Sub-millisecond VM sandboxes for AI agents via copy-on-write forking
curl -X POST https://api.zeroboot.dev/v1/exec \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer zb_demo_hn2026' \
-d '{"code":"import numpy as np; print(np.random.rand(3))"}'| Metric | Zeroboot | E2B | microsandbox | Daytona |
|---|---|---|---|---|
| Spawn latency p50 | 0.79ms | ~150ms | ~200ms | ~27ms |
| Spawn latency p99 | 1.74ms | ~300ms | ~400ms | ~90ms |
| Memory per sandbox | ~265KB | ~128MB | ~50MB | ~50MB |
| Fork + exec (Python) | ~8ms | - | - | - |
| 1000 concurrent forks | 815ms | - | - | - |
Each sandbox is a real KVM virtual machine with hardware-enforced memory isolation.
Firecracker snapshot ──► mmap(MAP_PRIVATE) ──► KVM VM + restored CPU state
(copy-on-write) (~0.8ms)
- Template (one-time): Firecracker boots a VM, pre-loads your runtime, and snapshots memory + CPU state
- Fork (~0.8ms): Creates a new KVM VM, maps snapshot memory as CoW, restores all CPU state
- Isolation: Each fork is a separate KVM VM with hardware-enforced memory isolation
Python — sdk/python
from zeroboot import Sandbox
sb = Sandbox("zb_live_your_key")
result = sb.run("print(1 + 1)")TypeScript — sdk/node
import { Sandbox } from "@zeroboot/sdk";
const result = await new Sandbox("zb_live_your_key").run("console.log(1+1)");Working prototype. The fork primitive, benchmarks, and API are real, but not production-hardened yet. Open an issue if you're interested.