One agent. Five ML signals. Pay per camera.
Cam-Mitra licenses are per-camera, billed monthly. No video storage fees, no per-event fees, no surprises — because inference runs on your hardware.
Trial
Try the full pipeline at no cost for 14 days.
- 1 camera included
- All 5 ML signals enabled
- WhatsApp + email alerts
- Smart alert rules
- Auto-learning loop
- Community support
Starter
One store, one camera. Perfect for small shops.
- 1 camera per store
- All 5 ML signals
- WhatsApp + email alerts
- PDF weekly reports
- 30-day clip retention
- Email support
Growth
Multiple cameras, multiple zones, smart rules.
- Up to 4 cameras
- All 5 ML signals + custom rules
- Multi-zone tripwires
- Heatmap analytics + dwell time
- 90-day clip retention
- Priority email support
- Auto-flag faces from theft events
Pro
Up to 99 cameras. Multi-site. SSO ready.
- Up to 99 cameras
- Multi-site / multi-store
- Custom SLA + dedicated support
- On-prem dashboard option
- 1-year clip retention
- API access + webhooks
- SSO + audit log (coming)
Your video stays on-premise
All five ML models run on a PC inside your store. Only JSON metadata reaches the cloud — never raw frames.
Cancel any time
Monthly billing. Cancel from the dashboard and keep access until the period ends. Export your data with one click.
14-day free trial
No credit card. One camera, all features. The system trains itself on your footage during the trial.
Common questions
Does my video footage actually leave my premises?
No. The edge agent runs on a PC inside your store. Only JSON metadata (~1KB per event) is uploaded to the cloud dashboard. The only exception is short, face-blurred clips attached to alerts — and those are never longer than 10 seconds.
What hardware do I need?
Any 4-year-old Intel Core i5 with 8 GB RAM is enough for one camera. For 4+ cameras we recommend adding a consumer NVIDIA GPU (GTX 1660 or better). The agent ships as a single Windows .exe — no Python, no Docker, no setup.
Can I bring my own cameras?
Yes — any IP camera that supports RTSP works. Hikvision, Dahua, CP Plus, Reolink, generic ONVIF. The agent reads RTSP via GStreamer for sub-50ms latency.
How does the auto-learning work?
When inference confidence lands between 35% and 65% (the “doubt band”), the agent uploads a short face-blurred clip. Gemini Flash labels it. Every 500 new labels triggers a YOLO fine-tune. If the new model's validation mAP doesn't regress, the agent hot-reloads it — no restart, no downtime.
What about the DPDP Act / GDPR?
Face-blurring happens at the edge, before any upload. Raw footage never crosses your network boundary. The clips we do upload are short, face-blurred, and stored encrypted in S3-compatible MinIO. We're built privacy-first because the entire pitch depends on it.
Can I cancel any time?
Yes. Monthly billing means you cancel from the dashboard and you keep access until the end of the current billing period. Your data export is one click away.