A Detailed In Design Product

Detect What Others Miss.
Online or On Prem.

Your analysts are drowning in dashboards. Your teams need shared access, but some environments still require local control. Narrative Capture Radar is a hosted narrative intelligence workspace with an optional Rust desktop deployment for offline and high-security environments — where missing a critical event is never treated the same as a false alarm.

Hosted collaboration • Optional Rust desktop • Deploy to the environment your mission requires

Narrative Capture Radar v1.0
EVIDENCE WINDOW
TOPOLOGY
VIABILITY
REGIME STATE

Built for teams in

Defense & Intelligence Financial Services Healthcare Critical Infrastructure Research & Academia
0:1
Max Bayes factor conviction
0
Parallel evidence windows
0 hr
Offline grace period
0
Data sent to the cloud

Sound Familiar?

These are the problems your team faces every day in the information environment. They don't have to.

Symmetric Collapse

Your current tools treat a missed cancer diagnosis the same as a false alarm. They optimize for "accuracy" — a metric that hides the most expensive failures behind a sea of correct trivial predictions.

NCR encodes domain-specific asymmetric costs into every detection decision. Missing what matters costs more than crying wolf.

Cloud Lock-In & Data Risk

Your most sensitive operational data sits on someone else's servers. You signed a BAA. They signed a multi-tenant architecture. Every query you run trains their next model. Every breach exposes everyone.

NCR runs entirely on your machine. Local SQLite. No telemetry. No API calls shipping data upstream. Your data is yours. Period.

Static Objective Drift

The world changes. Your detection system doesn't. That loss function tuned six months ago no longer matches reality, but nobody updates it because nobody knows which knob to turn without breaking everything else.

NCR evolves its expectations autonomously through safety-gated evolution. Three gates ensure updates never remove invariants or breach survival bounds.

Not Another Dashboard.
An Intelligence Engine.

Most platforms give you charts and hope you figure it out. NCR gives you grounded narratives, accumulated evidence, and expectations that evolve with your domain.

Not All Errors Are Equal

Asymmetric cost encoding from 3:1 to 81:1 Bayes factors. Multi-window evidence accumulation across fast, medium, and slow timescales. The system doesn't cry wolf — it accumulates conviction before alerting.

Bayes factors: 3:1 (anecdotal) → 81:1 (decisive)

Your Data. Your Machine. Period.

Built as a native desktop application. No Electron bloat. No cloud telemetry. No API calls shipping data to someone else's server. Local SQLite persistence, AES-256-GCM encryption, hardware-bound licensing.

AES-256-GCM | Argon2id key derivation | Ed25519 signatures

From Signal to Story in Real-Time

NCR doesn't just flag anomalies. It tracks violation topology, detects regime changes through structural shifts, and generates grounded narratives explaining what changed and why. LLM-powered with domain-gated security.

Domain-gated LLM | Typed admission | Content validation

Ingest From Anywhere

CSV files, JSON payloads, WebSocket streams, REST API polling, or manual observation entry. Paste data and NCR auto-detects the format. Built-in synthetic simulation for testing and training before going live.

CSV | JSON | WebSocket | REST polling | Synthetic simulation

Cross-Platform Desktop Native

One native codebase compiles to Windows, macOS, and Linux. No browser required. No server to maintain. Download, activate with your license key, and start detecting. Deploys in minutes, not months.

Windows (.msi) | macOS (.dmg) | Linux (.deb / .AppImage)

From Raw Data to Grounded Narratives

Three steps. Minutes to deploy. No consultants required.

1

Define Expectations

Set your domain expectations — what "normal" looks like, what violations cost, and how much evidence is required before acting. Or let NCR learn from your data and propose expectations automatically.

2

Stream & Detect

Feed data via CSV, WebSocket, REST, or JSON. The Clear Night Engine accumulates evidence across three parallel evidence accumulation windows. Violations require statistical consensus — not a single spike — before triggering alerts.

3

Understand & Evolve

Get grounded narratives explaining what changed, why it matters, and what to do next. The engine proposes expectation refinements through safety-gated evolution — shadow-tested before deployment.

The Clear Night Engine

A constraint-first epistemology. Not another dashboard.

Most monitoring tools predict what will happen next and alert when predictions fail. NCR inverts this. It starts with expectations — what the world should enforce — and only acts when those expectations are violated beyond tolerance.

The engine follows an efficiency-first design: if all expectations hold, no computation occurs. This isn't lazy — it's epistemically precise. Resources focus where violations emerge, not where metrics fluctuate.

Observation
↓
Violation Detection
↓
Evidence Accumulation
↓
Topology Update
↓
Narrative Generation
↓
Safety-Gated Evolution
Narrative Capture Radar
Narrative Capture Radar desktop application showing real-time expectation violation detection with radar topology, evidence accumulation, co-occurrence heatmap, and LLM narrative generation

What You See

A command center, not a confetti cannon of alerts.

Source Monitor: Live WebSocket streaming status, observation counts, violation rates, viability scoring, and transport controls for streaming, stepping, or resetting your data feed.

Violation Topology: Radar chart showing violation magnitudes across all active expectations. Shape changes between baseline and current state reveal structural shifts before individual metrics do.

Co-occurrence Heatmap: Which expectations violate together. Structural tracking detects regime shifts in the violation graph structure — catching systemic patterns that per-variable monitors miss.

Narrative Generation: LLM-powered explanations of what changed, why it matters, and what to do next. Domain-gated security layer prevents prompt injection. Evidence-grounded, never hallucinated.

Every panel updates in real-time as data streams in. No refresh. No polling. No waiting.

The $200K Question

You're evaluating options. We respect that. Here's what you're actually comparing.

Capability Cloud SaaS Platforms Narrative Capture Radar
Data residency Their servers (multi-tenant) Your machine (local SQLite)
Pricing transparency "Contact sales" Published pricing
Offline operation No — requires cloud connection Yes — 72-hour grace period
Asymmetric cost encoding No — symmetric loss functions Yes — domain-configurable 3:1 to 81:1
Evidence thresholds Alert on single spike Multi-window statistical consensus
Security framework Cloud security model Native desktop + AES-256-GCM
Expectation evolution Manual rule updates Safety-gated autonomous evolution
Desktop native No — browser only Yes — Windows, macOS, Linux
Annual cost $30K – $250K+ From $1,430.40/yr

We're not competing on brand recognition. We're competing on architecture.

The Math That Sells Itself

Cloud narrative intelligence platforms cost $30K–$250K+/yr. Here's what the alternative looks like.

Without NCR (Cloud SaaS)

Annual platform license $100K–$250K+
Data residency compliance risk $500K+ exposure
False negative cost (missed threats) $2M+ at risk
Overlapping tools your team manages 3–5 tools
Weekly alert triage hours (analyst time) 20+ hrs/week

With Narrative Capture Radar

Annual desktop license (Starter) $1,430.40/yr
Data stays on your machine $0 exposure
Asymmetric detection catches what matters 90%+ risk reduced
One platform, full pipeline 3–4 eliminated
Evidence-based alerts only 2–3 hrs/week

First-year savings: $100K–$250K+  •  Desktop license from $1,430.40/yr  •  Zero data residency risk

Security Is Not a Feature.
It's the Architecture.

Your detection platform should be the most secure tool in your stack. Ours runs natively on your machine, with zero cloud dependencies.

Built as a Native Desktop Application

Compiled native binary. No JavaScript runtime vulnerabilities. Memory-safe by default. Desktop-native performance with no browser overhead.

AES-256-GCM Encryption

All local data encrypted at rest. Hardware-bound license keys with machine fingerprinting. Your encryption stays with you — we never hold your keys.

Zero Cloud Dependency

No telemetry. No phone-home. No data exfiltration risk. NCR operates with a 72-hour offline grace period — no internet outage stops your detection.

Data Never Leaves Your Machine

Local SQLite database. No multi-tenant risk. No shared infrastructure. Your operational data, your expectations, your violations — all local.

Domain-Gated LLM Security

LLM narrative generation is protected by typed admission fields and content validation. Prompt injection attacks are blocked before reaching the model.

SOC-2 Available

Optional +20% add-on for compliance-regulated environments. Available on every tier — Starter through Enterprise.

Where Asymmetric Detection Matters Most

When the cost of missing a threat dwarfs the cost of a false alarm, symmetric tools fail. NCR doesn't.

10:1 Cost asymmetry ratio
Healthcare Compliance
Monitoring patient safety expectations where a missed violation — an unreported adverse event, a protocol deviation — costs ten times more than investigating a false alarm. NCR's asymmetric encoding ensures the engine errs on the side of patient safety.
Patient Safety & Compliance Hospitals • Research Institutions • Pharma
81:1 Maximum Bayes factor threshold
Financial Fraud Detection
Transaction monitoring where missing fraud is catastrophic. At 81:1, the engine demands overwhelming evidence before dismissing a suspicious pattern. Three parallel evidence windows prevent both premature alerts and dangerous silence.
Fraud & Risk Operations Banks • Insurance • FinTech
72hr Offline operation grace period
Critical Infrastructure
Air-gapped environments, intermittent connectivity, or field operations where cloud access is unreliable. NCR runs entirely on your machine with local persistence, hardware-bound licensing, and full detection capability — no internet required.
Infrastructure & Defense Energy • Utilities • Government • Military

Transparent Pricing. No Procurement Theater.

While competitors hide behind "contact sales," we show you exactly what it costs. Every plan includes the full Clear Night Engine. Tier 2 & 3 include a 20% discount.

Starter
$149.00 /month • $1,430.40/yr
No setup fees — desktop download

For analysts and small teams ready to move beyond symmetric detection.

  • Full Clear Night Engine
  • Narrative tracking & sentiment analysis
  • Asymmetric cost encoding
  • 3 parallel evidence windows
  • CSV, JSON, REST ingestion
  • Local SQLite persistence
  • 72-hour offline grace period
  • Basic alerting
  • Email support
Recommended
Professional
$399.20 /month • $3,832.32/yr
No setup fees — desktop download

For teams that need the full intelligence pipeline with LLM narratives and evolution gates.

  • Everything in Starter
  • LLM-powered narrative generation
  • Domain-gated security layer
  • Violation topology tracking
  • Evolution gates (Ginv, Gsurv, Gopt)
  • Shadow evaluation
  • WebSocket live streaming
  • Regime detection
  • Priority support
Enterprise
$1,199.20 /month • $11,512.32/yr
No setup fees — desktop download

For organizations with custom compliance needs and multi-seat deployment.

  • Everything in Professional
  • Multi-seat licensing
  • Custom policy bundles
  • Advanced regime analysis
  • Custom ingestion adapters
  • SOC-2 compliance option (+20%)
  • Dedicated support
  • Custom integrations
  • SLA guarantee

Desktop app — no setup fees. SOC-2 certification available on all plans (+20%). Annual billing saves ~20%. All plans include cross-platform desktop app (Windows, macOS, Linux).

Questions We Get Asked

Straight answers. No sales speak.

SIEMs and anomaly detectors flag deviations from learned baselines. They treat every deviation equally, alert on single spikes, and require you to figure out what it means. NCR inverts this: you define expectations (what the world should enforce), the engine accumulates evidence across three parallel time windows, and only alerts when statistical consensus is reached. More importantly, it encodes asymmetric costs — because in healthcare, fraud detection, or critical infrastructure, a missed event costs orders of magnitude more than a false alarm.

Three reasons. First, data sovereignty: your operational data never leaves your machine, which matters for defense, healthcare, finance, and any domain with regulatory constraints. Second, reliability: NCR works with a 72-hour offline grace period — no internet outage stops your detection. Third, performance: Native compiled code is faster than a JavaScript web app running through three layers of cloud infrastructure. No latency, no cold starts, no API rate limits.

Traditional systems optimize for accuracy — minimizing the total number of errors. But not all errors are equal. Missing a cancer diagnosis (false negative) is catastrophically more costly than ordering an unnecessary biopsy (false positive). NCR lets you define domain-specific cost ratios: healthcare might use 10:1, fraud detection 8:1, financial monitoring 5:1. The engine then accumulates evidence proportionally, requiring stronger proof to dismiss a potential threat than to flag one. This is based on our published research in statistical evidence accumulation.

NCR supports CSV file import, JSON payloads, live WebSocket streams, REST API polling at configurable intervals, and manual observation entry. Paste data into the application and it auto-detects the format. There's also a built-in synthetic data simulator for testing and training your expectations before connecting real data sources.

NCR generates human-readable explanations of violations and system state. The LLM integration is protected by a domain-gate security layer that validates all prompts through typed admission fields and content validation before they reach the model. This prevents prompt injection attacks. If the LLM is unavailable or you prefer not to use it, the system falls back to structured template-based explanations — every narrative is grounded in actual violation data, never hallucinated.

Windows (MSI installer), macOS (DMG package), and Linux (DEB package and AppImage). One native codebase compiles to all three. Activate with your license key (DID-XXXX-XXXX-XXXX-XXXX format) and start detecting immediately. The license is hardware-bound with machine fingerprinting, supporting seat-based enterprise deployment.

Every Hour Without Asymmetric Detection
Is an Hour Where Your Most Expensive
Mistakes Look Identical to Noise.

Request access to Narrative Capture Radar. See the difference between a dashboard that shows you everything and an engine that tells you what matters.

Native Performance No Cloud Dependency AES-256-GCM Encrypted Cross-Platform