Arcturus is an AI-driven control platform for ventilation across multi-site building portfolios. It runs on top of whatever building management system (BMS) is already on site, sets up a hardware integration where there isn’t one, and pulls indoor air-quality data from supported wireless sensor providers in both cases. The platform reads the documents and live data from each site to identify what is there, proposes a control configuration, and runs the per-site control loop once that configuration has been reviewed and deployed.

These pages are written for the building operators and HVAC integrators who set up and run sites through the in-app interface. The technical-coordination parts that involve a BMS vendor — credentials, watchdog setup, the round-trip on the address-list CSV — sit under the Building management system category.

What the platform does today

The path from an empty site to a controlled one runs through the new-site wizard — the eight-step guided flow under Site setup in this docs library, and the same flow that opens inside the platform when a new site is started. After a BMS or sensor-only setup is connected and the address-list CSV (plus any technical documentation) is uploaded, the platform proposes a control configuration. Once that configuration is reviewed and deployed, the per-site control loop runs continuously: reading ventilation state, applying adaptive control, writing setpoints back through a watchdog safety mechanism, and reporting telemetry into the dashboard.

1. Building management systems supported today: IWMAC (in production). The Niagara N4 over MQTT integration is wired through the wizard but not yet tested end-to-end against a live customer site.
2. Wireless sensor providers supported today: Airthings and Disruptive Technologies. Additional providers are added on request — Arcturus is supplier-agnostic and stays that way.
3. Sites without a BMS get a hardware integration set up on site — either control hardware we supply or external hardware that fits the building — so the per-site control loop has what it needs to read from and write to the building.
4. The wizard’s eight steps cover credentials, document upload, AI-driven detection of systems and components, manual review, watchdog setup (where the BMS supports write-back), and deploy.
5. Per-site adaptive control: occupancy-aware ventilation, free-cooling optimisation, mode switching between heating and cooling, with a watchdog that holds last-known-good values if the platform becomes unreachable.
6. Hosting: European Kubernetes cluster (Amsterdam region). Operational telemetry and configuration stay in-region.

Where the platform is going

Several pieces of the trajectory are live, several are still being built. The direction:

1. Auto-generated configuration: the wizard already drafts most of the site config; the goal is end-to-end auto-config — manual review only when something looks off.
2. Drawing comprehension: a separate computer-vision pipeline reads system schematics and P&ID drawings to extract components and connections. Working in development; not yet in production use.
3. More BMS vendors: integrations beyond IWMAC and Niagara are added on demand. Sites running something else can find the route under "What if your BMS isn’t supported yet?" in the BMS category.
4. Self-critiquing agents: each AI step in the pipeline is being wired to evaluate its own output and either confirm it or revise it. Rolling out one agent at a time.
5. Operational analysis across providers: today’s operational snapshot in the wizard is IWMAC-only; making it provider-generic is a next step.
6. Energy-system visualisation: a richer view of how heat, cooling, electricity and ventilation flow through the site, so operators can read the whole energy picture rather than only the ventilation portion.
7. Broader scope across the site: control coverage extends from ventilation today into heating, cooling, and lighting next.
8. Multi-system coordination: integration with battery storage, on-site solar production, and EV fast-charging so HVAC behaviour can be coordinated with the rest of the site’s energy use, instead of each subsystem optimising in isolation.
9. Automatic fault detection with root cause: anomalies surfaced from live data and traced to the underlying fault, not the symptom — the operator should read a cause, not infer one from a dashboard.

What Arcturus is not

Some boundaries the rest of the docs do not always restate explicitly:

1. Arcturus is not a replacement BMS. The platform reads from and writes to whatever building automation is already on site.
2. Arcturus is not a primary safety system. The watchdog protects against bad writes from the platform; the BMS itself is what keeps the building safe.
3. Arcturus is not zero-touch. The wizard guides the setup and the AI drafts the configuration, but the operator makes the call at the review steps — particularly Step 6, where the wrong pick can quietly bake a wrong control decision into the site.
4. Arcturus does not extend its own access. Only parameters the BMS vendor has exposed and tagged writable are reachable. See Parameter access under IT info & safety.

What these docs contain

The library is organised by the part of a setup the reader is working on:

1. Building management system — identifying what BMS the site runs, connecting it through the wizard, and what to do when the vendor is not yet supported.
2. Indoor sensors — when sensors are needed, which providers are supported, and how each provider’s setup translates into the wizard.
3. Control hardware — what to expect when there is no existing BMS and on-site control hardware needs to be set up to bring the site online.
4. Site setup — the new-site wizard walkthrough (the eight-step flow inside the platform that takes a site from empty to running) and the post-go-live checks.
5. IT info & safety — what data Arcturus reads and writes, network and firewall requirements, the watchdog and failsafe mechanism, credential handling, and access boundaries.

For a path tailored to a specific site, the Start here flow on the docs landing recommends the right sequence of guides based on three or four questions about the setup.