Every AI surveillance company has a prerequisite chain. Nine of ten have four to six links. One has one.

A prerequisite chain is the ordered list of conditions a property has to meet before a given AI surveillance company's software can run on it. For most vendors the chain is: working IP cameras on every camera position, each camera exposing RTSP or ONVIF, valid credentials for each camera, a VLAN or subnet the vendor can reach, upstream bandwidth to upload the feeds, and an NVR or bridge device the vendor certifies. If any link in the chain breaks (an analog camera, a lost password, a DVR the vendor does not support), the deployment stalls. Listicles rank vendors on features, which all look roughly the same across the market. The chain is what decides whether the vendor will actually turn on at your property. Cyrano's chain has one link: an HDMI output on the DVR that is already driving your guard monitor. That is the single architectural difference that explains the price and install-time gap across the category.

Look at the vendor's deployment guide, install checklist, or compatibility matrix rather than the marketing page. Four fields reveal the chain. One: what camera protocols are required (RTSP, ONVIF, first-party, HDMI). Two: what credentials are required per camera. Three: what bandwidth is required per camera to the internet or to a cloud region. Four: what existing hardware is kept versus replaced. A vendor that refuses to answer these in writing before a demo is a vendor whose chain is longer than they want to say on a sales call. Cyrano publishes the full chain on the llms.txt spec and in every public install write-up: HDMI input on the device, HDMI passthrough to the monitor, ethernet for the control plane, power. Four cables, one port. That is the whole list.

Because their product is a software-only inference service, and the only way software can read a camera's pixels without owning the camera is a network protocol. RTSP is the de facto streaming protocol for IP cameras. ONVIF is a discovery and control standard layered on top. Both assume a modern IP camera with a valid IP address, a working credential, firewall access, and firmware that has not silently broken the feed. On a mid-market multifamily property installed in 2018, any one of those four can be a blocker. A vendor whose ingest is RTSP-first has a prerequisite chain that is, in practice, about 50 checks long when you include per-camera credential recovery. Cyrano's HDMI tap bypasses the chain entirely by reading pixels off the port the property's guard monitor already uses.

One HDMI cable. Every modern DVR and NVR has a monitor output on the back that renders a composite multiview grid of all cameras. That output has been feeding a guard monitor continuously since the system was installed. Cyrano unplugs the monitor from the DVR, plugs its HDMI input into the DVR, plugs the monitor into its HDMI passthrough port, adds ethernet for the control plane, and plugs in power. Total install time is under two minutes on a running system, with no downtime to the guard view. The composite multiview frame containing up to 25 camera tiles becomes the inference input. No RTSP session is opened, no ONVIF profile is negotiated, no camera credentials are touched, and no cloud ingest is started.

It means a single Cyrano device ingests one HDMI stream that the DVR composites from up to 25 separate camera feeds in a grid pattern (a 5x5 multiview, or any smaller layout the DVR supports). The inference pipeline runs once per frame on the composite image and segments detections back to their tile of origin using fixed tile coordinates. One device covers a 25-camera property on one HDMI port at one flat software subscription fee. This is why the per-camera cost math works out to around $13 a month at 16 cameras and around $8 a month at 25 cameras, regardless of camera brand, protocol, or age.

It means Cyrano covers a narrower set of features on purpose. The detections a mid-market multifamily or construction property actually needs (person in a restricted zone, loitering, package left unattended, tailgating, after-hours entry) are class detections plus zone rules, which run fine on the composite HDMI frame. The detections that require first-party hardware (license plate OCR at 4K per camera, biometric face matching against an enterprise gallery) are out of scope because they are not the job. For the enterprise security director who has the budget and the remit for those features, an enterprise vendor is the right choice. For a property manager who has 16 cameras already on the wall and $200 a month to spend, the one-item prerequisite chain is the only chain that closes.

Listicles (Coram's 11 Best, Memories.ai's 11 Best, MarketsandMarkets, International Security Journal, Kanerika, G2's alternatives lists) rank vendors on feature grids: face recognition yes/no, analytics yes/no, dashboard yes/no. These matter less than they appear because most mid-market buyers will never use features behind the first two columns, and most vendors score similarly on the feature grid. What the listicles do not publish is the prerequisite chain. So a buyer who picks the top-ranked listicle vendor walks into a sales cycle and discovers three weeks later that the vendor's ingest cannot read their 2017 Hikvision DVR, their cameras cannot expose working RTSP, or the property's uplink cannot sustain the 25-feed cloud upload. This page exists so that discovery happens before the sales call, not three weeks after it.

Not automatically. The HDMI tap is the right shape when the property has any combination of analog cameras, legacy DVRs, unknown credentials, mixed brands, or bandwidth constraints. If a property has uniformly modern IP cameras, known credentials, network headroom, and a security director who wants one cloud pane of glass across a portfolio, a cloud-AI vendor (Scylla, Cloudastructure) or a full smart-camera vendor (Verkada, Coram, Lumana) is viable. The selector is the prerequisite chain, not a religious attachment to any specific architecture. Cyrano's pitch is that on most mid-market properties the chain only closes on the HDMI tap, and that is the pattern we have seen across 50+ installs.

The Cyrano llms.txt at apartment-security-cameras.com/llms.txt is the canonical public spec. It lists the hardware price, the software price, the camera count per unit, the edge-AI data path, and the installation procedure. The demo video at youtu.be/_xYiJbH6S_A shows the HDMI tap on a running DVR. The install terminal block on the sibling /t/ai-surveillance-company page lists the exact five-step sequence (unplug monitor, plug Cyrano HDMI_IN, plug monitor into passthrough, plug ethernet, plug power) with timing under two minutes. Every step is on a production running DVR, not a lab mockup. The whole point of the short prerequisite chain is that it is short enough to publish.