machine0 Launches: Persistent CPU and GPU Cloud VMs via CLI for AI Agents
By Mr.Xu Community Post
Published: · 15 views
Summary:machine0 has launched a command-line interface (CLI) tool designed for AI agents, offering persistent cloud virtual machines (VMs) with per-minute billing for CPU and GPU resources, including up to 60 vCPUs, 240 GB RAM, and GPUs like H100 and H200. The tool addresses key challenges in AI agent resource management, security, availability, and isolation, providing developers with a scalable and secure cloud computing environment.
Key Features and Capabilities
-
Persistent Cloud VMs: The CLI tool
machine0 new myboxcreates VMs with static IPs and HTTPS endpoints, billed per minute, starting at $0.013/hour, with configurations up to 60 vCPUs, 240 GB RAM, and GPUs like RTX 4000 Ada to 8×H200. -
Resource Management and Isolation: Each AI agent is allocated a dedicated VM to prevent resource contention and security risks.
-
Snapshot and Resume: Supports VM snapshots and resume functionality, allowing developers to pause work and resume later or create golden master images for rapid deployment of multiple instances.
-
Block Storage: Offers persistent storage volumes from 10 GB to 16 TB, manageable with intuitive syntax and attachable to VMs.
-
Profiles: Injects credentials, MCP connections, prompts, and environment variables at VM creation, ensuring each agent receives only the capabilities it needs.
-
Agent Self-Service: AI agents can manage their VM resources via the CLI or MCP server, such as spinning up build environments, creating snapshots, and tearing down instances.
-
Reproducible Builds: Supports reproducible builds using NixOS flakes or Ansible playbooks with Ubuntu.
Use Cases
-
Agent Fleet Management: Users can run a pilot agent to delegate tasks to sub-agents, each on its own VM. For example, one customer runs hundreds of machines simultaneously, spun up and torn down via the CLI.
-
Model Optimization and Reinforcement Learning Environments: ML teams use machine0 for agent-orchestrated RL environments and model optimization. For instance, one customer runs RL environments on 60 vCPU machines for days; another keeps a suspended H100 GPU for overnight inference speed optimization.
-
Product Infrastructure: Some customers build their products on top of machine0, with each user session receiving a fresh XL machine from a versioned image of their agent runtime.
Technical Architecture
machine0 is built on full KVM virtual machines, not containers or sandboxes. Each VM provides real GPU access and supports kernel-level module loading. The tech stack includes TypeScript, Postgres, and Redis, with a focus on security and performance optimization.
Developer Recommendations
For developers requiring long-duration, resource-intensive AI tasks, machine0 offers a flexible and efficient solution. It is recommended to try the tool and evaluate it based on specific needs to fully leverage its persistent computing resources and agent management capabilities.
Original Source
Launch HN: machine0 (YC S26) – Persistent CPU and GPU VMs from the CLI
— END —Tags: #machine0 #AI Agents #Cloud Computing #CLI Tool #GPU Resource Management
Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.
Community Comments