Documentation

Runra Sandbox gives your AI agents isolated cloud environments for running code, writing files, installing packages, and connecting to existing agent loops. The SDK is E2B-compatible, with Runra-native API keys and endpoints.

QuickstartCreate a sandbox and run code

Install the SDK, set RUNRA_API_KEY, execute Python, and clean up the sandbox.

LLM agentsConnect an LLM to Runra

Expose a run_python tool so an LLM can safely execute generated code.

FilesUpload and download files

Write inputs into the sandbox, run transforms, and read generated outputs.

PackagesInstall custom packages

Install Python and Node packages at runtime, or prebuild a reusable template.

What to read first

1. Quickstart - create your first sandbox.
2. Connect LLMs - use a sandbox as an agent tool.
3. Upload and download files - move data in and out.
4. Install custom packages - add dependencies.

Runtime model

A Runra sandbox is a short-lived cloud workspace. You create it from your application, run code or shell commands inside it, optionally persist work by writing files, and terminate it when the task is complete.

import { Sandbox } from "runra";

const sbx = await Sandbox.create();
try {
  const execution = await sbx.runCode('print("hello from Runra")');
  console.log(execution.logs.stdout.join("\n"));
} finally {
  await sbx.kill();
}

Environment variables

Variable	Required	Description
RUNRA_API_KEY	Yes	Your Runra API key, for example runra_....
RUNRA_API_URL	No	Defaults to https://box.runra.dev.
RUNRA_TEMPLATE_ID	No	Default sandbox template ID.