Spaces joins the building, equipment and operations systems a company already runs, plus camera-free RuView sensing, into one place model and one ordered event log, with evidence and approval boundaries for every person and agent that acts on it.
SALT LAKE CITY, UT / / November 11, 2026 / At 03:00, a cooling unit fails in a data hall. The building management system, the power distribution units and the rack temperature probes each raise their own alarm, each with its own tag name. None of them says where the fault is, which racks sit downstream or how long before those racks overheat. An engineer has to piece that together by hand, across three consoles, while the clock runs. Cognitum One™, Inc. (cognitum.one) today launched Cognitum Spaces™ at KubeCon + CloudNativeCon North America in Salt Lake City to close that gap. Because every system writes to the same zone in Spaces, the failure arrives as one alert on one place: cooling degraded in Pod 3, cold aisle A, 14 racks downstream, about nine minutes to 32 °C. In Cognitum One’s simulated data-hall scenario, the time to locate the fault and everything it affects falls from about 11 minutes to 20 seconds.
That is the problem Spaces exists to solve. Enterprises already run capable systems for buildings, equipment, bookings, maintenance and fleets, but each one holds a slice of a site and none of them holds the place itself. Spaces gives every site one place model and one ordered, replayable event log, written by the systems a company already runs. Where no existing system can see what is happening in a room, RuView™, Cognitum One’s open-source camera-free sensing project, adds presence, movement and occupancy from ordinary Wi-Fi signals, with the raw radio data kept at the edge.
The timing matters because enterprises are now putting AI agents to work on physical operations. An agent asked to rebalance cooling, release a hotel room or reroute a forklift needs more than a feed of sensor values. It needs to know where things are, what depends on them, whether the data is current, and what it is allowed to do. Spaces gives agents that context, the evidence behind every claim, and approval boundaries that keep people in charge of what happens next.
“Every enterprise wants agents that can act in the physical world, and almost none of them can tell an agent reliably what is happening in a building right now,” said Reuven “rUv” Cohen, Founder and Chief AI Officer of Cognitum One and creator of RuView. “The data is all there, scattered across a BMS, a ticketing system, a booking engine and a dozen probes. Spaces puts it in one place, in order, with the evidence attached, so an agent can reason about a site the way an experienced operator does, and can be stopped before it acts.”
Explore the Spaces demos: https://spaces.demo.cognitum.one
Explore RuView: https://cognitum.one/ruview
Developer resources and MCP tools: https://cognitum.one/mcp
What Spaces Is the System of Record For
Spaces does not replace the systems a company already trusts. It owns the records that none of them keep: where things are, what happened there, in what order, and who or what was allowed to act.
- The place model. The authoritative structure of each site, from campus to building, floor, hall, room and zone, and which equipment, rooms, lanes or blocks belong to each. Every other system’s data is tagged to a place in this model.
- The site event log. One ordered, replayable record of every state change written to a site, with the time, the source system and its provenance. Teams can scrub the log backward or forward and see the site exactly as it stood at any moment.
- Spatial facts and alerts. The conclusions that only exist once sources are joined, such as “cooling degraded in cold aisle A, 14 racks downstream” or “room 412 ready to sell,” each with a confidence score, its evidence and a status of live, delayed, stale or unavailable.
- The governance record. Tenant-scoped policies, automations, approval boundaries, agent proposals and the decisions made on them, so every action taken on a site can be traced back to who approved it and why.
Source systems stay authoritative for their own records. The building management system still owns setpoints, the booking engine owns reservations, the maintenance system owns work orders and the power system owns its readings. Spaces references those records and never overwrites them. When a person or an approved agent acts, the action goes back through the system that owns it.
Why Companies Add Spaces to the Stack They Already Run
Spaces lets teams understand and coordinate across systems that today stay disconnected.
- Equipment state tied to location and consequence. A cooling, power or probe alarm arrives already knowing which zone it is in and what sits downstream, so engineers start with the blast radius rather than a tag name.
- One answer to “is it ready?” Bookings, housekeeping, door locks and maintenance tickets write to the same room, so “ready” becomes one fact with a time and a source, rather than four systems that each hold part of the answer.
- Operations that know what they block. At a port, crane, fleet, gate and yard-plan systems write to the same lane, so a vehicle fault already knows which crane and which berth it is holding up.
- An agreed timeline after an incident. Record systems with clocks that disagree become one ordered ledger, so operators, insurers and regulators can replay what each system knew, and when, instead of contesting timelines for weeks.
- Occupancy where no system can see it. RuView sensing adds camera-free presence and movement in freezer aisles, wards and other spaces where cameras are unwelcome or impractical. Only the semantic event, such as “movement in zone,” leaves the room.
Built for Enterprise AI
- Reliable physical-world context. Agents query a Space for a site’s structure, current state and dependencies through a tenant-scoped API (GET /v1/spaces) or MCP tools, rather than stitching together raw feeds from many systems.
- Honest state, never filled in. Every value is labelled live, delayed, stale or unavailable. When a site goes quiet, Spaces says so, and never fills an outage with sample values. An agent can see that a reading is stale and decline to act on it.
- Evidence and provenance on every claim. Each fact carries a confidence score, the evidence behind it and its provenance, so an operator, an auditor or a regulator can ask why an agent did what it did.
- Approval boundaries. Agents inspect a Space and propose actions; policies scoped to the tenant decide what runs automatically and what needs a person to approve it, and every decision is recorded.
- Part of the Cognitum One platform. Spaces works alongside RuVector™ for agent memory and Ruflo™ for agent coordination, giving teams of agents shared, governed access to physical layouts, occupancy and conditions.
Privacy and Edge Governance
- The gateway stays authoritative. Where RuView sensing is deployed, the HomeCore gateway remains the source of truth on site. The Spaces dashboard renders only permitted state, and a user’s browser never connects to a private-network device or publishes directly to MQTT.
- Raw signal stays at the edge. Raw Wi-Fi Channel State Information, recordings, pose skeletons and vital-sign waveforms remain on the gateway unless separately authorized for release. Spaces publishes conclusions, not the underlying radio. Remote and multi-agent access uses bounded read clients under Cognitum One’s ADR-402 governance protocol.
In Practice
Cognitum One has also published eight interactive scenarios that show Spaces at work across different industries. Five connect existing operational systems with no sensing at all: a data hall, a hotel, a vineyard, a container port and a tank farm. Three add camera-free RuView sensing: flood response, a freezer warehouse and a hospital ward at night. Each scenario lets viewers switch Spaces on and off and see the live API call behind every change. In the hotel scenario, for example, the gap between a room being cleaned and being ready to sell falls from about 45 minutes to under one minute. All eight are scripted simulations of the target design, not captured measurements.
“Our customers’ AI programs keep running into the same wall: the agent is capable, but it doesn’t know what’s true on the ground and nobody can audit what it relied on,” said D. Christopher Keil, Founding CEO of Cognitum One. “Spaces gives infrastructure and operations teams a governed record of their physical sites that sits alongside the systems they already trust. It’s the same discipline the cloud-native community brought to workloads, now applied to the physical world.”
Availability
Cognitum Spaces is available today in the Cognitum console. RuView remains open source on GitHub, with more than 103,000 stars and over 13,400 forks. It is part of the ruvnet open-source ecosystem, whose 410 npm packages and 498 Rust crates have recorded more than 116 million downloads as of November 10, 2026. Spaces is delivered as part of the Cognitum One platform and can be deployed alongside Ruflo, RuVector and ruOS™ on infrastructure the customer controls. Hardware, model and infrastructure costs still apply.
Cognitum One recommends starting with one site and the two or three systems whose alarms or records most often need to be joined by hand, measuring the time it takes to reach an agreed answer before and after, and then adding sites, sensing and governed agents as each one earns its place. Attendees at KubeCon + CloudNativeCon North America who would like to meet with Cognitum One can email hello@cognitum.one. Developer resources and MCP tools are at cognitum.one/mcp, and the RuView project is at github.com/ruvnet/RuView.
About Cognitum One
Cognitum One is an artificial intelligence company building a sovereign agentic platform that orchestrates AI across the entire compute plane, from the public cloud to on-premises systems to the low-power edge, placing every workload where performance, cost, and security require. The platform is private by default, with real-time self-learning memory, distributed swarm coordination, contactless perception, cryptographic attestation, and native MCP integration, plus SDKs for Rust, Node.js, and Python. It is built for enterprises operating under data sovereignty and regulatory constraints. The company is led by Founder and Chief AI Officer Reuven Cohen, creator of the widely used open-source AI projects RuView, Ruflo, and RuVector, and Founding CEO D. Christopher Keil. To learn more, visit cognitum.one.
KubeCon and CloudNativeCon are trademarks of The Linux Foundation. Cognitum One is not affiliated with or endorsed by The Linux Foundation. Scenario figures, including the data-hall and hotel examples, are from Cognitum One’s scripted simulations, are illustrative of the target design and are not captured measurements. Results in real deployments depend on the connected systems, data quality, site layout and configuration. Cognitum Spaces is not a medical, life-safety or safety system. Star and fork figures are from the public ruvnet/RuView GitHub repository as of November 10, 2026. The download figure is the combined total across ruvnet’s npm packages since October 2025 and its Rust crates since publication, as of November 10, 2026.
Media Contact
Nick Ruest
Chief of Staff
Cognitum One, Inc.
(416) 477-3336
SOURCE: Cognitum One
