We watch jobsites. Not workers.

No facial recognition.

Our models detect people, equipment, and activity. They do not — and will not — match faces against any identity database.

No worker scorecards.

Workers are tracked only as anonymous IDs for counting and activity. We do not link a track to a named person or build per-employee scorecards — the dashboard answers 'how was the slab pour,' not 'how was Marc.'

No selling your data.

Your jobsite data is yours. We never sell it, license your footage, or share it with insurers. We may use footage to improve our own detection models — and only if you allow it.

No forced retention.

Raw video is deleted after processing unless you opt in to retention. Keep it to improve accuracy, or do not. Structured data (counts, durations, zones) is always yours to export or delete.

No surprise access.

Tectiv staff access your data only to operate the service or when you ask us for help — never to browse. Per-access audit logging is on our roadmap.

Camera

RTSP stream

Your hardware. Tectiv never owns the camera.

Secure ingest

Tectiv backend

The stream reaches our backend over the network; RTSP credentials are encrypted at rest.

GPU inference

Dedicated GPU

Frames are analyzed on a GPU. Detections — counts, zones, activity — are extracted; annotated, timestamped frames are retained per your settings.

Your dashboard

Vercel + Supabase

Counts, durations, and zones. Footage is retained per your settings and deletable on request.