Your cameras, understood.
Muraqaba AI turns the cameras you already own into a sovereign intelligence layer — it doesn't just detect, it understands the scene, explains what happened in Arabic and English, learns your site, and answers your questions. Runs entirely in-Kingdom: sealed on-premises, or on the Sigmix KSA Cloud.
See Muraqaba AI understand a scene
Under a minute and a half: how Muraqaba turns your existing cameras into a sovereign intelligence layer — detecting, understanding and alerting in Arabic and English, on-prem or on the in-Kingdom KSA Cloud.
Not another motion-alarm box. A system that understands.
HIK, Dahua and legacy analytics fire fixed rules you hand-draw — and bury you in false alarms. Muraqaba reasons about the scene, explains it in Arabic, and tells you what actually matters. These are the things fixed-rule systems structurally cannot do.
Instant Accident Notification
The moment a worker falls, is struck, collapses, or a vehicle collides — Muraqaba doesn't just see motion, it understands the accident, judges its severity, and pushes a described alert to every screen and phone in seconds. In a warehouse or on a giga-project, those seconds are a life. No rule to draw, no trigger to configure.
Ask your cameras in Arabic
“Show me every time the loading door was left open this week.” Natural-language search across weeks of footage — a question, not hours of scrubbing.
Understands, not just detects
“A worker fell and isn't getting up.” “The forklift driver looks asleep.” Descriptions of what happened — not a box labelled motion.
Daily brief in plain Arabic
Every morning: “2 PPE violations at Gate 3, one after-hours entry at 23:00, queue exceeded 10 twice at midday.” No dashboard-diving.
Learns your site — no rules
It watches for a week, learns your normal rhythm, then flags what doesn't fit. The “minimal-config” magic legacy systems never had.
The alert-fatigue killer
Because it understands, it sends fewer, smarter alerts than motion-spam systems — severity-tiered, human-in-the-loop, each with a reason attached.
Sees and hears
Audio-visual fusion — a breaking-glass or aggression sound corroborates the video, or flags a blind spot the camera missed.
Two tiers of intelligence — cheap eyes, deep reasoning
The architecture is what makes it affordable and reliable: a fast detector watches every camera all the time; the expensive vision-language model wakes only when something warrants a closer look.
Always-on detection
A modern detector runs continuously on every camera — people, vehicles, objects, zones, lines, counts and tracking — many cameras per GPU, in real time.
Vision-language reasoning
When Tier 1 flags something, the VLM interprets the clip — confirms the event, judges severity, and writes a bilingual description. It never runs on every frame, so cost stays flat.
Fewer, smarter alerts
Because it understands, it sends far fewer false alarms than motion-spam systems — severity-tiered, human-in-the-loop, with a clip and a plain-language reason attached.
This is what your cameras have been missing — until now
Fire, collisions, a worker down, unsafe behaviour on the floor. A motion box can't tell these apart from a passing truck. Muraqaba understands the scene, judges severity, and raises a described alert in Arabic and English — in under a second.
DETECTED
DETECTED
DETECTED
DETECTEDReal footage analysed by Muraqaba. Every alert carries a clip, a plain-language reason and a severity — so your operators act on what matters, not on motion spam.
Weeks of footage, answered in a sentence
Type or speak a question in Arabic or English and Muraqaba searches the meaning of what it saw — not just timestamps. “Every after-hours entry.” “The white van that parked and left within 90 seconds.” “When was the vault door opened?”
- Semantic, not keyword — it matches what happened, described in your own words.
- Arabic-native — the query, the answer and the reasoning, all in Arabic.
- Jump straight to the clip — every answer links to the exact moment.
A command dashboard your operators trust — bilingual to the core.
Live wall, AI incident feed, severity analytics and timeline playback — the entire experience works natively in Arabic and English, full right-to-left. Descriptions and briefings are generated by an Arabic-first model, not machine-translated.
One brain, many jobs — switched on per site
Enable what each site needs from an industry profile, then fine-tune with a checkbox. All capabilities run in Arabic and English.
Safety & PPE
Missing hard-hat, hi-vis or harness at height — per zone, in real time.
Fall & person-down
Someone collapses, is struck, or lies motionless too long → instant alert.
Danger-zone & vehicles
People too close to cranes, forklift paths, edges — near-miss detection.
Fire & smoke
Early visual detection of flames and smoke plumes in open spaces.
Intrusion & tailgating
After-hours presence, perimeter breach, two people on one badge.
Abandoned object
Unattended bags and packages near entrances and lobbies.
Loss & shrinkage
Concealment, asset removal, suspicious handling — flagged for human review.
Operations analytics
Queue length, dwell, occupancy heatmaps, is-the-post-manned, footfall.
Responsible by design
Staff-productivity features are framed as safety and operations — never worker-punishment tooling; the customer sets policy. Anything consequential requires human review. Face recognition and watchlists ship disabled, behind an explicit admin gate with a full audit log of who enabled it and who is on the list.
Built for banks, government and critical infrastructure
Footage is the most sensitive asset a site holds. Muraqaba is engineered so it stays under your control — designed to support your obligations under PDPL, SAMA CSF and NCA ECC-2.
Honest by design: Muraqaba gives your security team the evidence and controls to meet their obligations. Regulators assess your organisation, not vendor products — we provide the mapping and the audit trail, you hold the compliance.
Choose where it runs — the software is identical
Sigmix KSA Cloud
For distributed estates — branch networks, retail chains, multi-site operators. Cameras stay where they are; a small on-site connector links each site over an outbound-only encrypted channel — no port forwarding, no exposed recorders, ever.
- Start with one camera, grow to thousands — elastic capacity, new sites live in minutes, no hardware to buy.
- Remote by nature — web and mobile access with real-time bilingual alerts, wherever your team is.
- Always current — models and features improve continuously, managed by Sigmix.
- Shared or dedicated — dedicated in-Kingdom GPU capacity for regulated clients.
Sigmix i-CCTV Appliance
For sovereignty-critical sites — government, banking, critical infrastructure, industrial campuses. A sealed AI appliance on your network; every frame analysed inside your building, no internet required, and by design no remote-access path. Zero egress means zero egress.
- All intelligence local — detection, reasoning, bilingual alerts and insights, on your metal.
- Your existing cameras — consumes Hikvision and standard IP cameras / NVRs over RTSP & ONVIF.
- Your channels — control-room dashboard, email, and webhooks into your own systems.
- Optional managed care — via a customer-controlled, fully audited outbound tunnel you enable and revoke.
Both run the same software and the same intelligence. Many clients combine them — sealed appliances at sovereignty-critical sites, KSA Cloud across the branch estate.
Appliance tiers, sized by camera count
Indicative sizing — final sizing is validated against your sites' activity. Grow between tiers by adding accelerators. (KSA Cloud scales elastically — no tier to choose.)
- Branch, showroom, villa compound, small warehouse
- Standard reasoning
- All capabilities · AR / EN
- Large branch, factory hall, logistics depot
- Standard reasoning
- All capabilities · AR / EN
- Enterprise campus, mall, large plant
- Enhanced reasoning
- All capabilities · AR / EN
- HQ campus, industrial complex, giga-project zone
- Flagship reasoning
- All capabilities · AR / EN
Indicative at typical site activity; sites with very high event rates are validated during sizing — the reasoning tier is sized to alert traffic, not just camera count. Storage scales with your retention policy (days × cameras). The appliance is a single-node product; run two units for continuous availability. Beyond 150 cameras on one site, or for multi-site estates, Sigmix engineering designs a clustered or cloud/hybrid deployment. Fixed annual license by tier, supported in-Kingdom.
Fixed-rule CCTV vs. Muraqaba AI
| Traditional analytics | Muraqaba AI | |
|---|---|---|
| What it does | Detects motion / objects | Understands the scene |
| Configuration | Hand-drawn rules per camera | Learns the site · minimal config |
| Search | Scrub the timeline | Ask in Arabic, get the moment |
| Alerts | Motion spam | Fewer, explained, severity-tiered |
| New camera brand | Often locked to one vendor | Your existing HIK / ONVIF cameras |
| Data & sovereignty | Varies / cloud-tethered | In-Kingdom · zero-egress option |
The alert reaches your team — wherever they are
On KSA Cloud deployments, operators get the full incident feed and bilingual push alerts on mobile — acknowledge or escalate a critical event from anywhere in the Kingdom. Sealed on-premises sites exclude remote access by design: zero egress means zero egress.
Answers for your security, operations & procurement teams
Do we need to replace our cameras?+
Does the on-prem appliance touch the internet?+
Where does the KSA Cloud store our footage?+
How accurate is it — and can we trust the AI's judgement?+
Is it really Arabic, or translated?+
What about face recognition and privacy?+
See Muraqaba AI on your own footage
Bring us a few clips from your site — we'll show you the incidents it finds, in Arabic and English, and walk your team through on-prem and KSA-Cloud deployment.
Muraqaba AI · مُراقبة — sovereign video intelligence, in-Kingdom.