Pricing
Alien provisions Sandboxes in your customer's cloud. The customer pays the cloud provider for the underlying compute, and Alien charges a small management fee. See Pricing for Alien's current rates.
AWS prices are for US East (N. Virginia) on Graviton, Azure prices for East US. Google's Agent Platform pricing does not name a region. Last verified: 2026-09-26.
Compute
| Provider | Service | vCPU | Memory | Billed |
|---|---|---|---|---|
| AWS | Lambda MicroVMs | $0.0000276944 per vCPU-second | $0.0000036667 per GB-second | Per second while running. Baseline resources for the whole run; resources above baseline only for the seconds they are used. |
| GCP | Agent Platform sandboxes | $0.085 per vCPU-hour (Agent Compute) | $0.009 per GiB-hour (Agent Memory) | On allocated vCPU and memory, rounded to the nearest second. |
| Azure | Container Apps Sandboxes | $0.000024 per vCPU-second | $0.000003 per GiB-second | Per second, at Consumption plan rates. Microsoft charges no CPU or memory fees while a sandbox is stopped. |
On AWS, what you pay for depends on the size Alien picks from your declared limits (see AWS Sizes). Declaring cpu: "1", memory: "2Gi", and disk: "8Gi" runs a MicroVM with a 512 MiB baseline that bursts to 2 GiB, and you pay the baseline for the whole run. A sandbox with no .limits runs a 2 GB, 1 vCPU baseline that bursts to 8 GB and 4 vCPU.
AWS Snapshots
Lambda MicroVMs start from a snapshot image and keep a suspended MicroVM's state as a snapshot.
| Charge | Price |
|---|---|
| Snapshot storage (images and suspended MicroVMs) | $0.08 per GB-month |
| Snapshot data written (on suspend) | $0.0038 per GB |
| Snapshot data read (on start and resume) | $0.00155 per GB |
MicroVM image storage has a 1-week minimum retention. Data transfer is billed at standard AWS rates.
Google and Microsoft do not publish a separate charge for a paused sandbox's state on these pages. On GCP and Azure, the cost drivers are the vCPU and memory a sandbox holds while it runs, and how long it runs.
Kubernetes and Local
On Kubernetes, the customer pays for the cluster. Each Sandbox resource keeps a warm pool of 2 idle pods. Their declared limits also become their resource requests, so the idle pods reserve that cpu and memory on the nodes while they wait. On Local, sandboxes run on your machine.
Example: 10,000 Runs of 5 Minutes per Month
Each run holds 1 vCPU and 2 GiB of memory for 300 seconds, with no bursting and no pause. On AWS, that is a sandbox declared with no .limits, and each run launches from a 2 GB image.
| Provider | vCPU | Memory | Snapshots | Total |
|---|---|---|---|---|
| AWS | $83.08 | $22.00 | $31.16 | $136.24 |
| GCP | $66.58 | $14.10 | none | $80.68 |
| Azure | $72.00 | $18.00 | none | $90.00 |
The GCP figures apply its free usage tier (below). Without it, GCP is $85.83. The AWS snapshot column is 10,000 launches reading 2 GB each ($31.00) plus one month of storage for the 2 GB image ($0.16).
Free Tiers
- AWS Lambda MicroVMs: none listed.
- GCP Agent Platform: the first 50 Agent Compute vCPU-hours and the first 100 Agent Memory GiB-hours each month, per account.
- Azure Container Apps: the Consumption plan lists a monthly free grant of 180,000 vCPU-seconds and 360,000 GiB-seconds. The pricing page does not say whether sandbox usage draws on it.
Sources
- AWS Lambda Pricing: MicroVM compute rates, snapshot charges, image retention
- Vertex AI Pricing: Agent Compute and Agent Memory rates for sandbox environments
- Azure Container Apps Pricing: Sandboxes follow Consumption plan per-second pricing
- Azure Retail Prices API, Container Apps in East US: Consumption plan vCPU and memory rates
- Azure Container Apps Sandboxes overview: no CPU or memory fees while stopped