Configure once. Let multi-agent engines select, qualify, and execute — while risk caps, kill switches, and an independent watchdog keep capital under control.
Configure Alpaca and AI agents once per tenant. Each user connects Webull or Robinhood, starts their own worker, and gets an isolated book — with selective autonomy, multi-agent consensus, and a decision journal that explains every trade or refusal.
Shared Alpaca feeds and OpenAI / Claude keys for everyone in the tenant — not a global singleton. Other tenants keep their own integrations.
Each user stores Webull or Robinhood credentials, chooses paper / shadow / live, and starts or stops their own trading worker — without seeing another user’s book.
A luxury ops dashboard for P&L, symbols, orders, Settings, setup wizard, and one-click pause / panic / flatten — designed for calm intervention.
Read the board top to bottom: intelligence scores the idea, hard gates can refuse it, then you and an independent Watchdog stay in the loop. Click any phase to explore.
Operator map
Scores & proposes — never bypasses risk
Any refuse blocks ENTRY · EXIT stays open
Waiting on your rules
Always able to intervene
Separate process · always on
Phase 1 · Your rules
Budgets, max loss, and paper/day mode come first. The system cannot invent risk beyond what you allowed.
Strategy theater
Signals in. One intent out.
Normalizing signals…
Only evidence-backed ideas continue to risk — never a raw signal storm.
Signals are normalized, conflicts resolved, and only evidence-backed intents reach the broker — after qualification and risk.
Trend follow, mean reversion, breakout, ORB, VWAP reversion, momentum pullback, MACD, SuperTrend, confluence, and RVOL breakout packs.
Directional single-leg, vertical spreads, and protective / hedge overlays with explicit buy/sell permission gates.
Harmonics, structure, candles, and anti-chop filters produce soft or hard blocks before capital is committed.
Risk gate
Entries earn permission.
Pause can block new entries — flatten and exit ladders stay privileged.
Daily loss limits, per-symbol caps, stale-data breakers, and kill switches block unsafe entries. Exits and flatten ladders stay privileged so recovery is never locked behind a pause.
Watchdog
Separate process. Always awake.
If heartbeats stall or broker probes fail, Watchdog opens an incident, alerts operators, cancels risky entries, and can escalate a controlled flatten using broker truth — even when the main app is down.
Explore how intelligence, risk gates, and an independent Watchdog work together — paper-first, auditable, and supervised when it matters.
