Show HN: Mudroom – Run coding agents in a VM, review the diff before it lands

Category: devtools

Tags: sandboxing, coding-agents, vm, diff-review, network-policy, swift, ai-safety

Score: 7.8/10 (Innovation: 8, Technical: 8, Documentation: 8, Utility: 7)

Mudroom runs coding agents like Claude Code, Codex, and Aider inside an isolated Linux micro-VM on a copy-on-write clone of your project, then presents the agent's changes as a reviewable diff you can apply per-file or per-hunk after the fact. It restricts network egress to allowlisted hosts and never mounts the real working directory, giving paranoid-mode agent execution without sacrificing autonomy. The project ships a native macOS SwiftUI app plus a cross-platform CLI with Docker/Podman backends.

Target audience: developers using AI coding agents, backend and platform engineers, devops/security-conscious teams

Repository: https://github.com/Kernel-Hunter/mudroom · Swift · MIT · 11 stars

View on Hacker News