Usage-based, billed by the second, like a utility. Active vCPU and memory while your machine runs, persistent storage while it exists, and free bandwidth. Idle machines scale to zero and stop billing compute. No seats, no minimums.
Active only. Billed per second while the machine is running.
Active only. Per second while running. Scale to zero pauses it.
Durable storage that follows your VM across migrations.
No bandwidth charges. Egress, ingress, and in-region traffic are all free.
Live migration is included on every machine. You're never billed for the move. Snapshots $0.05/GB-mo.
Drag to size a machine and see the monthly bill. Note how running fewer hours (scale-to-zero) cuts compute cost.
Every plan bills the same usage rates. Higher tiers add support, guarantees, and committed-use discounts.
No monthly fee. Pay per second for what runs. $5 trial credit to start.
For real projects. $20 of usage included each month.
For production fleets that have to stay up.
Dedicated capacity, your cloud or ours.
Same workload everywhere: 1 vCPU + 2 GB RAM, running 24/7. Rounded list prices, for illustration.
| Platform | ~ monthly · 1 vCPU · 2 GB · 730h | Egress | Scale-to-zero | Live migration |
|---|---|---|---|---|
| mashines.dev | ~$29+ storage · cheapest here | Free | Yes | Yes |
| Fly Machines (performance) | ~$31shared CPU is cheaper but oversubscribed | Free tier, then paid | Yes | No |
| Railway | ~$40+ plan base | Metered | No (always-on) | No |
| AWS Fargate | ~$36 | Paid | No | No |
| Google Cloud Run | ~$76always-allocated | Paid | Yes | No |
| E2B / Modal / Daytona | ~$30–100not built for always-on; sandboxes time-box | Varies | Yes | No |
Illustrative list prices (~2025) for a like-for-like always-on instance; verify current numbers with each vendor. We aim to sit just under comparable dedicated-VM pricing, keep egress free, and add live migration none of them offer.
When a machine is idle and you've enabled auto-stop, it suspends and stops accruing vCPU and memory charges. You keep paying only for the persistent volume (storage), which is a fraction of compute cost. The machine resumes from snapshot in under a second on the next request.
No extra charge. Migration is part of the platform, you continue paying your normal per-second compute rate while the VM runs, and the move itself (memory transfer, the millisecond pause) is on us.
Like Railway, you pay for usage rather than fixed instance sizes. Unlike a PaaS that bills an always-on base, mashines.dev can scale compute fully to zero, you pay only for seconds your machine actually runs, plus storage. No per-seat charges.
A 1 vCPU / 1 GB machine running 24/7 is roughly $0.020 + $0.010 per hour ≈ $22/mo of compute, plus storage. Bandwidth is free. Run it only part of the day and the compute cost drops proportionally.
No perpetual free tier. Because every machine is a real microVM holding real CPU, RAM, and disk, free compute attracts abuse (crypto mining, spam, scanning). Instead you get a $5 one-time trial credit with a verified payment method, scale-to-zero so it stretches, and no monthly fee. New accounts have starter quotas and outbound limits that lift once verified.
