School & Campus Safety Software

Pick your campus, see the coverage change

Hundreds of cameras, one safety team. Entrances, hallways, cafeterias and grounds watched continuously, with district policy deciding what happens next.

After-hours entry. Perimeter and restricted-area events raised the moment they occur.
Aggressive interaction. Activity detection flags physical altercations for a person to assess -- triage, not discipline.
FERPA-aware by design. Role-based access, retention controls and redaction before any footage leaves.

An open campus is the point -- and the problem. Libraries, residence perimeters, parking structures and event venues on one estate, watched without walling it off.

Campus police as buyer. Built for the public-safety operation a university actually runs.
Person-down detection. Falls, collapse and prolonged immobility raise an alert with footage preserved.
Cross-camera tracking. One person or vehicle followed across buildings during an active concern.

Commuter campuses, lean staff. Safety coverage that does not assume a 24/7 operations centre.

Alerts routed, not watched. Events reach the right phone instead of a wall of monitors.
Weapon detection. Brandished-weapon detection on existing cameras, verified by a person before escalation.
Runs on what you have. RTSP and ONVIF over the existing estate and VMS.

Small teams, full accountability. Enterprise-grade detection without an enterprise security department.

Zone rules you draw. Restricted areas and off-limits hours defined per camera.
Visitor anomalies. Unexpected presence flagged where and when nobody should be.
Evidence, not screenshots. Incidents land as records with custody intact.

Bus yards, loading zones and routes. The incidents that today surface as a parent complaint, found on footage in minutes.

Loading-zone monitoring. Activity around vehicles during boarding and drop-off.
Yard security. After-hours movement in depots and lots.
Complaint-to-clip. Natural-language search across transport footage instead of scrubbing.

What You Get

What a governed safety layer actually does

Detection that never looks away

Unauthorized access, aggressive physical interaction, crowding, person-down events and visual fire and smoke cues raised in real time from the cameras already installed.

Live AI DetectionZone Rules

Weapons, verified by people

Brandished-weapon detection flags the frame for immediate human verification -- the model raises it, a person confirms it, policy decides what follows.

Weapon DetectionHuman in the Loop

Context, seconds behind the alert

The detection is immediate. A multi-modal model then describes what the scene showed, so responders get more than a class label before they act.

Prompt-Driven Analysis

Response by policy, not by model

Workflows you draw decide who is notified, what escalates, and where a human approval gates the step. Commands to external systems go through your integrations -- the AI does not lock doors or dial 911 on its own.

Graph WorkflowsHuman Approval

The incident becomes a record

Footage around an event is preserved with chain of custody, retention policy and audited access, so the review afterwards works from evidence rather than recollection.

Chain of CustodyAudit Trail

Search instead of scrubbing

Ask across recorded video in plain language -- a person in a specific hallway after hours, a crowd forming near the gym -- and get the clip back.

Natural-Language Search

FERPA-ready disclosure

When footage must be released, faces and identifying detail are redacted on the copy while the original stays intact, with role-based access throughout.

RedactionRole-Based Access

Runs where the district requires

Cloud, on-premises or fully air-gapped with self-hosted models, so student footage stays inside the boundary the institution sets.

Self-Hosted InferenceAir-Gapped

How It Works

Detect, decide, preserve

Detect

Computer vision reads the live streams and raises an event the moment a condition is met -- real-time, deterministic, on the cameras already there.

Understand & decide

The event triggers a workflow. A multi-modal model describes the scene, and institution-defined policy determines the response path, with human approval gating anything consequential.

Act & preserve

People are notified, external systems receive validated events through your integrations, and the footage is preserved as an evidence record for the investigation and the review that follow.

FAQ

School and campus safety, asked and answered

What is school and campus safety software?

Software that turns the cameras a district or university already owns into a continuously watched safety layer. Computer vision detects the conditions you define -- after-hours entry, aggressive physical interaction, crowding, a person down, a brandished weapon -- and a policy-controlled workflow decides who is notified and what happens next, while the footage is preserved as a record.

Does it replace our access control, alarm or emergency notification systems?

No, and it should not be evaluated as one. Those systems execute door control, mass notification and dispatch. This provides the visual intelligence layer that can send validated events to them through integrations -- REST APIs, webhooks and the workflow node catalog -- rather than replacing them. There are no proprietary lock or notification systems here to buy.

Can the AI lock down the school or call 911 by itself?

It is not designed to. Response runs through workflows the institution defines, with human approval gating consequential steps and escalation rules set by policy, not by the model. An AI-raised event can reach an external system only through the integrations you configure, under the controls you set.

Does it detect bullying?

No system can reliably detect bullying, and claims otherwise deserve suspicion. What this detects is narrower and observable: aggressive physical interaction, crowding, unusual activity and scene context that helps a person assess what deserves attention. It is a triage aid for staff, not an autonomous disciplinary decision-maker.

How does weapon detection work?

Detection models watch for brandished weapons on the live streams and flag the frame the moment one appears. The flag goes to a person for verification before anything escalates, and confidence thresholds are configurable per camera. Detection accuracy gates the whole chain, so the verification step is a design feature, not a disclaimer.

Is the analysis real-time?

The detection is -- it reads the live stream and fires as the condition occurs. The added scene description runs on cached frames on a cadence, so it arrives moments after rather than in the same instant, and the alert does not wait for it.

How is student privacy protected?

Role-based access controls who sees footage, retention policy governs how long it exists, every access is audited, and redaction removes faces and identifying detail from any copy that leaves -- the original stays intact. Deployment can be on-premises or fully air-gapped where footage must not leave the institution, and school video handled as an education record stays inside those controls.

What happens after an incident?

The footage around the event is already preserved with chain of custody. Investigators search the estate in plain language, assemble what happened across cameras, and produce a record that survives a district review, a parent meeting or a legal process -- with a redacted copy for anything disclosed.

Do we need new cameras?

No. Detection runs over RTSP and ONVIF against the existing estate, alongside whatever VMS or NVR is already in place, with camera-health alerts so a dead camera is found before it matters.

Every Camera Watched. Every Response Governed. Every Incident Answerable.