</>MCP Agents Market
Skill

shuru

by superhq-ai851RustUpdated 2026-08-05

A local-first microVM sandbox for running AI agents safely on macOS & Linux

Claude CodeCursorCopilot

Shuru is a local-first microVM sandbox agent skill that enables AI agents to safely execute code in isolated, ephemeral Linux environments on macOS and Linux. It leverages Apple's Virtualization.framework on macOS and KVM on ARM64 Linux hosts to boot lightweight virtual machines that reset after every run, preventing code execution from affecting the host system. Developers can install it via Homebrew or the install script, then add it to their agent configuration so assistants automatically use sandboxed environments for running untrusted code, installing packages, and testing scripts.

Key Features

Ephemeral microVMs that reset rootfs on every run for disposable, isolated code execution
Native macOS support via Apple Virtualization.framework on Apple Silicon (macOS 14+)
Experimental Linux ARM64 support using KVM backend
Network control with host allowlisting, DNS resolver selection, and per-host restrictions
Checkpoint system to save and reuse VM disk states across runs without rebuilding environments
Port forwarding over vsock for accessing guest services without network device requirements
Secret injection via proxy that substitutes real API keys only on HTTPS requests to specified hosts
Directory mounting with VirtioFS supporting read-only overlays or direct read-write access to host filesystems

Use Cases

  • 01AI agents safely executing generated code snippets in isolated environments without risking host contamination
  • 02Running package installation and dependency testing in ephemeral VMs that never persist changes
  • 03Building reusable development environments with checkpoints for Python, Node.js, or compiled language projects
  • 04Testing API integrations with secret injection that keeps real credentials on the host
  • 05Running CI/CD validation steps in fresh VMs with custom CPU, memory, and disk configurations
  • 06Developing agent workflows that need temporary filesystem access via overlays without modifying source directories

Related Skills

View more

shuru — FAQ

What is shuru and what does it do?+

Shuru is a local-first microVM sandbox agent skill that lets AI assistants run code in isolated, ephemeral Linux virtual machines on macOS and Linux. Every sandbox resets after each run, so agents can execute untrusted code, install packages, and test scripts without affecting your host system.

How do I install shuru on macOS?+

On macOS with Apple Silicon, install via Homebrew with 'brew tap superhq-ai/tap && brew install shuru' or use the install script with 'curl -fsSL https://raw.githubusercontent.com/superhq-ai/shuru/main/install.sh | sh'. macOS 14 (Sonoma) or later is required.

Which AI agents and clients work with shuru?+

Shuru works as an agent skill with Claude Code, Cursor, Copilot, and other AI assistants that support agent skills. Install it via 'npx skills add superhq-ai/shuru' or manually copy the skill directory into your agent's configuration folder so it can use sandboxed execution automatically.

Does shuru work on Linux?+

Yes, but Linux support is experimental and only available for ARM64 hosts with KVM access (/dev/kvm). Install via the install script or download the linux-aarch64 tarball from GitHub Releases; Homebrew is macOS-only.

Do I need API keys or external services to use shuru?+

No, shuru runs entirely locally using your machine's virtualization capabilities (Apple Virtualization.framework or KVM). The optional secret injection feature can reference your existing API keys on the host, but shuru itself requires no external credentials.

Is shuru free and open source?+

Yes, shuru is open source and available at github.com/superhq-ai/shuru. You can install and use it freely on supported platforms.

How do I install shuru?+

Open the source repository on GitHub and follow its README. shuru is a skill — MCP Agents Market links you directly to the official repo.

Is shuru free?+

shuru is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches