Virtual desktops and private AI, running in your own server room. No per-user licenses, no rent to foreign clouds — and your data stays on your hardware. Built on open technologies, quietly supported by us.
Open technologies, no per-seat licensing. Cancel support any time — everything keeps running.
Files, projects and AI prompts live on your own server and NAS. A stolen laptop is an empty screen.
A fault-tolerant cluster with automatic on-site failover in under 60 seconds — and rollback for the bad days.
After the market consolidation, companies pay per-user licenses and sharply higher annual fees — just for the right to use their own hardware.
Teams want AI for contracts, CAD documentation and code, but won't upload company data to public models. So the benefit waits — for years.
Equipping an office with CAD and office machines is a large outlay that is obsolete in 4–5 years, with data scattered across dozens of drives — a ransomware target.
With subscriptions and DaaS the client never owns the infrastructure, and moving away is expensive and painful.
We build a private cloud on your premises. Every workplace is a virtual machine streamed to inexpensive terminals — in the office or from home. You own everything.
100% open technologies (Proxmox/KVM, ZFS). Support can be cancelled at any time — the system keeps running.
Files, projects and AI prompts live on your own server and NAS. Nothing is sent to us to make the system work. GDPR / NIS2 / EU data residency by design.
Automatic on-site failover in under 60 seconds — continuity individual workstations will never have.
Enterprise server with hardware vGPU partitioning (SR-IOV) and NVMe ZFS storage — dozens of office and CAD workplaces on one machine, with hardware video encoding.
Stateless VDI over a 40–100GbE fabric: the machines are replaceable, the data — isolated and protected. ZFS replication to the backup node every few minutes.
Hardware H.264/HEVC streaming (Sunshine / Moonlight · RDP), multi-monitor — in the office or from home. CAD feels like sitting at a supercomputer.
Main server, backup node and a quorum device form a cluster in your server room. If the main server fails, the backup takes the load automatically. Data never leaves your building for this to work.
For fire, flood or theft you can add an off-site replica — encrypted before it leaves, with keys only you hold — to a location you choose: your second office or our EU cloud. Recovery is started by you, never by us. Without this option, nothing is replicated off-site.
Our public platform is not a separate product. It is how you meet, trial and quietly operate your infrastructure.
Get a GPU workspace and feel the streaming quality in minutes — the platform runs the exact stack we install on-premise.
Every supported client gets a working copy of their infrastructure's shape: the same VLAN per department, the same office, dev, design and machinery workspace types, NAS share structure, routers, switches and department services — all of it empty. Not one file is copied. It exists so your IT team and ours can try a new service for one department or all, change the layout, break things — then roll it out to your own hardware.
The same platform can hold your encrypted off-site replica. On a total site failure, you trigger recovery and your machines come up here within minutes — with keys only you hold.
Most companies want AI in their workflow before they own hardware that can run it. So we split the two: a sealed instance on the platform to teach and tune, and your own premises to run what you've taught.
Your AI instance lives behind an encrypted tunnel, encrypted at rest with keys only you hold. Nobody outside your VPN — and nobody without your keys, us included — can read what's inside.
Feed it your contracts, drawings, procedures. Shape agents around how your departments actually work. Your data is yours the whole time; the workflow you teach is yours too.
When an agent works flawlessly, export the skill to your own premises for in-house use. Clients with their own AI node use the enclave as a test bench: prove a new skill there, then deploy it to the model in their rack.
Every workspace is a real machine with a real GPU partition — not a browser tab, not a shared session — networked together privately and streamed to any device. On the platform they run on our hardware; on-premise, on yours.
Full control — install your own stacks and toolchains on a dedicated machine.
Hardware-accelerated graphics for CAD, 3D and creative work.
Standard productivity tools on the shared private network.
Shared NVMe reachable from every machine in your constellation.
GPU capacity on demand for heavy processing jobs.
Private AI agents inside a sealed, encrypted instance — keys only you hold. Teach them your workflow here, then take the skill home to your own hardware.
Streaming: ~10–15 Mbit per live session · H.264 / H.265 / AV1 · up to 4K @ 60 FPS · Infrastructure in Plovdiv, Bulgaria
ZFS snapshots of every workplace and of the NAS, kept immutable and outside the reach of any user account. If something encrypts a desktop at 10:14, you are back at 10:00 by 10:20.
Desktops, files, printers and the private AI live on your own network. A cut cable outside the building stops the internet — not the office.
Anything that optionally leaves your site — off-site replicas, archived machines, the AI enclave — is encrypted with keys only you hold. Not even we can read it.
The cluster is designed for battery and inverter backup, and its two servers draw a fraction of what a room full of workstations does — so a power cut outside stops the lights, not the work. Our own facility runs the same way: battery-backed, efficiently cooled.
The platform's uptime, incidents and maintenance windows are published for everyone to see, not reported on request. Calm is easier when you can check for yourself.
A one-off project: we deliver the servers, deploy the cluster and migrate your environment.
Service tiers from next-business-day response to 24/7 with proactive monitoring and parts replacement.
A company AI enclave behind your VPN, encrypted with your keys — zero upfront hardware. Teach it, then bring the skill home.
For heavy usage: a dedicated inference server in your own rack, supported like every other critical system.
Public Beta opens at the end of October — in parallel with our first on-premise client deployment, both on the same stack. Request access to try a live GPU workspace, or talk to us about infrastructure in your own server room.