Shuru

What it does

Shuru provides lightweight, ephemeral Linux virtual machines (microVMs) that run natively on macOS (Apple Silicon) and Linux ARM64. It uses Apple's Virtualization.framework on macOS and KVM on Linux, so there is no emulation layer or Docker dependency. VMs boot from a clean rootfs each time by default, and changes vanish on exit unless explicitly saved. Shuru supports checkpointing—saving disk state as named snapshots that can be restored, branched, and reused like git commits. It also offers opt-in networking (offline by default), configurable CPU/memory/disk, directory mounts (read-only by default, with optional host writes), and port forwarding via vsock without network access.

Who it is for

Shuru is designed for developers and AI agent builders who need safe, isolated execution environments on macOS or Linux. It is especially useful for running AI-generated code, letting agents install packages and use system tools, running reproducible evaluations in parallel, and creating disposable Linux environments for testing and debugging. The tool ships as an "agent skill" that integrates with Claude Code, Cursor, GitHub Copilot, Gemini CLI, OpenAI Codex, and more, so AI agents can automatically use Shuru for sandboxed execution without manual prompting.

Why it matters

Running untrusted code—especially AI-generated code—requires strong isolation to protect the host system. Traditional solutions like Docker can be heavy and require daemon setup, while full VMs are slow. Shuru offers near-native speed on ARM64 with minimal overhead, ephemeral sandboxes that start and tear down instantly, and checkpointing for reproducible environments. Its local-first design means no cloud dependency, and its agent skill integration makes it easy for AI tools to adopt safe execution automatically. This lowers the barrier for developers to experiment with AI agents without compromising security.

Launch signal

Shuru was launched as a Show HN on Hacker News with the title "Local-First Linux MicroVMs for macOS." The project is open-source (GitHub) and offers a simple CLI install via Homebrew or a script. The website and GitHub repository provide full documentation, including configuration via a shuru.json file for secrets, network allowlists, and resource limits.

Brand and naming

The name "Shuru" is short, memorable, and easy to type—important for a CLI tool. It does not obviously relate to "sandbox" or "virtual machine," which may make its purpose less intuitive at first glance. However, the tagline "Local-First Sandboxes for AI Agents" clearly positions it as a developer tool for safe code execution. The brand emphasizes speed, ephemerality, and agent integration, appealing to the AI developer community.