← DIRECTORY
AI TOOL

machine0

CLI- and MCP-driven persistent cloud VMs for long-running AI agents, with snapshots, suspend/resume, static IPs and CPU/GPU configurations.

8.2OUT OF 10 · LOW CONF.

Product facts

Category
—
Pricing
Pay-as-you-go per minute. CPU VMs start at $0.013/hour; GPU VMs start at $0.836/hour; larger configurations scale upward.
Free access
NO
Platforms
CLI, Cloud, MCP
API
YES
Open source
NO
Integrations
MCP, GitHub, Claude Code, Codex
Last updated
2026-09-06

Product sources: machine0 official product/pricing surface + Y Combinator S26 company page

machine0 review

Updated 2026-09-08

Product & capabilities

machine0 is a cloud-VM control layer designed for long-running AI agents rather than short-lived sandbox sessions. Its CLI and remote MCP interface can provision persistent NixOS or Ubuntu machines, start and stop them, suspend and resume state, create snapshots and clones, expose static IPs and HTTPS endpoints, and select CPU or GPU configurations.1 Profiles can inject GitHub access, Claude Code, Codex, remote MCP servers, prompts and environment variables into a new VM, while recent releases added reproducible image building from Git repositories and GPU-ready NixOS images.2

The product is deliberately infrastructure-focused: users receive root SSH access to machines running on DigitalOcean infrastructure rather than a constrained application sandbox.3 Y Combinator independently corroborates the Summer 2026 company and the same CLI/MCP persistent-agent-compute positioning.4

Pricing & access

Pricing is usage-based and billed by the minute. CPU VMs start at $0.013 per hour and GPU VMs at $0.836 per hour, with larger CPU, memory and accelerator configurations priced progressively higher.1 Suspending a VM stops compute billing while preserving state; stored images continue to incur storage charges. The Terms state a $5 minimum account top-up.3

There is no general free tier on the current public surface. For small always-on CPU workloads the entry price is competitive with ordinary cloud hosting, while larger machines and GPUs can become substantial ongoing spend. The economic advantage is therefore strongest when suspend/resume and per-minute provisioning are actively used rather than leaving oversized machines running.

Evidence & trust

The Privacy Policy distinguishes machine0's role as controller for account, authentication, billing and infrastructure metadata from its processor role for data stored inside customer VMs. It says VM contents are not routinely accessed or inspected; remote-command output and optional HTTP-proxy traffic are relayed in real time rather than stored or logged by machine0.5 Account data is hosted through Railway in the United States, VM data resides in the selected region, and the policy lists DigitalOcean, Cloudflare, Stripe, Resend, Google, EU-hosted PostHog and optional Sentry as subprocessors.5

The same policy publishes retention periods and security controls including TLS, SSH, bcrypt password hashing, rate limiting and encryption at rest for managed SSH private keys; those private keys are retrievable through the API.5 The Terms confirm full root SSH access, customer ownership of VM data and DigitalOcean as the underlying VM provider.3 These disclosures improve transparency, but they remain vendor-authored controls rather than an independent security audit. Likewise, the advertised 99.99% uptime is a first-party/infrastructure claim, not an independently validated reliability measurement.1

Independent product evidence is still limited. Y Combinator confirms the active company and product surface, while APIs.io independently catalogs machine0's developer API and persistent VM lifecycle; neither provides a substantive reliability or security benchmark.46

Who it's for

machine0 is best suited to developers and infrastructure teams running coding agents, autonomous research jobs, agent swarms or other workloads that need a real machine to remain alive for hours or days. CLI/MCP control, JSON output, profiles, snapshots and reproducible images make it particularly relevant when the agent itself must create or manage compute.12

It is less suitable for users who want a fully managed no-code experience, a broad free tier, or independently certified infrastructure controls. Teams handling sensitive workloads should review the account-data, credential and subprocessor boundaries in the Privacy Policy and decide whether a root-accessible VM on the stated infrastructure matches their own security model.5

Strengths & weaknesses

Strengths

persistent agent-native VMs; CLI and MCP lifecycle control; CPU and GPU choices; static networking and HTTPS; suspend/resume, snapshots and cloning; reproducible NixOS workflows; useful GitHub, Claude Code and Codex integration; transparent per-minute pricing; and unusually specific public privacy, retention and subprocessor disclosures for an early-stage product.125

Weaknesses

no general free tier; a very young product with limited independent evidence; vendor-supplied uptime/security assertions that should not be read as third-party validation; and cloud-held account/authentication metadata and managed credentials even when VM data itself is region-selected.53

SOURCES

Sources & references

6 sources
  1. Official sourcemachine0 — Powerful & Persistent Virtual Machines
    OFFICIAL
  2. Official sourcemachine0 Changelog
    OFFICIAL
  3. Official sourcemachine0 Terms of Service
    OFFICIAL
  4. Sourcemachine0: AWS for Agents | Y Combinator
    REPORT
  5. Official sourcemachine0 Privacy Policy
    OFFICIAL
  6. SourceMachine0 | APIs.io Providers
    REPORT

Related AI tools

Products matched by shared tasks, audience, workflow, product type and capabilities.