Runra Sandbox Documentation

import { Sandbox } from "runra";

const sandbox = await Sandbox.create();
const result = await sandbox.runCode(
  'print("hello from Runra")'
);

console.log(result.text);
await sandbox.kill();

Runra is a sandbox infrastructure layer for agent platforms. Create a VM-backed workspace on demand, run generated code or shell commands, move files in and out, pause state when work is idle, and terminate the environment when the task is complete.

Run code in minutes

Use the Runra SDK to create a sandbox and execute code from your application.

JavaScript and TypeScript

npm install runra
# or
bun add runra

export RUNRA_API_KEY=runra_your_api_key
# Optional. Defaults to https://box.runra.dev
export RUNRA_API_URL=https://box.runra.dev

import { Sandbox } from "runra";

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

try {
  const execution = await sandbox.runCode(`
name = "Runra"
print(f"hello from {name}")
2 + 2
`);

  console.log(execution.logs.stdout.join("\n"));
  console.log("Result:", execution.text);
} finally {
  await sandbox.kill();
}

Python

pip install runra e2b-code-interpreter

from runra import Sandbox

sandbox = Sandbox.create(timeout=300)
try:
    execution = sandbox.run_code('print("hello from Runra")')
    print("\n".join(execution.logs.stdout))
finally:
    sandbox.kill()

Start building

Pick the path closest to your workload. The cards below keep the existing docs flow, but present it like a product API surface instead of a plain table of contents.

Core API surface

The most common production flow is create, execute, optionally pause, and delete. Use the full API reference for request and response schemas.

How to use the docs

• New to Runra: start with the Quickstart and then connect an LLM agent.
• Moving data through sandboxes: read Upload and download files.
• Custom runtime needs: read Install custom packages and Configuration.
• Building against HTTP endpoints: read the API overview and Authentication.

Environment variables