A documented comparison for teams that shortlisted Lumana for being camera agnostic. Its detection catalog and its response layer are both published, so the open question is narrower: what the appliance in the middle replaces, and what happens when the detection you need is not on the list.
Lumana is camera agnostic in its own words, and its Core device replaces your DVR or NVR while storing and processing video on site. It publishes one of the fullest detection catalogs in this category, four named AI agents, a response layer down to loud speaker and lockdown, and an annual license set by camera feeds, days of storage and term. Teams shop for an alternative when they want the recorder decision left open, a way to build the detection nobody ships, or a response with nobody to verify it first.
Both platforms run on the cameras a site already owns. They differ on what happens to the box those cameras record to, and that is the part of the project a per-camera price does not show.
Every cell is something the vendor publishes, or an explicit not publicly specified.
Swipe the table sideways to see every column.
Lumana's row was read off lumana.ai, its pricing page and its Response product page, re-read on August 21, 2026, and it carries the camera-agnostic wording, the detection catalog, the named response actions and both of Lumana's own answers on the recorder. The other five rows come from each vendor's public documentation reviewed in July and August 2026.
Two of the reasons teams shortlist Lumana are not reasons to look elsewhere, so deal with those first. It is not a hardware-first vendor trying to sell you cameras: it states plainly that it is completely camera-agnostic and compatible with any IP camera supporting standard streaming protocols. And it publishes what it detects rather than describing it, roughly thirty named event types under four AI agents, with alerts drawn from over 100 triggers.
The first real reason is what sits in the middle. Lumana's own answer on whether Core works alongside an existing recorder is no: Core replaces your DVR or NVR, and Core devices store and process the video. That is a clean architecture, and for a site with an aging recorder it is an advantage, because one box leaves and one arrives. If your NVR is new or under warranty, the swap is a cost the quote will not show.
The second is the route to a detection the catalog does not cover. Lumana publishes roughly thirty event types and a continuously learning model behind 100 or more triggers, and no builder, no request process and no build time for anything outside that. The other gap sits in the response path: the automated actions run in software, but emergency dispatch is published as AI response combined with human verification, and who verifies, at what hours and to what response time is not stated.
Camera compatibility will not separate these platforms and neither will the detection list or the response layer, because Lumana publishes all three. What replaces your recorder, and what happens when the detection you need is not on the list, will.
What each one is, where it is strong, and what to check before you commit.
Spot AI is a software-led video AI platform that turns the IP cameras a business already owns into AI coworkers. An on-site Intelligent Video Recorder keeps full-resolution video in the building and sends only event metadata to the cloud, so search and multi-site management stay cloud-based while footage does not leave the facility.
Verkada is a cloud-managed platform that sells its own cameras, access control, sensors, alarms and intercoms under one console, and publishes per-device MSRPs. Third-party cameras come in through Command Connector, which Verkada states is an ONVIF Profile S conformant client and can also ingest RTSP feeds.
Eagle Eye Networks is a cloud video management system that connects to virtually any ONVIF-conformant camera and digitizes analog feeds through analog-ready Bridges and CMVRs, then layers cloud analytics on top of the estate. Its documentation names ONVIF Profile S and dual codec streaming as the integration path.
Coram AI is a camera-agnostic platform built around AI search and investigation across cameras a business already owns, with named detections for firearms, falls and PPE violations, and alerts that can be built in plain English. An on-site appliance called Coram Point is purchased upfront.
Rhombus sells cloud-managed cameras, sensors, access control and alarm monitoring as one system, described by the company as a cloud-edge system that operates offline. Third-party cameras come in through the Relay Connector N100, and the company publishes SOC 2, NDAA, GDPR and TAA compliance badges.
Six questions that separate these platforms faster than any feature list.
This is the question that separates the shortlist. Lumana states that Core replaces your DVR or NVR. Spot AI's Intelligent Video Recorder is added alongside the cameras and covered by the subscription. Coram Point is bought upfront. Eagle Eye moves the video to its own data centers. Four answers, four different first-year costs.
Name the certification pack before the demo rather than after it. Lumana publishes SOC 2 Type I and Type II alongside TLS 1.2 and AES-256 encryption, single sign-on through Okta SAML and multi-factor authentication. Rhombus states SOC 2, NDAA, GDPR and TAA. Spot AI states NDAA-compliant and SOC 2 practices. Eagle Eye's status is not stated on the pages read.
Pin down what happens after closing, when the alert has no one to reach. Eagle Eye documents sirens and talk-down alerts. Verkada gates automated audio to the BZ11 or an Intercom. Spot AI runs talk down, strobes and horns through standard speakers already on site. Lumana publishes staff notification, a loud speaker, lockdowns, GPIO and webhooks, plus emergency dispatch with human verification, so ask who verifies and when. Coram and Rhombus publish none.
Lumana publishes roughly thirty event types spanning weapon, fire, fall, violence, PPE, tailgating, loitering, tampering, object and vehicle events, and no way to add one it does not cover. Spot AI ships 15+ pre-trained Video AI Agents plus Iris for custom detections in about eight minutes. Write down the five detections you need, then check each against both lists.
Days of storage is a pricing input for Lumana and for Coram, so retention is partly a commercial lever. Lumana does publish a ceiling: real-time cloud backup up to 365 days, with unlimited cloud archiving for clips. What it does not publish is an export path or what happens to footage already on Core if you leave.
If your list includes forklift near-misses, hazard-zone crowding, PPE compliance or SOP adherence, ask which of those ship pre-trained. Lumana publishes PPE detection and proximity monitoring, and frames solutions across physical security, environmental health and safety and operations. Spot AI ships forklift near-miss, falls and hazard-zone crowding as named agents.
Spot AI fits when the cameras are already mounted, the recorder should be additive rather than swapped, and the questions span security, safety and operations. It runs on any ONVIF or RTSP camera plus legacy analog through the Intelligent Video Recorder, keeps full-resolution video in the building, and ships 15+ pre-trained agents including PPE, forklift near-miss, falls and hazard-zone crowding.
Lumana stays a strong call for a site whose recorder is due for replacement anyway, that wants AI on the cameras already mounted, and that values a licensing model with no tiers and every feature included. On camera-agnostic positioning, on publishing what it detects and on publishing what happens next, it is ahead of most of this page, and a comparison pretending otherwise would not survive your first vendor call.
A camera-agnostic platform lets you test the AI on your own cameras before any hardware decision, which is the cheapest way to de-risk the choice.
Watch the AI coworkers work on your live feeds, not a demo reel.
Customer-reported outcomes from named Spot AI customers.
Bridge33 Capital standardized video across 25 plus commercial properties on whatever cameras each acquisition already had, and cut footage search from hours to minutes.
Cambridge City cut the time it takes to find footage from two hours to thirty seconds.
Don Franklin Family of Dealerships lifted camera utilization 60% as the system became an analytical tool across departments rather than a security archive.
"With Spot AI, we were able to standardize video surveillance across existing and newly acquired assets, regardless of what cameras were already installed."
The five covered here are Spot AI, Verkada, Eagle Eye Networks, Coram AI and Rhombus. Spot AI fits teams keeping their cameras and their recorder while adding named agents and deterrence. Verkada suits a greenfield single-vendor build. Eagle Eye Networks suits mixed and partly analog estates. Coram AI suits investigation-first teams, and Rhombus mid-market teams wanting one vendor.
Yes, and it is unusually direct about it. Lumana states it is completely camera-agnostic and was designed to transform any camera into a smart device. It does not publish a protocol or conformance profile on the pages read, so ask for the stream requirements in writing and check them against your oldest cameras rather than your newest.
Lumana answers this two ways on its own site. Its pricing FAQ says no, Core replaces your DVR or NVR, and that Core devices store video data and process footage on site. Its platform section describes Core as an AI engine and video processor that replaces or works with your NVR. Get the one that applies to your site in writing.
Lumana states the license is an annual recurring one set by the number of camera feeds, the days of storage and the license term, that all features and capabilities come standard so there are no tiers, and that it typically covers the Core devices alongside the VMS+ software.
Lumana publishes the list rather than describing it: roughly thirty named event types covering weapon, fire, fall and violence detection, PPE, tailgating, loitering, tampering, plates, faces, occupancy counting and activity heatmaps, under four AI agents called Insight, Monitor, Investigate and Respond. The response layer is published too, down to loud speaker, lockdown, GPIO and emergency dispatch with human verification. What is not published is any route to a detection outside the catalog.