker. It audits every decision in real time, catches silent regressions before they cost you, and gives you a complete audit trail — without holding your funds or your credentials.
Paper trading enabled by default. No live orders until you explicitly opt in.
Most algo traders have a signal source and a broker. What they don't have is visibility into what happens in between.
You change a parameter — position size, cooldown, a symbol in your allowlist. The strategy keeps running. You don't find out it broke until you're down 3% and staring at a spreadsheet trying to reconstruct what changed.
Your webhook fired. Did the webhook land? Was it a duplicate? Did it fail the daily loss limit or a cooldown window? Without a trace, every unexplained miss is a debugging session that eats into your time.
Markets change. Your signal quality drifts. Gate failure rates creep upward, critical mismatches accumulate, and by the time the pattern is obvious it's already done damage. You need a system that watches for this and tells you before you go looking.
Titan Fortress is a hosted evaluation layer, not a trading bot. It never places orders.
TradingView fires a webhook, your Pine Script alert triggers, or your Python bot calls /v1/pretrade/eval. Titan Fortress authenticates the request using your agent key.
The signal passes through a sequential chain of configurable risk gates — duplicate detection, allowlists, cooldowns, loss limits, market hours, position exposure, volatility filters, and more. Every gate result is recorded.
Titan Fortress responds with PASS or FAIL plus the exact gate that triggered, a reason string, and a link to the full trace. No ambiguity.
If PASS, the lightweight forwarder running on your machine reads the verdict and routes the order to your broker using credentials that never leave your machine. Titan Fortress's server never sees them.
Gates run in order. The first failure stops evaluation and records the exact reason. Nothing is silently swallowed.
The most differentiated feature in Titan Fortress.
Pin a baseline of your recent live signals. Define a candidate config — different position sizing, a tighter loss limit, a new cooldown. Replay the same signals through both configs and get a scored verdict before you touch production.
The confidence score compares outcomes signal-by-signal. It degrades conservatively and hard-caps near zero when critical mismatches appear. A PASS means your candidate config would have made the same decisions on your actual history.
Confidence scoring is evidence-backed, not probabilistic. Each mismatch is a real signal from your history, not a simulation.
Titan Fortress watches rolling windows of your decisions and fires alerts the moment something shifts.
Gate failure rates, critical mismatch counts, integrity score trends across rolling windows. Not just current state — directional change.
Alerts are acknowledged, not dismissed. If the underlying fingerprint changes after acknowledgement, the alert automatically reopens. You can't close out a recurring issue by clicking it away.
The dashboard surfaces a real-time health state — STABLE, WARN, or FAIL — derived from active alerts, crit mismatches, and deploy gate results. Not a vanity metric.
Call /v1/pretrade/eval from any language. Get back an allow/deny decision, a trace link, and a reason. Your bot handles execution.
The dashboard gives you a live picture of system health — not a chart of equity, but a signal-integrity view your algo actually needs.
The Explain This button provides a natural language investigation of your current health state backed by real trace evidence, drift alerts, and deployment diffs — scoped to your account. It does not generate trading signals or advice.
Signal volume, gate health over time, P&L tracking, and CSV export. Every decision Titan Fortress has ever recorded is queryable — not just a rolling window.
Daily and weekly signal counts broken down by symbol, side, and verdict. See gate failure rates at a glance.
Which gates are firing most? Which symbols keep hitting cooldowns? The analytics view surfaces it without you having to build a dashboard.
Full CSV export of your trace history. Bring it into your own tooling, backtesting environment, or spreadsheet without any lock-in.
The forwarder pattern keeps your credentials on your machine. Always.
| Broker | Order types | Notes |
|---|---|---|
| Alpaca | Market, limit, bracket | First-class support. Paper account works out of the box. |
| Robinhood | Market, limit | Setup guide included in Titan Fortress docs. |
| IBKR | Market, limit, bracket | Via TWS or IBKR Gateway on your machine. |
| Schwab / TD Ameritrade | Market, limit | OAuth flow handled by forwarder locally. |
| Tradier | Market, limit, bracket | REST-based forwarder; straightforward setup. |
| Webull | Market, limit | Unofficial API via forwarder. Community-maintained. |
No server to deploy. No Dockerfile. Sign up and you're running.
Create a free account. Paper trading mode is on by default — nothing goes live until you configure it.
Hit the Send Test Signal button on the dashboard. You'll see a complete trace — gate results, verdict, reason — appear in the trace browser immediately. Or paste your webhook URL into a TradingView alert and fire it manually.
Use the webhook URL with TradingView, point your Python bot at /v1/pretrade/eval, or set up the forwarder for live execution. The setup guide walks through each integration in detail.
The architecture reflects the product's values. Here's what that means in practice.
No credit card. Paper trading on by default. First trace in under 60 seconds.
Sign up freeAlready have an account? Sign in