The modern runtime for agent-native development

Website • Join Waitlist • Discord • Contact

---

What is VM0?

VM0 helps you build agents and automated workflows using natural language.

It also provides a full agent runtime with:

• Secure cloud sandbox: isolated execution environments for running agents safely
• Versioned storage: track, reproduce, and roll back agent state and artifacts
• Persistent execution: resume, fork, and iterate on agent runs with full history
• Complete observability: logs, metrics, and network visibility for every run

You can build and extend agents using:

• Natural language workflows: define agent behavior in plain text, without complex graphs or pipelines
• Pluggable skills: add capabilities through reusable agent skills

Supported agents:

• Claude Code
• OpenAI Codex (in beta)
• Gemini CLI (coming soon)
• More coming soon...

Installation

npm install -g @vm0/cli

VM0 is in beta. Join the waitlist to get access.

Quick Start

# 1. Login
vm0 auth login

# 2. Create project directory
mkdir my-agent && cd my-agent

# 3. Initialize project
vm0 init

# 4. Get Claude Code token and save to .env
claude setup-token
echo "CLAUDE_CODE_OAUTH_TOKEN=<your-token>" > .env

# 5. Run your agent
vm0 cook "let's start working."
# Agent follows AGENTS.md: curates AI news from HackerNews and writes to content.md

What just happened?

• vm0 init created vm0.yaml (agent config) and AGENTS.md (agent instructions)
• vm0 cook initialized storage, composed the agent, and ran your prompt
• Results are in the artifact/ directory

Next steps

• Edit AGENTS.md to customize agent instructions
• Edit vm0.yaml to configure agent skills

For more examples, see vm0-cookbooks.

CLI Reference

Use vm0 --help or vm0 <command> --help for detailed usage information.

Authentication

Command	Description
vm0 auth login	Login to VM0
vm0 auth logout	Logout from VM0
vm0 auth status	Check authentication status
vm0 auth setup-token	Output auth token for CI/CD environments

Tutorial Commands

Command	Description
vm0 init	Initialize a new project (creates vm0.yaml + AGENTS.md)
vm0 cook "<prompt>"	One-click setup and run
vm0 cook continue "<prompt>"	Continue from last session
vm0 cook resume "<prompt>"	Resume from last checkpoint
vm0 cook logs	View logs from last run

Agent Commands

Command	Description
vm0 compose <config.yaml>	Create/update agent from config
vm0 run <agent> "<prompt>"	Run agent with prompt
vm0 run continue <session-id> "<prompt>"	Continue from session
vm0 run resume <checkpoint-id> "<prompt>"	Resume from checkpoint
vm0 logs <run-id>	View run logs

Storage Commands

Command	Description
vm0 artifact init	Initialize artifact in current directory
vm0 artifact push	Upload artifact to cloud
vm0 artifact pull [version]	Download artifact from cloud
vm0 volume init	Initialize volume in current directory
vm0 volume push	Upload volume to cloud
vm0 volume pull [version]	Download volume from cloud

Concepts

Concept	Description
Agent	Runs in sandbox with filesystem, mounts volumes and artifact
Volume	Mounts agent instructions, skills, and datasets (read-only)
Artifact	Versioned workspace where agents read/write files
Checkpoint	Immutable snapshot of run state (artifact, volumes, conversation)
Session	Continue conversation with latest artifact version

Resources

• vm0-cookbooks - Ready-to-run examples
• vm0-skills - Agent skills library
• Contributing guide - Development setup
• Website - Official website
• Discord - Join our community
• Email - Questions and support

License

See LICENSE for details.