Video Systems
Camera estates grow by accretion. A few hundred cameras from one integrator, a few hundred more from an acquisition or a grant cycle, and suddenly there are six management consoles, four codecs, and no single answer to how much footage exists or where it lives.
Mentat consolidates that into one operating layer. Cameras become inventoried devices with health telemetry. Streams are normalized. Storage is tiered so that the last 72 hours is instantly available and the last 18 months is affordable.
Analytics sit on top rather than being welded into a single vendor stack, so detection models can be added, evaluated, and replaced without re-architecting the estate around whichever vendor was chosen first.
What we build
Know which cameras are actually working. Not according to last quarter’s spreadsheet, but according to telemetry arriving right now.
ONVIF, RTSP, and proprietary vendor feeds converge into a consistent internal format so downstream tooling stops caring about camera brand.
Video is the largest line item in most estates. Tiering, retention policy, and honest capacity modeling are what keep it from consuming the budget.
Detection and search capabilities run as replaceable components against a common video substrate, so the estate outlives any single analytics vendor.
How we get there
Automated discovery of what is actually deployed, including the cameras nobody has documented since the integrator left.
A common ingest layer that ends the per-vendor integration tax and gives every downstream tool one interface to build against.
Retention and cost modeled together, so the policy you write is the policy you can afford to run for the full retention window.
Detection capabilities added incrementally, evaluated against real footage, and kept replaceable by design.
Governance & safeguards
The questions arrive eventually: who watched this, why, what does the analytics actually detect, and how well does it work on the population it is pointed at. A system built to answer those questions is worth far more than one that cannot.
Live view and recorded retrieval both carry attribution and purpose, with elevated review requirements for sensitive locations and archives.
Any detection model in production has documented performance characteristics and a defined human review step before it drives enforcement action.
Deletion is verified across every tier including archives and backups, with an auditable record that the policy executed.
Built on
We build on established open-source and cloud-native building blocks rather than proprietary black boxes, so the platform stays inspectable, portable, and yours.
Video layer
Storage
Platform
Related
Camera estate outgrowing its tooling?
We can inventory what you are actually running, model the real storage trajectory, and design the ingest layer that gets you off per-vendor integration work.