Compute you control.
ANAROQ is building one control layer over compute capacity: machines for you and for the AI agents you allow, inside limits you set.
- In development
- Public tools first
- No sign-up needed
One control layer over compute.
ANAROQ is not another VPS host. The product is the control layer: you ask for compute, set its limits and see what it costs, in one place. The capacity underneath can come from external providers, rented servers and, later, hardware of our own. You don't manage that complexity.
- Persistent compute
- Machines that run for as long as you need them.
- Temporary compute for agents
- Short-lived environments that agents create and destroy, inside limits you set.
- Capacity you never manage
- Where a workload runs is our job. You see one platform.
ANAROQ is in development. Nothing on this page can be ordered yet; the public tools are the first step.
Limits, not keys.
Giving people and AI agents access to infrastructure should come with real boundaries. These are the controls the platform is designed around.
- Budget caps
- Enforced before money is spent, not discovered on an invoice.
- Time to live
- Resources expire on schedule, by policy.
- Concurrency limits
- A ceiling on how much an agent can run at once.
- Approvals
- Sensitive or costly actions wait for a person.
- Cost estimate first
- See the cost and the policy result before anything runs.
- Audit trail
- Who did what, from which interface, and when.
- Kill switch
- Revoke agent access and stop its activity at once.
- Scoped permissions
- Each token or agent gets only what it needs.
Design principles of a platform in development, not features available today.
One platform. Every surface.
The console, the API, the CLI and MCP are planned as four ways to reach the same control plane, with the same policy and the same audit trail. MCP is not a separate product; it is one more way to do the same thing.
- UI
A console for people.
- API
The single interface everything else uses.
- CLI
The same actions from a terminal.
- MCP
The same actions for AI agents, under the same policy.
Start with the free tools.
Measure, understand, then deploy. The first tools help you size and set up self-hosted apps, with every number traced to an official source.
- Choose computeStack SizerPick self-hosted apps, a load preset and options; get vCPU, RAM and disk with a line-by-line breakdown and sources.In preparationAbout the Stack Sizer
- Set up an appSelf-hosted appsComponents, load presets, options and official sources for popular self-hosted apps, starting with eight.Profiles in preparationBrowse the apps
- Check infrastructureInfrastructure checksChecks for DNS, TLS and network reachability of a server you run.Coming soon
Self-hosted apps we're profiling.
A shared, versioned library of resource profiles. Each profile is published only when its numbers are traced to official sources.
Measure first. Then decide.
Our first public tools are for sizing and setting up self-hosted apps. The rest of the platform follows, one verified step at a time.