Pricing
Free during the private beta. To get your installation enabled, write to hello@hypervise.sh. The prices below apply once the beta ends. We will give at least 30 days' notice before billing starts.
One price: $0.04 per core-minute
You pay for the physical cores your job's machines hold, at $0.04 per core-minute ($2.40 per core-hour). Usage is metered per second, with a one-minute minimum per job. Recorded and unrecorded time cost the same.
| Your job is… | It is billed for |
|---|---|
| recorded (the default) | 1 core |
unrecorded, after hypervise update-config --physical-cpus K --no-recording |
K cores |
| split into N shards | each shard's cores, plus 1 core for the coordinator while it waits for its shards |
Included at no extra charge:
- Recordings of failed jobs, kept for 30 days.
- The replay check that confirms each failing recording reproduces.
- Job summaries and logs on the dashboard.
- Network traffic.
Recordings of passing jobs and shards are deleted when the job ends, and are not billed.
Replay sessions (re-running a kept recording to inspect it) are billed like a job: per core-minute, at the same price.
Example
A Rust project that builds unrecorded on 4 cores, then runs its tests as 4 recorded shards (the getting started workflow):
| Phase | Machines | Time | Core-minutes |
|---|---|---|---|
| checkout and build, unrecorded on 4 cores | 1 × 4 cores | 3 min | 12 |
| tests, 4 recorded shards + coordinator | 4 × 1 core + 1 | 2 min | 10 |
| total | 5 min | 22 core-minutes = $0.88 |
What we don't charge for
- Time a job waits in the queue.
- Machine start-up from the pre-booted image.
- Jobs that fail because of us: a host fault, a quarantined host, or a lost runner.