Private pilot · May 2026

Iran Monitor Inference API

Submit a novel Iran-conflict scenario. Get a probability, per-framework reasoning, and a signed audit record drawn from a verified intelligence base.

For institutional risk desks. Asynchronous API; ~25 min per Premium query. Built on the same perspective stack as the canonical 6-hour cycle.

What it answers

Beyond the 8 canonical scenarios our cron-driven monitor tracks, you can submit a bespoke scenario and have it assessed through 14 independent perspective agents — Tetlock-style superforecasting, Schelling bargaining, Wack predetermined-elements analysis, plus domain lenses on IRGC, regional security, nuclear posture, and US-policy options.

Example queries pilot buyers have asked:

P(Iran-attributed cyber attack on German critical infrastructure in 30d)
P(formal bilateral oil-for-yuan agreement between Iran and China in 30d)
P(Houthi multi-vessel Red Sea attack triggering Article-5-equivalent posture in 30d)

What you get back

Probability + interval

Aggregated point estimate and uncertainty interval. Divergence flag when perspectives spread > 15pp.

14 perspective agents

Each agent's own probability + reasoning, grounded in published methodology. Independence enforced at the process level (verified by spike).

Major disagreements

Where perspectives split, what each side argues, why their frameworks lead them there.

High-elasticity events

What would move the probability >5pp up or down, with specific signals to monitor.

Briefing

Human-readable Markdown briefing your risk officer can drop into a memo. 500–1500 words.

Signed audit record

Ed25519 signature over the canonical JSON. Bind every output to the exact intelligence base used.

How a query looks

POST https://iran-a2a.globaltradealert.org/v1/queries
Authorization: Bearer <pilot-key>
Content-Type: application/json

{
  "scenario": "Iran (IRGC or proxy) launches a meaningfully disruptive
               cyber attack on German critical infrastructure within
               the next 30 days, with German government public attribution.",
  "horizon":  "30d",
  "tier":     "premium",
  "deliver_to": "risk-desk@your-fund.com"
}
↓
202 Accepted: { "query_id": "...", "estimated_completion_utc": "..." }

Poll GET /v1/queries/{id} until status: "completed" (~25 min for Premium). The completed response includes result.briefing_markdown, structured per-perspective outputs, and the signed audit record.

Same API surface for direct backend integration, a Python SDK, or your own LLM agent invoking the API as a tool. See ONBOARDING.md for the three integration patterns.

What makes this different

Multi-perspective by construction. Not a single LLM response — 12 independent framework agents, each grounded in published methodology.
Verified intelligence base. Sources fetched, publication dates verified, peripheral signals tracked across cycles. Same base the canonical 6-hour report uses.
Auditable. Every answer ships with the Ed25519-signed intelligence-base hash. Reproducible. Disputable.
Disagreement is the product. Where the perspectives split is often more useful than the consensus number — that's what surfaces fat-tail risk before it becomes consensus.

Request access

Pilot access is currently limited to institutional risk desks (hedge funds, sovereign wealth, commodities, reinsurance). Send a short note describing your use case and we'll respond within two business days.

Request pilot access →

Or contact johannes.fritz@sgept.org directly.

Limitations

Novel scenarios cannot be Brier-calibrated against ground truth — quality signal is reasoning-trace integrity + inter-perspective divergence, not historical calibration.
Iran-focused. Other conflicts (Ukraine, Taiwan) are separate monitors.
Async by design — no synchronous chat surface in the current product.
No free tier; pricing per pilot bilaterally.