# Tenki Sandbox — Disposable Linux VMs for AI Agents (https://tenki.cloud/products/sandbox)

> For the complete documentation index, see [llms.txt](https://tenki.cloud/llms.txt)

Tenki Sandbox gives AI agents isolated, fully managed compute environments — full Linux VMs that boot in under 2 seconds. Give your agent root without giving it yours.

[Start for Free](https://app.tenki.cloud/auth/registration) · [View Docs](https://tenki.cloud/docs/sandbox/quickstart.md)

Companies deploy faster with Tenki.

## Sandbox SDK — Drive sandboxes from your code.

TypeScript, Go, and Python SDKs over the same VMs. Authenticate with an API key and create sessions from your application.

- TypeScript: `npm install @tenkicloud/sandbox`
- Go: `go get github.com/LuxorLabs/tenki-sdk-go/sandbox`
- Python: `pip install tenki`

Example (Python):

```python
from tenki import Sandbox

with Sandbox.create(cpu_cores=2, memory_mb=4096) as sb:
    result = sb.exec("bash", "-lc", "uname -a && whoami")
    print(result.stdout_text)
```

## Capabilities — Full Linux VMs that boot in under 2s, with isolation no container can touch.

Each session is a disposable environment where agents write, run, and ship code. [Read the docs.](https://tenki.cloud/docs/sandbox/quickstart.md)

- **Isolated Environments** — Every session runs in an isolated VM with zero cross-contamination between workloads. Your code, dependencies, and secrets never leak across boundaries.
- **Bring your own agent** — Works with Claude Code and Codex; more harnesses coming soon. The sandbox is the workspace, your agent is the driver.
- **Massive Parallelization** — Spin up hundreds of sandboxes concurrently for batch workloads, evals, or multi-agent orchestration, each one fully isolated with dedicated resources.
- **Per-task sandboxes** — Every agent task runs in its own ephemeral sandbox, with file system, network, and secrets fully isolated from your host machine.
- **Vibe Coding** — Build it, run it, share it. Expose ports to ship live apps instantly. Any language, any framework.
- **Instant Provisioning** — Go from API call to running sandbox in under 2s, no cold starts, no waiting on image pulls. Agents get a live environment before the next token is generated.

## Workflow — Keep sandboxes awake, pay only when they work.

Most sandboxes cold-start on every task and bill you for idle time. Tenki keeps yours warm and ready, with no compute charge while it sits unused.

| Dimension  | Other Sandboxes                                            | Tenki Sandbox                                         |
| ---------- | ---------------------------------------------------------- | ----------------------------------------------------- |
| Startup    | Cold start on every task, with seconds of lag.             | Kept awake and ready when your agent needs it.        |
| While Idle | Billed for idle compute, or torn down to save cost.        | No compute charge when a sandbox is not in use.       |
| State      | Filesystem and context are lost once the box is destroyed. | Filesystem and memory saved. Work resumes seamlessly. |
| Scaling    | Throughput is capped by how fast new boxes spin up.        | Fan out hundreds of warm sandboxes in parallel.       |

## Sandbox Security — Enterprise-grade security.

Isolation is the product. Each sandbox is a disposable, hardware-isolated VM, governed by the same controls your security team already audits.

- **Hardware-isolated VMs** — Every session runs in its own VM with a dedicated kernel boundary, far stronger isolation than shared-kernel containers.
- **Wiped on teardown** — Filesystem and memory live only for the session. When a sandbox is destroyed, its state is gone for good, with no residue between runs.
- **Secrets never cross boundaries** — Credentials are scoped to a single session. Code, dependencies, and secrets never leak between sandboxes or onto your host.
- **Encrypted in transit** — Sessions are reached over tunneled, authenticated connections, with no public IPs or open inbound ports.

Backed by the compliance program of our parent company, Luxor Technology. [How Tenki Protects Customers](https://tenki.cloud/company/security)

## Sandbox ADE

The Agentic Development Environment (ADE) offers another way to use Sandboxes. It is a native desktop app for macOS and Linux that orchestrates Sandbox sessions for your agents. Launch a session, point your agent at it, and review every change before anything ships.

Downloads: [macOS](https://tenki.cloud/api/download?platform=mac) · [Linux (.deb)](https://tenki.cloud/api/download?platform=linux-deb) · [Linux (.AppImage)](https://tenki.cloud/api/download?platform=linux-appimage)

## Pricing and Limits — Pay for what runs. Upgrade your workspace to raise limits.

| Compute Cost      | Per Hour  | Per Second  |
| ----------------- | --------- | ----------- |
| CPU (per core)    | $0.0504   | $0.000014   |
| Memory (per GiB)  | $0.0162   | $0.0000045  |
| Storage (per GiB) | $0.000108 | $0.00000003 |

| Resource Limits | Starter Workspace | Team Workspace |
| --------------- | ----------------- | -------------- |
| Snapshots       | 10                | 50             |
| Templates       | 10                | 50             |
| Volumes         | 10                | 50             |
| Storage         | 100 GiB           | 500 GiB        |
| Active Sessions | 5                 | 25             |
| Preview URLs    | 20                | 100            |

Billed per-second across CPU, memory, and storage; the per-hour rate is shown for reference. First 5 GiB of storage included. Usage draws from your workspace monthly credits. [See all plans](https://tenki.cloud/pricing)

## Testimonials — Don't believe us. Believe them.

Customer quotes from founders, CTOs, and engineering managers on Tenki's speed, cost savings, and support.

## FAQs

Have a question we haven't answered? Reach out to hello@tenki.cloud.

### What is a sandbox?

A sandbox is an isolated, fully managed compute environment where developers and AI coding agents can write, run, and test code without affecting production systems. Think of it as a disposable virtual machine spun up on demand.

### How fast do sandboxes start?

Tenki sandboxes go from API call to a live environment in under 2 seconds, with no cold-start image pulls between sessions. That latency budget is what makes sandboxes usable inside an agent loop: you can spin up a fresh environment between tool calls without making the user wait.

### Is this a good fit for AI coding agents?

Full VM sandboxes are a great fit for agents that need to run tests, compile code, or interact with Docker. Every agent task gets its own isolated workspace, so agents can work with root access without ever touching your host machine.

### Can I run Docker inside a sandbox?

Yes. Full VM-based sandboxes support Docker, system packages, and background services. This makes them ideal for integration tests and workflows that mirror a CI/CD environment exactly.

### Can I customize the environment with my own dependencies?

Yes. You can define your environment using standard Dockerfiles, check them into source control, and share reproducible environments across your team. Sandboxes also support pre-captured environment snapshots to speed up boot times.

### How does pricing work?

Pricing is per-second across CPU, memory, and storage. You only pay while a sandbox is actively running, and paused or idle sandboxes incur no compute charges. The first 5 GiB of storage is included, and resource limits scale with your workspace plan.

### How do I get started?

Spin up your first environment in seconds with the TypeScript, Go, or Python SDK (or the CLI), or download the Sandbox ADE desktop app, launch a session, and point your agent at it.