What WitCloud is
WitCloud is the on-premise layer of OmnibusCloud — the same distributed-computing engine that runs the public network, installed on your own hardware, inside your own walls.
Connect your machines and they become a private distributed cluster: work is matched to capable machines, distributed by benchmark, run only while a machine is idle, and aggregated back — all under your own policy, with nothing leaving your network. Where the public OmnibusCloud opens the network to the internet, WitCloud keeps it entirely within one organization.
The idle machines you already own
Most office machines are busy for a fraction of the day and idle the rest — overnight, over weekends, between tasks. That idle time is real compute you've already bought and powered.
WitCloud puts it to work on your own jobs, turning the machines across your offices into a private cluster that runs while no one is using them — without buying dedicated infrastructure and without sending anything to a public cloud.
Built for your workloads
WitCloud runs the kinds of work that spread across many machines — simulation, data processing, batch jobs, analysis, model training, encoding, and whatever else your organization runs at scale. Anything that can be split into pieces, distributed across the fleet, and checked afterward runs on it.
Each kind of work runs as a controller — a component that defines how that work is prepared, split, executed and validated. New workloads are added by building a controller for them: on the open SDK by your own team, or delivered as a controller adapted to your systems and your data. The platform stays the same; the work it runs is whatever you need it to.
Built on the same engine
WitCloud inherits everything the engine does, and a private fleet exercises all of it:
- Heterogeneous by capability. Offices don't have identical machines. WitCloud benchmarks each one and assigns work proportionally — a developer's workstation takes more, an office PC less, and both contribute. There are no manual tiers to configure.
- Invisible to the people using the machines. The agent runs only when a machine is idle, and yields the moment someone returns. The person at the keyboard doesn't notice it.
- Fault-tolerant. If a machine goes offline mid-task — someone shuts a laptop, a desktop reboots — its work is reassigned automatically.
Where WitCloud sits
WitCloud is one layer of a single platform, not a separate product. The same core runs in a single office, across a global organization, or on the open internet — what changes is not the engine, but how widely it reaches.
- WitEngine — the execution core: units of work, capability-based distribution, retries and fault tolerance, and the controllers that carry workload logic.
- WitCloud — the deployable server: the layer you install on your own hardware, adding client registration, controller distribution, health monitoring, scheduling and result aggregation.
- OmnibusCloud — the same server opened to the internet as a public network.
Deployment modes
The same stack runs at three scopes, and an organization can move along them as trust grows:
- On-premise — one office, one network, full IT control. The simplest trust model: nothing leaves the organization.
- Cross-organization — offices in different time zones contributing to one private network, so compute follows the sun across a company: when one site sleeps, another is just starting its day.
- Public network — the open OmnibusCloud, where anyone can join.
Start behind your own firewall, and scale outward as trust grows.
Your hardware, your data
On-premise means the data boundary is yours. Jobs run on your machines, results stay in your network, and you control access and policy — with the same distribution and validation the public network uses. Work that can't leave a controlled environment runs where it already lives.
Get started
- Architecture and docs: omnibuscloud.io — how the platform is built and how to deploy it.
- Controllers and SDK: github.com/OmnibusCloud — the open controllers and the SDK for building your own.
- A deployment, or a controller built for your workload: get in touch.
Let's build the ultracomputer together!