Rhombus gives outside AI agents a map of the cameras before they answer anything
A July 2026 post by Rhombus's Jeremy Gulley describes a command-line tool that exports a live snapshot of every site and device, and a Claude plugin that loads it before each query. The premise is that an agent that does not know where the cameras are cannot be trusted with much.
Most of the agent announcements in this industry are about the model. Jeremy Gulley's post for Rhombus, published July 8, 2026 on the company's blog, is about something less glamorous and arguably more limiting: the agent has no idea what a customer's deployment looks like.
Gulley, Rhombus's global director of ecosystem, starts from how camera inventories are actually kept: a spreadsheet that was accurate on install day, or a floor plan printed and pinned to a wall. Neither gets updated when a camera is moved or a site is added, and every automation project that follows pays what he calls a silent tax to reconstruct that information by hand.
The snapshot
The fix Rhombus shipped is a command in its rhombus-cli tool, "rhombus context generate," which pulls a current snapshot of every location, every device and representative imagery from each camera directly out of the platform. The data comes straight from the system of record, so it is correct at the moment it is generated, and it can be regenerated whenever it is needed. Nobody has to maintain it.
The second piece is a Rhombus plugin for Claude that loads the snapshot before each query. With that context in place, an external agent can answer a question like which cameras are in the east parking lot without a person looking it up, and can then go on to do whatever the operator asked for with those cameras. Gulley's argument is that the whole chain depends on that first step. "Your AI agent is only as good as the context it has," he writes, and the usual alternative, in his words, is "a document nobody opens." Generate the context from the source system and everything downstream gets faster and more accurate.
Why it differs from a video agent
What Rhombus built here is the map. The agent that reads it is whatever the customer already runs. The company's own in-product AI work is search, launched in April 2025 and extended since, and the July post concerns the plumbing that lets a customer's own tools, including general-purpose assistants, ask the VMS questions. Most other vendors are building a proprietary agent and hoping customers use it. Rhombus is betting the customer already has one.
For integrators the implication is that the deployment record, the thing that used to live in a binder, becomes a live export and a product feature. For customers experimenting with agents on top of their security stack, it removes the step where someone spends a week typing camera names into a config file. That is a small piece of work, and I suspect it is the piece most agent projects in this industry are stuck on right now.
Read the original at Rhombus.