machine0 - Persistent NixOS VMs from the CLI

TL;DR

TL;DR: machine0 is a YC-backed CLI tool that provisions persistent NixOS and Ubuntu VMs with static IPs, per-minute billing, and a remote MCP server so AI agents like Claude Code and Hermes can manage infrastructure programmatically.

Source and Accuracy Notes

• Project page: machine0.io
• Documentation: docs.machine0.io
• Source repository (images): github.com/fdmtl/machine0-nixos, github.com/fdmtl/machine0-ubuntu
• HN launch thread: news.ycombinator.com/item?id=48543245
• Source last checked: 2026-06-16

What Is machine0?

machine0 is a CLI-first cloud VM platform designed for developers who want persistent, reproducible environments without dashboard-driven workflows. Every VM runs real KVM/QEMU with dedicated CPU and RAM, a static public IP, and an HTTPS endpoint at <vm-name>.mac0.io.

The project launched on Show HN with 86 points and active discussion from developers already using it for hosting OpenClaw, Hermes, and web apps.

Key differentiators from traditional VPS providers:

• NixOS with Flakes — define your entire OS as code, deterministic builds, one-command rollbacks
• CLI-only operations — every command supports --json output for scripting and agent integration
• Remote MCP server — AI agents can create, manage, and destroy VMs programmatically
• Suspend/resume — snapshot a VM’s full state, stop billing, restore later
• Per-minute billing — starts at $0.013/hr ($9/month) for a small instance

Setup Workflow

Step 1: Install the CLI

curl -LsSf https://machine0.io/install.sh | sh

Or via npm:

npm install -g @machine0/cli

Step 2: Create Your First VM

# Create a NixOS VM with default size
machine0 new my-vm

# Specify image, size, and region
machine0 new devbox --image nixos-25-11-loaded --size large --region us

# Ubuntu with pre-installed tools (Docker, Node, Python, Claude Code, Codex)
machine0 new ubuntu-dev --image ubuntu-24-04-loaded --size medium

Step 3: SSH In and Start Working

machine0 ssh my-vm

# The VM has root access via sudo, a dedicated IP, and HTTPS at:
# https://my-vm.mac0.io

Step 4: Manage VMs

machine0 ls                    # list all VMs
machine0 get my-vm             # show details (IP, size, status)
machine0 stop my-vm            # stop (still billed, resources reserved)
machine0 suspend my-vm         # snapshot + destroy (pay only storage)
machine0 start my-vm           # resume from stopped or suspended state
machine0 rm my-vm              # destroy permanently

Step 5: Snapshot and Clone

# Create a golden master image from your configured VM
machine0 images new my-vm my-snapshot

# Remove the original
machine0 rm my-vm

# Spin up clones from the snapshot
machine0 new my-clone --image my-snapshot --size xxl

Pricing and Sizes

machine0 uses per-minute billing with the same pricing across all four regions (US, UK, EU, Asia). Suspend VMs to stop billing — you only pay for image storage ($0.078/GB/month) while suspended.

| Size | vCPU | RAM | Disk | $/hour | ~$/month | |------|------|-----|------|--------|----------| | small | 1 | 1 GB | 25 GB | $0.013 | $9 | | medium | 2 | 2 GB | 40 GB | $0.026 | $19 | | large | 2 | 4 GB | 80 GB | $0.052 | $38 | | xl | 4 | 8 GB | 160 GB | $0.104 | $76 | | xxl | 8 | 16 GB | 320 GB | $0.208 | $152 | | xxxl | 16 | 64 GB | 200 GB | $0.825 | $602 | | 4xl | 32 | 128 GB | 480 GB | $1.980 | $1,445 | | 5xl | 48 | 192 GB | 720 GB | $2.970 | $2,168 | | 6xl | 60 | 240 GB | 900 GB | $3.714 | $2,711 |

GPU instances (H100, H200, L40S, MI300X) are available for ML workloads.

MCP Integration for AI Agents

machine0 ships with a remote MCP server that lets AI agents manage infrastructure without SSH keys or manual intervention. This is particularly useful for autonomous coding agents like Claude Code, Codex, and Hermes that need their own isolated environments.

Example workflow with Claude Code:

# Create a VM and SSH in
machine0 new openclaw --image nixos-25-11-openclaw --size large
machine0 ssh openclaw

# Start the agent's onboarding flow
openclaw onboard --install-daemon

# Back up data and suspend when done
machine0 sync pull openclaw:~/.openclaw/ ~/Backups/openclaw
machine0 suspend openclaw

The MCP server exposes tools for creating VMs, listing sizes, managing images, executing commands remotely via SSH, and more. Agents can drive the entire lifecycle programmatically.

NixOS: Your OS as Code

One of machine0’s core features is first-class NixOS support with Flakes. Your entire operating system — packages, services, users, networking — is defined in a flake.nix file.

# Clone the official machine0 NixOS repo
git clone https://github.com/fdmtl/machine0-nixos.git ./

# Modify the config (e.g., add PostgreSQL)
# Then provision the VM from the flake
machine0 provision devbox ./flake.nix#loaded

Benefits of this approach:

• Deterministic builds — same flake produces identical VMs every time
• No configuration drift — manual changes outside the flake are detected and can be reset
• One-command rollbacks — if something breaks, rebuild from the previous flake state
• Agent-friendly — AI coding tools can author and verify NixOS configs programmatically

How machine0 Differs from Alternatives

vs. Traditional VPS (DigitalOcean, Linode, Hetzner): VPS platforms give you dashboards and require dozens of decisions. machine0 gives you a CLI and MCP server, NixOS declarative provisioning, suspend/resume with full state snapshots, per-minute billing, and pre-built images with developer tools.

vs. Ephemeral Sandboxes (E2B, Modal, Fly.io): Sandboxes are short-lived containers optimized for quick tasks. machine0 is persistent VMs you own — static IPs, real KVM virtualization, data persists across reboots, NixOS declarative configs. Built for workloads that run for hours or days.

vs. Cloud IDEs (GitHub Codespaces, Gitpod): Cloud IDEs are tied to specific repos and have limited customization. machine0 gives you full root access to a persistent VM with any tools you want, plus the ability to snapshot and clone environments.

Practical Evaluation Checklist

Before adopting machine0, consider:

• [ ] Do you need persistent VMs? — If your workloads are ephemeral (CI jobs, quick tests), sandboxes like E2B may be more cost-effective
• [ ] Are you comfortable with NixOS? — The NixOS learning curve is real, but Ubuntu images are available if you prefer traditional package management
• [ ] Do you need GPU instances? — machine0 offers H100, H200, L40S, and MI300X GPUs for ML workloads
• [ ] Is per-minute billing important? — If you spin up/down VMs frequently, per-minute billing saves money compared to hourly billing
• [ ] Do you use AI coding agents? — The MCP server integration makes machine0 ideal for Claude Code, Codex, and Hermes workflows

Security Notes

• VMs run on dedicated KVM/QEMU infrastructure with a 99.99% uptime SLA
• Each VM gets its own static IP and HTTPS endpoint — no NAT, no tunnels
• machine0 submits to regular external security audits and publishes a Privacy Policy and DPA
• SSH access uses your ~/.ssh/id_rsa.pub by default; you can manage multiple keys via the CLI
• All VM images are open source (NixOS, Ubuntu) — you can audit exactly what’s installed

FAQ

Q: Can I use Ubuntu instead of NixOS?

A: Yes. machine0 offers Ubuntu 24.04 images pre-installed with Docker, Node.js, Python, Claude Code, and Codex. You get the same CLI, MCP server, and per-minute billing regardless of which OS you choose.

Q: What happens to my data when I suspend a VM?

A: Suspending creates a full state snapshot of your VM (memory, disk, running processes) and destroys the instance. You pay only for image storage ($0.078/GB/month) while suspended. When you resume with machine0 start, everything is restored exactly as it was.

Q: Can I run my own custom NixOS flake?

A: Yes. Clone the official machine0-nixos repo, modify the flake.nix to add packages, services, or users, then run machine0 provision <vm-name> ./flake.nix#<configuration>. The VM rebuilds deterministically from your flake.

Q: Is there a dashboard or web UI?

A: Yes, there’s a dashboard at app.machine0.io for viewing VMs, billing, and usage. However, the primary workflow is CLI-first — every operation is available via machine0 commands with --json output for scripting.

Q: How does machine0 handle network security?

A: Each VM gets a static public IP and HTTPS endpoint at <vm-name>.mac0.io. There’s no NAT or tunneling — you have direct network access. You’re responsible for configuring firewalls and security groups within the VM itself.

Q: Can I migrate my VM to another cloud provider?

A: With NixOS, your OS is defined as code in a flake, which is portable to any NixOS host. You can pull your data anytime with machine0 sync pull. There’s no vendor lock-in — the flake can be deployed elsewhere.

Conclusion

machine0 fills a gap between ephemeral sandboxes and traditional VPS providers by offering persistent, CLI-managed VMs with NixOS declarative configuration and AI agent integration via MCP. The per-minute billing, suspend/resume functionality, and pre-built images with developer tools make it particularly well-suited for:

• AI coding agents that need isolated, persistent environments
• Developers who want reproducible dev environments without dashboard overhead
• Self-hosters who need static IPs and long-running VMs at reasonable cost
• ML engineers who need GPU instances with flexible billing

The NixOS-first approach is a double-edged sword — powerful for reproducibility and agent automation, but requires learning the Nix ecosystem. Ubuntu images are available as a fallback.

At $9/month for a small instance with per-minute billing, machine0 is competitive with traditional VPS providers while offering significantly more developer-friendly workflows.