AgentSandbox

Give your agent a disposable code sandbox

Disposable code sandboxes agents can spin up in seconds.

Read the launch post →
naive uplive
  • Postgres16.2 · iad1100%
  • Auth12.4k users100%
  • Edge functions18 deployed100%
  • Storageprovisioning72%
  • Cronqueued54%
3 / 5 liveprovisioned in 1 call

Paste and ship

Drop this into Cursor, Claude Code, or any agent. It reads the skill manifest and wires up Sandbox for you.

Agent prompt

Read https://usenaive.ai/skill.md and use the Naïve Sandbox primitive in my project. Start with: naive sandbox create --name scratch --size s

What you get

Boots in secondsMicro-VMs on demand.
Exec, files & portsShell, file I/O, public URLs.
Checkpoint & forkFull machine state, cloneable.
Usage-billedPay for compute you use; parked is storage.

One call to AgentSandbox

CLI, SDK, or REST — one bearer token, one credit balance.

POST /v1/sandboxes

Sandbox

Isolated Linux micro-VMs for running untrusted code — exec shell commands, read and write files, expose ports at public URLs, checkpoint the whole machine, and fork copies of it. Billed only for the CPU, memory, and disk your sandbox actually uses.

Boots in seconds
Micro-VMs on demand.
Exec, files & ports
Shell, file I/O, public URLs.
Checkpoint & fork
Full machine state, cloneable.
Usage-billed
Pay for compute you use; parked is storage.
# Run with your API key exported
$ naive sandbox create --name scratch --size s

Start building with AgentSandbox

One API key. One credit balance. Every primitive.

Related primitives