Sandbox cost estimator
Estimate a monthly sandbox compute bill from run count, run length, sandbox size and a provider's published per-hour rates.
How we calculated this
We turn runs times seconds into sandbox-hours, multiply by your vCPU and GiB counts and the per-hour rates you set, then add a flat percentage for idle time, queueing and retries.
Default rates are mid-range published list prices checked on 2026-10-20: E2B and Daytona list $0.0504 per vCPU-hour and $0.0162 per GiB-hour; Fly Sprites $0.0385 per CPU-hour actually used; AWS AgentCore $0.0895 per vCPU-hour; Vercel Sandbox $0.128 per active CPU-hour; Modal $0.1419 per physical core-hour; Cloudflare $0.072 per active vCPU-hour.
Providers meter differently — per second, per 10ms of active CPU, per GB-RAM-second or per request — and plan limits, included usage, storage and egress change the real bill.
Method checked on 2026-10-20.
Estimates only. Metering units, included usage, storage and egress differ per provider; check the linked pricing page before you commit.
Questions
- Are storage and egress in the estimate?
- No. They are metered separately by most providers; add them to the overhead allowance or compute them from the provider's pricing page.
- Why is my real bill lower than the estimate?
- Some providers bill only active CPU or scale a sandbox to zero while it waits on the model. If your runs idle a lot, lower the compute hours or the allowance.
- Where do the default rates come from?
- Each provider's public pricing page; the listing rows in this index link to the cited page for each rate.
Find sandbox providers
Use this tool on your site
The snippet includes a link back to this page. Please keep it.