What's inside.
The inventory, as your team will see it.
| Host | Hardware | Owner | Utilisation | Cost / h |
|---|---|---|---|---|
| Busy: gpu-a-01 | 8 × H100 | Owned | 94% | — |
| Busy: gpu-a-02 | 8 × H100 | Rented | 88% | $26.40 |
| Partly used: gpu-a-03 | 8 × H100 | Borrowed | 41% | — |
| Idle: gpu-b-07 | 4 × A100 | Rented | 0% | $9.80 |
| Idle: gpu-b-08 | 4 × A100 | Rented | 0% | $9.80 |
| Busy: cpu-c-12 | 96 vCPU | Owned | 67% | — |
Two rented A100 boxes have been idle for nine days. That's what Clastra is for.
Where it reads from
AWS, GCP, Azure, Lambda, CoreWeave, bare metal over SSH.
Where it goes
A web view, a CLI, and a JSON API your scheduler can ask.
What it touches
Nothing. Read-only, unless you decide otherwise.
See it on your fleet.
Request access
Private beta.