Autonomous forex market scanner
A quantitative terminal that also tells you when not to trade.
P-Trades Hub scans XAUUSD, GBPAUD and EURUSD across H4, H1 and M15, grades the ABC retracement structures it finds, turns the survivors into a fully specified trade plan sized to your own risk settings — and then measures, in R, what your decisions actually earned.
- Instruments monitored
- 3
- Timeframes per scan
- H4 · H1 · M15
- Scan cadence
- 15 min
- Grade tiers
- A+ · A · B · C
The loop
Scan · Plan · Measure
- 01
Scan
Three instruments, three timeframes, every 15 minutes. Each ABC retracement structure is scored on four confluence pillars and graded A+, A, B or C. The default answer is No Trade — when a scan cycle qualifies nothing, the terminal says so instead of filling the screen.
- 02
Plan
Each surviving structure becomes a fully specified profile: a limit entry, a maximum acceptable entry so you never chase, an ATR-buffered structural stop, structure-capped targets, R:R, a confidence score, and the lot size your own risk settings allow.
- 03
Measure
Log a setup as taken or skipped, add your real fills, and read expectancy in R against an explicit basis. Small samples remain visible as a record, but row/day gates label them immature rather than treating them as evidence.
How it behaves
- Deterministic rules
- Grading, targets and stops come from fixed, versioned rules — not from a model guessing an outcome. A grade describes structure quality; the confidence score is rule satisfaction, never a win probability.
- Explicit provenance
- Every figure is labelled by where it came from: broker candles, engine calculation, deterministic replay, or your own reporting. Margin is shown as an estimate from the contract specification and your leverage, never as a broker quote.
- No forced trades
- No global daily quota. Each account sets its own daily setup cap — unlimited by default — and a filtered empty view means only that nothing matches that view.
- Nothing invented
- When an input is missing or stale — no equity set, no contract specification, a conversion rate too old to trust — sizing refuses and names the reason instead of guessing.
- Alerts never carry orders
- The notification path — web/Android push and email — has no route to a broker: it can only tell you something qualified. Placing an order is a separate system with its own switches, and no alert can turn it on.
- Execution is opt-in and gated
- You may connect a MetaTrader account and leave it in Observe, arm automatic orders on a demo account, require per-order confirmation on a real account, or allow automatic real orders. Each step is armed deliberately and passes independent global, account, policy and exposure gates every time; the first send of any bridge is a dry run.
- Broker evidence stays separate
- Your own journal, evidence read back from your broker, a controlled benchmark account and research replay are reported as four distinct sources and never merged. Broker figures the account does not report stay marked unavailable — Settings values are never substituted for them.
- Broker telemetry is monitoring
- MetaStats history and Risk Guardian drawdown events are observations timestamped by the broker, shown when the account reports them. They describe what happened; they are not a pre-submit safeguard, and no reported breach does not prove no risk.
- Built for AI assistants
- Fourteen MCP tools let ChatGPT, Claude or Claude Code read your setups, automatic orders and risk holds, size a position and maintain your journal as you — using the same eligibility rules, sizing service and R mathematics as the screen.
Open the terminal
Set your equity, risk per trade and instruments once. From then on the feed shows only what qualified — and an in-app guide explains every figure, including what it does not claim.