Home / Hosting
Private AI hosting, so you get your own AI desktop without running the hardware
If you would rather not operate servers, we host your workspace and its AI desktop on hardware we operate. AI requests go to an open-weight model: the untuned 4B on our hardware, or a larger one through OpenRouter. A deployment used by you alone is planned and not yet delivered.
What it is
Hosting means we run the model for you on hardware we operate, and give each client its own hosted desktop where the agent works, driven from a web dashboard. Nothing is single-tenant today: desktops are separate containers on one shared machine. AI requests go to an open-weight model: the untuned 4B model on our hardware, or a larger open-weight model through OpenRouter. Requests sent through OpenRouter leave our hardware and go to OpenRouter and the provider it routes to. In the web app, chat and tasks in a new workspace start on a larger open-weight model through OpenRouter. They run on the untuned 4B model on our hardware instead when that route is not available, when the wallet holds no real money or when the plan’s monthly model cap is reached. The AI desktop agent runs on the 4B model on our hardware, or on your own model once one is delivered. A workspace that turns training on moves to its own model once it is trained. That model runs on our hardware. Training is opt-in: it happens only if an owner of your workspace turns it on, on our own hardware, and the model is used for your workspace only.
A model trained for you that you can export is planned and has not been delivered yet. A separate machine for each client is not offered; moving the platform to a dedicated server is planned, with no date yet.
Who it is for
Teams that want the privacy and cost control of private AI but have no one to run GPU servers, and companies that want to start hosted and move in-house later.
What you get
- A hosted model. Your requests go to an open-weight model: the untuned 4B model on our hardware, where a request does not leave it, or a larger open-weight model through OpenRouter, where it does.
- Isolation. Each client's desktop is a separate container with its own storage, disk limit and network rules, on one shared machine. A separate machine for each client is not offered. A private API endpoint per client is planned and has not been delivered.
- Monitoring and upkeep. Monitored by the founder. There is no uptime guarantee yet.
- An exit path (planned). When per-client models launch, yours will be exportable, and we will help you move it to your own servers.
How it is done
- Agree the terms. Expected load, availability needs, data handling and retention are written down before anything is deployed.
- Provision. Your workspace and AI desktop are created when you register, and you can try your own examples on a practice company.
- Connect. We connect the accounts you approve and set the access rules. A private API endpoint for your own applications is planned.
- Operate. We run it and report on usage.
What we need from you
- An expected volume of requests, even a rough one.
- Your rules on data location and retention.
- The list of people and accounts allowed to use it.
- A contact for planned maintenance and incidents.
Questions, answered.
Is hosting really single-tenant?
Not yet. Today every client's desktop is a separate container on one machine that we operate, and AI requests go to an open-weight model: the untuned 4B model on our hardware, or a larger open-weight model through OpenRouter. The AI desktop agent itself runs on the 4B model on our hardware. A separate machine for each client is not offered. Moving the platform to a dedicated server is planned, with no date yet.
Can we move to our own servers later?
That is the plan. Per-client models, and handing over their weights, are in development, so today there is no per-client model to move. We can scope an installation on your own servers, but we have not delivered one yet.
Do you use our prompts for anything?
Requests are used to answer you. Requests sent through OpenRouter are handled under the terms of OpenRouter and the provider it routes to. We train on your data only if an owner of your workspace turns training on, and that model is used for your workspace only; it is never used to train a model for another client. Retention is described in our privacy policy.
Other services
- DistillationTrain a compact model for one task of yours.
- InstallationOn your own servers. Scoped per client, not yet delivered.
- Agentic systemsAgents for multi-step work, with the approvals you set.
Start on a practice company.
Request an invite and try it on made-up customers first. Nothing reaches real people while you learn.