Skip to main content
Moru provides isolated cloud sandboxes for AI agents that need full computer access—like Claude Agent SDK—to run commands, write files, and execute code.

​
Why Moru?

Hardware Isolation

Each sandbox runs in its own Firecracker microVM, providing strong security isolation between workloads.

Full Environment

Complete Linux environment with filesystem, networking, and command execution - everything your agent needs.

Real-time Streaming

Stream command output, file changes, and logs in real-time as your agent works.

Session Persistence

Pause and resume sandboxes to maintain agent state across sessions.

​
Quick Example

Create a sandbox, run a command, and read the output: 
from moru import Sandbox

# Create a sandbox
sandbox = Sandbox.create()

# Run a command
result = sandbox.commands.run("echo 'Hello from Moru!'")
print(result.stdout)  # Hello from Moru!

# Clean up
sandbox.kill()

​
Core Concepts

​
Sandboxes

A sandbox is an isolated compute environment where your agent can safely execute code, manipulate files, and run commands. Each sandbox:
  • Runs in a dedicated Firecracker microVM
  • Has its own filesystem (10 GB default)
  • Can execute arbitrary shell commands
  • Supports network access (configurable)
  • Auto-terminates after a timeout (default 5 minutes)

​
Templates

Templates define the starting state of a sandbox - the base image, installed packages, and startup configuration. Use built-in templates or create custom ones with your specific dependencies.

​
Files and Commands

Sandboxes provide full access to:
  • Filesystem operations: Read, write, list, watch files and directories
  • Command execution: Run shell commands with streaming output
  • PTY support: Interactive terminal sessions for REPL environments

​
Get Started

Quickstart

Create your first sandbox in 5 minutes.

Sandbox Overview

Learn how sandboxes work and their lifecycle.

Python SDK

Complete Python SDK reference.

JavaScript SDK

Complete JavaScript/TypeScript SDK reference.

​
Use Cases

Moru is designed for AI agent workloads:
  • Code execution agents: Run code generated by LLMs in isolated environments
  • Development assistants: Provide agents with full development environments
  • Data processing: Execute data transformation pipelines securely
  • Testing and CI: Run tests in clean, reproducible environments

Claude Agent SDK Integration

Learn how to run Claude Agent SDK in Moru sandboxes with session persistence and tool execution.