Cost controls
Concurrency caps, cache reuse, runner selection, and per-project budgets — and why owner-edge runs cost no NOME credits.
Owner-edge runs are free of NOME credits
Jobs that run on a machine you own consume no NOME compute credits — you already paid for the hardware. NOME credits are spent only on NOME-managed cloud runners and hosted inference. Routing routine builds to owner-edge runners is the single biggest cost lever.
Warm cache reuse
A warm BuildKit cache means most builds reuse unchanged layers, so you pay for the changed work, not the whole graph. Keeping the cache healthy is both faster and cheaper. See BuildKit cache.
Concurrency caps
You can cap how many jobs run concurrently per project so a busy day doesn't spike spend. Caps apply to NOME cloud runners; owner-edge concurrency is bounded by your own hardware.
Per-project budgets
Set a per-project budget so spend on NOME-managed compute is bounded and visible. Every run shows its cost in the receipt, so you can see where credits go and rebalance jobs to owner-edge runners where it makes sense.
Ready to try it?
Open NOME