Welcome to Runra Sandbox

Today we're excited to introduce Runra Sandbox — hosted sandbox infrastructure purpose-built for AI agents.

Why we built this

AI agents are evolving fast. They're no longer just chatbots — they clone repos, run tests, install packages, start dev servers, browse files, and operate inside real environments.

But running agent-generated code safely is hard. You don't want untrusted scripts, shell commands, and package installs running next to your production application.

Runra Sandbox gives every agent a secure, isolated, disposable execution environment — without requiring you to operate VMs, Kubernetes, or sandbox infrastructure yourself.

What makes Runra different

Built for agents, not functions

Traditional serverless platforms are designed for short-lived, stateless functions. Agents need long-running tasks, mutable workspaces, file system state, background processes, and the ability to pause and resume.

Runra Sandbox is designed for these agent-native workloads from day one.

VM-backed isolation

Each sandbox runs in its own VM-level environment with full kernel separation. File system isolation, process isolation, resource limits, and network policy controls — all built in.

Pause and resume

Agents work in sessions, not requests. Runra lets you preserve workspace state across agent runs without keeping every environment fully active. Pause when the agent is idle, resume when it has more work to do.

Open and pluggable

Runra is open source. Use the agents, LLM providers, observability tools, and sandbox backends that fit your stack. Start hosted with Runra Sandbox, or self-host via CubeSandbox.

Observability built in

Runra emits structured lifecycle and execution events — sandbox created, command executed, file changed, port exposed, errors — that you can send to Axiom, OpenTelemetry, Postgres, or your own pipeline.

Getting started

Check out the Quick Start guide to create your first sandbox:

import { Sandbox } from "runra";

const sandbox = await Sandbox.create({ timeout: 300 });

try {
  const result = await sandbox.runCode('print("Hello from Runra!")');
  console.log(result.text);
} finally {
  await sandbox.kill();
}

What's next

We're in early access and actively shipping. Here's what's on the roadmap:

• More agent adapters (Claude Code, Codex, OpenCode)
• Custom sandbox images
• Advanced network policies
• More regions
• Enterprise SSO/SAML

Start building today — no credit card required for the developer preview.