AI Agents Supervisor

WANT AN AI PORTFOLIO MANAGER WATCHING YOUR FLEET 24/7? 

Run your fleet on the same brokers I do: IC MARKETS, IC TRADING . Unlike the centralized stock market, Forex has no single, unified price feed. Every broker sources liquidity from different providers, creating unique data streams. The Supervisor reads the same feed your EAs trade on, so its risk decisions match your real fill conditions.

ONLY 3 COPIES OUT OF 10 LEFT AT $199!

After that, the price will be raised to $299.

EA will be sold in limited quantities to ensure the rights of all customers who have purchased.

USER GUIDE & SETUP FOR AI AGENT SUPERVISOR

AI Agent Supervisor is not just another Expert Advisor — it is a multi-agent AI supervisor that sits on top of your entire account and manages the EAs you already run. Where every other EA on this market opens trades, AI Agent Supervisor never trades its own chart: it watches, decides, and intervenes. It auto-discovers every EA on the account by magic number, classifies each as one of your own EAs or a third-party EA, and applies a five-agent decision pipeline to manage the whole fleet — pausing, scaling down, locking direction, tightening stops, or flattening positions when conditions demand it. One Supervisor — your whole portfolio — one decision pipeline.

Under the hood, every portfolio decision is debated by five specialised AI agents — Technical Analyst, Bull Analyst, Bear Analyst, Research Manager and Portfolio Manager — each with its own role, prompt, and veto power. The Bull and Bear argue across multiple debate rounds; the Research Manager weighs the evidence; the Supervisor Portfolio Manager has the final word and translates regime, exposure and correlation into a concrete intervention for each target EA. This adversarial design eliminates the single-model blind spot that plagues every "ChatGPT EA" you have seen so far.

Crucially, the Supervisor earns your trust before it acts. It runs in three escalating modes — Shadow (log every decision, touch nothing), Approve (propose actions and wait for your one-click approval), and Live (act automatically). Every decision — the regime read, the agents' reasoning, the target EA and the action taken — is written to a per-instance audit file. You watch the Supervisor reason in Shadow, approve it in Approve mode, and only hand it the keys once it has proven itself.

THREE CORE STRENGTHS

1. A Fleet Supervisor, Not Another Strategy. Most EAs add one more strategy to your account and hope it survives. AI Agent Supervisor manages the EAs you already run. Point it at the whole account and it auto-discovers every magic; or name specific magics. EAs you own are treated as cooperative — the Supervisor steers them gently (pause, resume, reduce lot, lock direction) through a lightweight signal your EA reads via a drop-in include. Third-party EAs are treated as adversarial — the Supervisor enforces broker-side (force-close, block exposure) because they will not take instructions. One layer governs your entire portfolio.

2. Five-Agent Debate Pipeline. A single AI call is a guess. Five agents arguing is a process.

• Technical Analyst — reads regime, volatility and price action across the account's symbols.
• Bull Analyst — argues to keep exposure and let winners run.
• Bear Analyst — argues to cut risk, citing correlation clusters and drawdown.
• Research Manager — arbitrates the debate, scores conviction.
• Portfolio Manager — turns conviction into a concrete per-EA action: pause, scale, lock direction, tighten stop or flatten.

Configurable debate depth lets you tune speed-vs-thoroughness; in deep mode the Bull and Bear iterate multiple rounds until the Research Manager is satisfied the case has been stress-tested.

3. Three-Tier Trust Model + Independent Risk Governor. The Supervisor escalates with your confidence: Shadow → Approve → Live. Wrapped around the AI is a hard risk governor that runs independently of the agents — equity floor, daily and weekly loss caps and high-impact news blackout can pause or flatten the whole fleet even when the AI's conviction says otherwise. The Supervisor can only ever reduce risk — pause, scale down, or close. It never opens a new position or increases exposure.

BUILT-IN INSTITUTIONAL-GRADE RISK MANAGEMENT

The decision pipeline is wrapped in a multi-layer risk governor that runs independently of the AI — even if the agents' conviction is high, hard limits cannot be overridden:

• Daily / weekly loss caps with auto-lockout on breach
• Equity-floor protection vs initial deposit — account-wide kill switch
• High-impact news blackout (NFP, FOMC, CPI …) broadcast fleet-wide
• Per-target actions: pause / resume, lot scale-down, direction lock, tighten-stop, force-close
• Cooperative vs adversarial classification with a per-hour action rate-limit on third-party EAs
• Dual-confirm gate and catastrophic-action guard before any account-wide flatten
• Per-magic file & global-variable isolation — safe to run alongside everything else on the terminal
• Full decision audit trail (one JSONL line per decision) for after-the-fact review

NO MARTINGALE. NO GRID. NO RECOVERY ZONES. NO NEW RISK. The Supervisor never opens or averages a position. It only pauses, scales down, locks direction, or closes — risk is reduced, never added.

SUPPORTED AI MODELS

Plug in any of the major API providers via a single API-key field — the same key powers all five agents:

• OpenAI GPT family (GPT-4o, GPT-4o-mini and newer)
• Anthropic Claude family
• Google Gemini
• Custom OpenAI-compatible endpoints (self-hosted, Azure, OpenRouter …)

You choose the model, you control the cost, you keep your key private — nothing routes through my server. Run the Supervisor on the simple built-in regime policy with no key at all, or arm the full five-agent pipeline when you add one.

ADVANCED FEATURES

- Multi-agent debate pipeline with configurable depth (Fast / Balanced / Thorough)

- Whole-account auto-discovery, or supervise specific magics only

- Cooperative control of your own EAs via a free drop-in include; broker-side control of third-party EAs

- Three execution modes: Shadow (log) / Approve (manual) / Live (auto)

- Independent hard risk governor: equity floor, daily & weekly caps, fleet-wide news blackout

- File-based Q&A agent — ask the Supervisor about your fleet in plain text

- Per-decision JSONL audit trail + shadow log

- Mass-concurrency safe — per-magic file and global-variable namespacing

- Built-in fleet-health dashboard panel with one-click manual override (pause all / resume / force-close)

…

- EA CONFIGURATION: