Skip to main content
LIVE INTELLIGENCE PLATFORM

Geopolitical
Intelligence.
Operationalized.

Sovereign AI platform scoring 208 countries across 5 PESME axes in real time. ML-powered briefings. 3D threat visualization. Crisis war room.

Try as Guest — Full AccessRequest Demo
Powered by ORDRPoli AI — ML · RL · SLM
LIVE
Global Threat Monitor·12 active
—Events Ingested
—Risk Scores
—Countries Tracked
—Sources Indexed
—Knowledge Entities
—AI Briefings

Platform

See It Running

Command Center

Unified intelligence dashboard with real-time KPIs, threat nexus, and AI briefing summaries.

Enter Platform

Risk Scoring Engine

Five-Axis Geopolitical Risk Model

POL

ECO

SOC

MIL

ENV

208 countries scored across 5 axes in real-time. The PESME model quantifies Political, Economic, Social, Military, and Environmental risk — computed every 60 seconds from 65,131 live data sources.

Risk Distribution

Intelligence Pipeline

65,131 sources. One operating picture.

Ten source registries feed a real-time NestJS + FastAPI pipeline. Every event is tagged, scored, and surfaced through 49 platform pages and 144 REST endpoints.

INGEST

65,131 sources across 10 registries

RSS, Twitter/X, government APIs, world media, NGOs, cabinet press, business and political leaders. Adaptive scheduler escalates polling during crises.

SCORE

PESME model, 60-second cadence

Per-country political, economic, social, military, environmental axes computed continuously. ML risk-computation worker with ISO3 country-code defense in depth.

BRIEF

5 Kimi services + 22 API modules

AI briefings, cascade predictions, alert enrichment, market narratives, analytics synthesis. Every call is metered, sanitized, and tenant-scoped.

Source registry breakdown

30,897

RSS Feeds

25,253

Twitter/X

2,299

Officials

1,973

Media

1,500

Business

1,007

Influencers

1,004

Politicians

822

Cabinet

312

NGOs

64

Gov APIs

Adaptive scheduler

Crisis-aware polling. Sources tied to active war-room countries are pulled every minute; quiet regions cycle hourly. Backpressure prevents fan-out spikes.

BullMQ + Redis

Job queues for ML sync, risk computation, briefing generation, and webhook dispatch. Failed jobs retry with exponential backoff; dead-letter topology visible in admin.

Defense in depth

ISO3 country code validated at three layers: source-side normalization, pre-enqueue regex filter, worker entry check. Postgres check constraint catches the rest.

PESME computation

Five axes per country. NLP entity extraction, sentiment, severity, and event clustering feed the deterministic scoring kernel. Recomputed on every event ingestion.

Last ingestion: —

AI Intelligence Layer

Five Kimi-backed services. One disciplined contract.

Every AI surface follows the same pattern: versioned prompt, structured payload, fail-open fallback, per-field sanitization, cost metering. No hallucinated state ever reaches the database.

Cascade Predictor

cascade.prediction.v3

What it does

Probabilistic scenario branching for a trigger event.

Returns a multi-order consequence tree (2 to 5 levels deep) with per-node probability, timeline, confidence band, indicators to watch, and a historical analog when one exists.

POST /intelligence/cascade

Guardrails

  • Output depth and breadth capped to prevent malicious LLM responses from exploding the JSONB row or the V8 stack
  • Input bounds: trigger ≤ 2000 chars, focus arrays ≤ 50 items
  • Every domain, confidence, and probability value is clamped against a fixed enum before persistence

Copilot

copilot.suggestions.v3 / copilot.actions.v1

What it does

Live analyst assistant. Suggests relevant questions and translates assessments into actionable commands.

Suggestions endpoint surfaces 3 to 5 cued questions based on top risk countries, recent events, and unresolved alerts. Actions endpoint turns an assessment into concrete next steps: create alert, open scenario, build cascade, generate briefing, mark on globe.

GET /intelligence/copilot/suggestions · POST /intelligence/copilot/actions

Guardrails

  • DTO-bounded input (question ≤ 4000 chars, answer ≤ 8000 chars)
  • Server-side answer slice (800 chars) before prompt assembly
  • Action type clamped to a five-value enum so the UI never receives an unknown intent

Alert AI

alert.ai-context.v2

What it does

On-demand deep context for any triggered alert.

Returns an executive summary, severity assessment, escalation path (multi-stage), historical analogs with similarity scores, market-impact items per asset, and prioritized recommended actions with rationale and owner.

GET /alerts/:id/ai-context

Guardrails

  • LIKE-wildcard escape on alert.title before Postgres `contains` queries (108k-row table)
  • Tenant-scoped at the Prisma extension layer — no cross-org leakage possible
  • Fail-open: deterministic fallback context when KIMI_API_KEY is unset

Markets Narrative

markets.narrative.v2

What it does

Natural-language explanation of a forecast move or correlation shift.

Wraps the deterministic markets engine with prose: why the geopolitical-risk overlay moved the forecast cone, which analog events drive the comparison, what positions are most exposed.

GET /markets/narrative?symbol=...

Guardrails

  • Rate-limited at 15 calls per 120s per workspace
  • Costs accounted in `llm_usage` against the calling user and prompt id
  • Symbol parameter validated against the instruments catalog before any LLM call

Analytics Narrative

analytics.synthesis.v3

What it does

Prose layer on the platform analytics radar.

Synthesizes how the workspace is using the platform: regions of focus, query archetypes, alert pickup latency, briefing engagement. Surfaces patterns a dashboard alone would not.

GET /analytics/narrative-radar

Guardrails

  • Deterministic offline fallback when Kimi is down — no blank panels
  • Tenant-scoped data inputs, no cross-org context leak in the prompt
  • Output sanitized through the same per-field clamp pattern as every other service

Fail-open contract

If KIMI_API_KEY is missing, every service returns a deterministic, structured fallback rather than throwing. The platform stays usable end-to-end without an LLM provider. When a key is present, every call is metered to llm_usage with the prompt id, token counts, latency, and USD cost — visible per workspace in the billing surface.

Platform Surface

22 modules. 49 pages. One operating picture.

Every module is wired directly into the live API — no demo data, no mock layer. Each card below is a real route in the platform.

risk

Risk Engine

The five-axis PESME model scoring 205 countries every 60 seconds. Heatmaps, comparison, war-room aggregation, and scenario stress.

globe / graph

Globe & Knowledge Graph

Spatial and relational layers of the same data. WebGL globe for geographic reasoning, force-directed graph for entity reasoning.

Who This Is For

Built for serious operators.

Each persona below maps to a real workflow across real platform routes — not a brochure summary. Hover a step to see the page it lands on.

Intelligence Analysts

Run the full OSINT cycle on a single pane. From the raw event firehose to a defensible written assessment, every artifact is traceable back to source.

Operating outcomes

  • Cut briefing assembly time from hours to minutes — Kimi drafts the prose, you keep editorial control
  • Never lose a thread — Copilot recall lists your last queries, last cascades, last alerts in one place
  • Cite everything — every event card links to original source URL, hash, ingestion timestamp

Day in the platform

  1. 01

    Triage the firehose with severity + country filters

    /events

  2. 02

    Streaming natural-language query with cited evidence

    /ask

  3. 03

    Pivot on entities to find connections you missed

    /knowledge-graph

  4. 04

    Run a probabilistic consequence tree on the trigger

    /cascade-intelligence

  5. 05

    Draft, version, export PDF/CSV with full audit trail

    /reports

Risk Managers

Watch 205 countries in real time. PESME scores update every 60 seconds, war room aggregates active crises, and scenarios let you stress portfolios before the move.

Operating outcomes

  • See the regime shift before the headline — adaptive scheduler escalates polling when a country heats up
  • Build defensible counterfactuals — scenarios pin assumptions, propagate impact, surface second-order effects
  • Tie geopolitics to P&L — markets overlay shows which positions the regime change actually touches

Day in the platform

  1. 01

    Live heatmap → PESME matrix per country → market overlay

    /risk-analysis

  2. 02

    Open multi-crisis tabs, scrub the timeline, export JSON for stakeholders

    /war-room

  3. 03

    Build counterfactual, run impact model across instruments + regions

    /scenarios

  4. 04

    Slice country × threat axis × time window to compare regimes

    /risk-cube

  5. 05

    Promote pattern to a rule with AI-enriched escalation context

    /alerts

C-Suite Executives

Three minutes a morning. Read the daily AI briefing, scan the command-center band, then dive only where the deltas matter. Strategic, not encyclopedic.

Operating outcomes

  • Decision-grade briefings — every claim is sourced, every confidence is scored, every analog is dated
  • Spot the change, not the noise — 30-day risk timeline + top-10 target package surface only what moved
  • Translate intel to action — Copilot turns an assessment into "create alert, open scenario, draft briefing" with one click

Day in the platform

  1. 01

    Daily / flash / weekly briefings with confidence scoring

    /intelligence

  2. 02

    Executive situation board with operational ticker

    /command

  3. 03

    Geopolitical-risk overlay on watchlist, forecast cones, narrative prose

    /markets

  4. 04

    Narrative radar on workspace usage patterns and pickup latency

    /analytics

  5. 05

    Pull versioned briefings for the board package

    /reports

Government & Defense

Sovereign deployment, air-gapped enclave, no LLM dependency required. Every page degrades gracefully without Kimi; the deterministic engine is the platform.

Operating outcomes

  • Run fully air-gapped — on-prem Postgres + Redis, self-hosted inference, no external LLM call needed
  • Audit every action — actor, organization, IP, user agent recorded on every mutation, GDPR-aware erasure built in
  • SSO and provisioning that meet the bar — SAML with XML-DSig signature validation, SCIM ready, session revocation on credential change

Day in the platform

  1. 01

    Multi-crisis operations with timeline scrubber + classified annotations

    /war-room

  2. 02

    Deck.gl WebGL globe — country-level risk coloring + tactical layers

    /globe

  3. 03

    Stress trees for doctrine validation

    /cascade-intelligence

  4. 04

    SAML SSO, 2FA, audit log, GDPR erasure workflow

    /settings/security

  5. 05

    Scoped programmatic access with SSRF-guarded webhooks

    /settings/api-keys

Security & Compliance

Production controls. Not a slide.

Every control listed below is wired into the running code, not a policy document. Each item is pinned by a test that fails if the defense regresses.

144

API endpoints

24

Tenant-scoped models

102

Test suites green

22

Hardening commits this sprint

Identity & Access

  • httpOnly cookie auth, 15-minute access tokens, 7-day refresh tokens
  • bcryptjs hashing at 12 rounds, 10+ character policy with mixed-case, digit, and symbol requirements
  • SAML SSO via @node-saml/node-saml with XML-DSig signature validation
  • TOTP-based two-factor with backup codes and step-up auth on credential changes
  • Per-session revocation on password change — every old session is invalidated

Tenant Isolation

  • Prisma client extension auto-injects organizationId on every read and create
  • Check-then-act pattern on single-row mutations — no cross-tenant updates possible
  • Explicit TENANT_BYPASS list for global-derived models, audited on every PR
  • Transactional writes pass organizationId explicitly to every query
  • WebSocket gateway verifies JWT from handshake auth, never from query string

Role-Based Access

  • Four-tier RBAC: owner, admin, analyst, viewer with workspace-owner override
  • API keys carry independent read / write / admin scopes enforced per HTTP method
  • JwtAuthGuard + RolesGuard composition required on every mutation endpoint
  • Mass-assignment defense at the service layer — explicit allowlist for every editable field
  • Audit log records every action with actor, organization, IP, and user agent

Outbound Safety

  • SSRF guard on every webhook destination — blocks private IPs, link-local, credentials in URL
  • Stripe webhooks verified with HMAC-SHA256 + timingSafeEqual + 5-minute replay window
  • LIKE-wildcard escape on every user-controlled `contains` query against large tables
  • Bounded JSON validators on alert rule conditions (16 KB cap) and ingestion payloads
  • Output recursion caps on AI responses — depth and breadth bounded before persistence

Privacy & Data Rights

  • GDPR data-erasure workflow strips IP and user agent from audit logs in the same transaction
  • Intelligence query history is deleted as part of erasure, not just user records
  • Active sessions filtered by revokedAt and expiresAt — no stale token usage
  • Per-workspace data retention controls with automatic cleanup jobs
  • Customer data never sent to any LLM provider without an active KIMI_API_KEY configured by the workspace

Cost & Reliability

  • Every LLM call recorded to llm_usage with prompt id, tokens, latency, and USD cost
  • AI services rate-limited per workspace at the throttler tier (default, ai, narrative)
  • DTO input bounds prevent multi-megabyte payloads from reaching paid endpoints
  • Fail-open AI contract: deterministic fallback when the provider is unavailable
  • Render container bootstrap runs prisma migrate deploy, baselines only on P3005

Sovereign deployment option

Government and defense customers can run PoliSophic in an air-gapped enclave with on-premise PostgreSQL, Redis, and self-hosted inference. The Kimi narrative layer is optional — every page degrades gracefully without it.

Pricing

Intelligence at Every Scale

From individual analysts to sovereign government deployments. Every plan includes real-time data.

Explorer

$0/month

1 seat

  • Basic PESME risk scores
  • 30-day data retention
  • 5 AI briefings/month
  • Email support
  • 5 countries monitoring
Start Free
Most Popular

Analyst

$299/month

Up to 5 seats

  • Full PESME scoring + history
  • 1-year data retention
  • 100 API calls/day
  • Unlimited AI briefings
  • War Room access
  • 3D Globe + Threat Cube
  • Knowledge Graph
  • PDF/CSV exports
  • Custom alert rules
  • Collaborative workspaces
  • Priority support
Start 14-Day Trial

Enterprise

$999/month

Unlimited seats

  • SSO / SAML integration
  • Custom data retention
  • Dedicated support engineer
  • On-premise deployment option
  • Custom data source integration
  • SLA guarantee
  • Audit log + compliance reports
  • Custom AI model tuning
  • White-label option
  • Unlimited API access
  • Unlimited countries
Contact Sales

Ready to operationalize intelligence?

Full platform access. No credit card required.

Try as Guest — Full AccessRequest Demo