Sandbox
Pricing
Per-hour pricing for Tenki Sandbox tiers, vCPU, memory, and storage allocations from Nano to XLarge.
Sandbox is billed per-second across vCPU, memory, and storage. The table below rolls that up into a flat hourly rate for each standard tier, so you can pick a size and reason about cost at a glance.
| Tier | vCPU | RAM | Storage | $ / hour | Best for |
|---|---|---|---|---|---|
| Nano | 1 | 2 GiB | 5 GiB | $0.08 | Quick code-exec calls |
| Small (default) | 2 | 4 GiB | 5 GiB | $0.17 | Standard AI agent sandbox |
| Medium | 4 | 8 GiB | 10 GiB | $0.33 | Light builds, browser automation |
| Large | 8 | 16 GiB | 20 GiB | $0.66 | CI-like builds, parallel test runs |
| XLarge | 16 | 32 GiB | 40 GiB | $1.32 | Power workloads |
How billing works
- Per-second metering. The hourly rate is the equivalent of summing the underlying per-second rates for the tier's vCPU, memory, and storage. You only pay for the time a session is active.
- Storage. The first 5 GiB of storage is included; usage beyond that follows the per-second storage rate.
- Credits. Sandbox usage draws from your workspace's monthly credit balance. See pricing plans for credit inclusions and limits.
- No idle minimums. Tear a session down the moment your agent is done and the meter stops.
Picking a tier
- Nano is the cheapest way to expose a stateless code-exec endpoint to an agent for short, bursty calls.
- Small is the default for typical agent sessions, code generation, light scripting, file I/O.
- Medium and Large suit builds, browser automation, and parallel test runs where you need real CPU.
- XLarge is for compute-heavy workloads that would otherwise need a dedicated VM.
For custom shapes, enterprise contracts, or volume discounts, email [email protected].