Skip to main content

PLATFORM · DEPLOYMENT

Your data. Your network. Your choice.

The same intelligence layer for your security operations center runs three ways: fully managed in our cloud, on-premise inside your own network, or fully air-gapped with no internet at all. One codebase, the same features in every mode. Where your most sensitive data lives is a decision you make — not a constraint the platform imposes on you.

Cameras
Any ONVIF / RTSP
On-site Agent
ONNX AI · local, on CPU
Cloud or On-Prem
events · faces · evidence
Operator Console
respond · investigate · govern
Edge AI · no cloud GPU · footage stays in your network
3 modes
One codebase
cloud · on-premise · air-gapped
< 30 min
On-prem live
five-step browser setup wizard
0
Cloud GPUs required
inference runs on the on-site agent
12
Languages
Arabic RTL first-class

DEPLOYMENT MODELS

Choose your deployment based on your data policy — not your feature list.

Every mode is the full platform. Pattern-of-life, cross-camera tracking, sealed evidence, and the tamper-evident audit all work the same way whether we host it or you do. Pick the model your mandate requires; the capability set never changes.

We run it. You keep control.

Cloud-managed

The fastest way to stand Sentinel up. We operate the platform on a hardened cloud cluster — Aurora MySQL, pgvector, object storage, Redis, TLS, and custom-domain routing. Stood up in minutes, automatic scaling and managed updates. Best for multi-site operations and teams without a server estate to manage.

Sentinel inside your own network.

On-premise

A self-contained Docker Compose stack — application, MySQL, pgvector, Redis, and MinIO object storage — all behind your firewall. An offline HMAC-signed license means no phone-home. A five-step setup wizard takes you from docker compose up to your first camera in under 30 minutes.

Fully offline. Nothing leaves the perimeter.

Air-gapped

For the most sensitive environments: zero internet connectivity required, ever. Install by transferring the image bundle on secure media; license validation happens entirely on your server. AI models can be sideloaded so even model updates never touch the internet.

DATA SOVEREIGNTY

Your footage. Your intelligence. Your network.

Cloud-only platforms ask you to trust that someone else is holding your most sensitive data correctly. Sentinel does not make you take that on faith. In on-premise and air-gapped mode, footage, face embeddings, event records, and audit trails all live on infrastructure you own and control.

Footage stays on your hardware

In on-premise and air-gapped mode, all video, face embeddings, event records, and audit trails live on infrastructure you own. Nothing has to leave your network for the platform to do its job — including the AI.

Same features in every mode

This is not a stripped-down local edition. Pattern-of-life, cross-camera tracking, lawful-basis tracking, sealed evidence, and the tamper-evident audit all work identically whether we host it or you do.

Accountability is mode-independent

The privacy controls, the lawful-basis requirement, and the tamper-evident audit are identical in every deployment mode. Accountability never depends on where you host.

ON-PREMISE

One stack. Your firewall. Every feature.

The on-premise package is complete and self-contained — no external services required. All services run as Docker containers, managed with a single docker compose up, with health checks, automatic restarts, and named volumes for data persistence. Every feature works offline: face recognition, cross-camera tracking, pattern-of-life, plate search, threat detection, evidence packs with chain of custody, and the tamper-evident audit.

Application (Next.js)
MySQL 8.0
PostgreSQL + pgvector
Redis
MinIO object storage

ON-SITE AI

The AI runs where the cameras are.

A lightweight on-site agent does the heavy lifting next to your cameras, so video is understood locally and never needs to make a round-trip to a cloud GPU. It pulls RTSP, relays it as HLS to the console, and runs the full vision pipeline in ONNX Runtime on your own hardware.

On-device AI in ONNX Runtime

Every model — faces, plates, people, vehicles, behavior, threats — runs on your own hardware with no cloud GPU and no video leaving the site for inference.

ONVIF & RTSP auto-discovery

The agent discovers ONVIF cameras automatically via WS-Discovery, or connect directly by RTSP URL. Multi-agent topology spans buildings and cities under one tenancy.

Auto-updates with rollback

The agent updates itself safely over the air and rolls back automatically if an update fails — no manual intervention required for the on-site component.

White-label ready

Built per partner brand: bundled logo, colors, product name, and about text. The software your customer installs on their own network carries your brand from the installer onward.

HOW IT WORKS

On-premise setup, live in under 30 minutes.

1

Pull the stack

Transfer the Docker image bundle to your server — via internet pull or secure media for air-gapped installs. A single docker compose up brings all five services online: application, MySQL, pgvector, Redis, and MinIO.

2

Open the setup wizard

A browser-based five-step wizard guides an administrator from a running stack to a streaming camera. Enter your offline license key, create the first organization and administrator account, and set the hostname.

3

Install the on-site agent

Download and run the agent installer on a Windows, Linux, or macOS machine on the same network as your cameras. The agent connects to the console and registers itself automatically.

4

Add cameras

The agent discovers ONVIF cameras on the local network via WS-Discovery, or add any RTSP stream by URL. Cameras appear in the console within seconds of being added.

5

Go live

The full platform — face recognition, plate search, cross-camera tracking, threat detection, evidence packs, the tamper-evident audit — is live on your own network. Every feature works offline.

SPECIFICATIONS

Deployment at a glance

On-site agent OS supportWindows, Linux, macOS — Windows service, systemd unit, or launchd daemon; system-tray UI
Camera compatibilityAny ONVIF camera; WS-Discovery auto-detection or direct RTSP; multi-agent topology across sites
AI processingOn-device in ONNX Runtime on the agent — no cloud GPU required; server-side AI worker available for batch re-processing
On-premise stackDocker Compose: application (Next.js) · MySQL 8.0 · PostgreSQL + pgvector · Redis · MinIO object storage
Cloud-managed stackManaged cluster: Aurora MySQL · PostgreSQL + pgvector · object storage · Redis · TLS and custom-domain routing
License modelOffline HMAC-SHA256 signed key, verified locally — no phone-home, air-gap friendly
Air-gap supportFull — install via image-bundle transfer on secure media; models sideloadable; zero internet requirement
Setup timeOn-premise live in under 30 minutes via the five-step browser setup wizard
UpdatesCloud-managed by us; on-premise via image pull and restart, with bundled database migrations that run automatically
BrandingWhite-label end to end — console, agent, and emails, under your own brand and custom domain
Languages12 supported; Arabic RTL first-class
Storage tiersHot (instant) · Cold (6-mo min, 3–5 h restore) · Deep Freeze (12-mo min, 12–48 h restore)

Specifications describe shipped platform capabilities. We will confirm the deployment configuration that fits your requirements during your demo.

Deploy on your terms.

Request demo access and we’ll configure a live environment for you — or walk you through the Docker deployment on your own hardware, end to end. Full feature access in cloud, on-premise, or air-gapped. The intelligence is the same in every mode; only the address changes.