Skip to content
Runra Sandbox Docs

Introduction

Runra Sandbox is hosted sandbox infrastructure for AI agents. It provides isolated cloud workspaces where agents can safely run code, use tools, modify files, start processes, expose ports, pause, resume, and terminate.

What you can do

Section titled “What you can do”
  • Run Claude Code, Codex, OpenCode, or custom AI agents in secure sandboxes
  • Execute untrusted code, scripts, and shell commands safely
  • Preserve workspace state with pause and resume
  • Expose ports for previews, dev servers, and web apps
  • Export structured observability events to your own stack
  • Plug into the open-source Runra runtime or use the API directly

Architecture

Section titled “Architecture”
Your App
  ↓
Runra Runtime (open-source)
  ↓
Runra Sandbox API (hosted)
  ↓
CubeSandbox (VM-backed engine)

Key features

Section titled “Key features”
  • VM-backed isolation: Each sandbox runs in its own VM-level environment
  • Pause & resume: Preserve state without keeping environments active
  • Observability: Structured events for Axiom, OpenTelemetry, JSONL, Postgres
  • Pluggable: Works with your choice of agents, LLMs, and observability tools
  • Hosted or self-hosted: Use Runra Sandbox hosted or self-host via CubeSandbox

Next steps

Section titled “Next steps”