Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Hyper-V Sandbox MCP Server Run this PowerShell script in an isolated VM and show me the output"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Secure agentic sandbox for autonomous testing, validation, and debugging of PowerShell, CMD/BAT scripts, and system utilities inside isolated Hyper-V virtual machines.
Hyper-V Sandbox MCP Server is a Model Context Protocol (MCP) server that turns your local Microsoft Hyper-V hypervisor into an AI-agent-managed testing ground for code. The agent (Antigravity 2.0, Claude Desktop, etc.) gains the ability to autonomously revert VMs to clean snapshots, transfer files out-of-band via VMBus, execute commands inside the guest OS using PowerShell Direct, analyze output, and self-correct on errors (Self-Healing Loop).
🌟 Key Features
🛡️ 100% Host Protection (Zero-Risk Sandbox): Potentially destructive scripts (system service modifications, registry edits, network adapter changes, file deletions) execute strictly inside an isolated virtual machine.
⚡ PowerShell Direct (Zero-Network Dependency): Guest OS management operates via the hypervisor's internal VMBus. No network setup, open RDP/SSH/WinRM ports, or firewall tweaks required.
🔄 Deterministic Rollback (CleanBase Checkpoint): Instant VM resets to a clean baseline snapshot before every run ensure fully reproducible tests without leftover state.
📦 Seamless File Transfer (Guest Integration Services): Native Copy-VMFile usage to transfer scripts, libraries, and utilities (e.g., PsExec) directly from host to guest.
🤖 Self-Healing Code Loop: The agent captures real ExitCode, stdout, and stderr outputs alongside exception stack traces to autonomously patch code and re-test until green.
This server cannot be deployed
Validates AI infra code on real VMs. Self-corrects until it works. No containers, no sandboxes.
Build, validate, and deploy multi-agent AI solutions from any AI environment.
Persistent Linux microVMs for agents: root, internet, sub-second resume and a public URL.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
