vm0

Original：🇺🇸 English
Translated

VM0 API for running AI agents in secure sandboxes. Use this skill to execute agents, manage runs, and download outputs (artifacts) and inputs (volumes) via the VM0 platform API.

2installs
Sourcevm0-ai/vm0-skills
Added on

NPX Install

npx skill4agent add vm0-ai/vm0-skills vm0

Tags

Translated version includes tags in frontmatter
vm0-apiai-agent-managementsandbox-executionartifact-downloadvolume-management

SKILL.md Content

View Translation Comparison →

VM0 API

Execute AI agents in secure sandboxed environments and manage their inputs/outputs.
Note: If you have the VM0 CLI installed and the vm0-cli skill available, prefer using the vm0-cli skill instead. It provides a higher-level interface with more convenient operations.

When to Use

  • Execute AI agents in isolated sandbox environments
  • Monitor and manage agent runs (status, logs, metrics)
  • List and download input files (volumes) provided to agents
  • List and download output files (artifacts) created by agents

Prerequisites

bash
export VM0_TOKEN=vm0_live_your-api-key

Get API Key

  1. Install the VM0 CLI: https://docs.vm0.ai/docs/getting-started
  2. Run 
    vm0 auth login
    to authenticate
  3. Run 
    vm0 auth setup-token
    to view your API key
  4. Copy the token starting with 
    vm0_live_
Important: When using 
$VAR
in a command that pipes to another command, wrap only the curl command in 
bash -c '...'
, then pipe to jq outside. Due to a Claude Code bug, environment variables are silently cleared when pipes are used directly.
bash
bash -c 'curl -s "https://api.vm0.ai/v1/agents" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data'

Quick Start

List Your Agents

bash
bash -c 'curl -s "https://api.vm0.ai/v1/agents" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data[] | {id, name}'

Run an Agent

Write to 
/tmp/vm0_request.json
:
json
{
  "agent": "my-agent",
  "prompt": "Hello, please introduce yourself"
}
Then run:
bash
bash -c 'curl -s -X POST "https://api.vm0.ai/v1/runs" -H "Authorization: Bearer $VM0_TOKEN" -H "Content-Type: application/json" -d @/tmp/vm0_request.json' | jq '{id, status}'

Check Run Status

Replace 
<run-id>
with your run ID:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/runs/<run-id>" -H "Authorization: Bearer $VM0_TOKEN"' | jq '{id, status, error, execution_time_ms}'

Get Run Logs

bash
bash -c 'curl -s "https://api.vm0.ai/v1/runs/<run-id>/logs" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data[]'

Core Operations

Agents

List all agents:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/agents" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data'
Get agent details:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/agents/<agent-id>" -H "Authorization: Bearer $VM0_TOKEN"' | jq '{id, name, description}'
See references/agents.md for version listing.

Runs

Create a run with variables:
bash
curl -s -X POST "https://api.vm0.ai/v1/runs" \
  -H "Authorization: Bearer $VM0_TOKEN" \
  -H "Content-Type: application/json" \
  -d @- << 'EOF'
{
  "agent": "my-agent",
  "prompt": "Process the data file",
  "variables": {
    "DEBUG": "true"
  },
  "volumes": {
    "input-data": "latest"
  }
}
EOF
Cancel a running execution:
bash
bash -c 'curl -s -X POST "https://api.vm0.ai/v1/runs/<run-id>/cancel" -H "Authorization: Bearer $VM0_TOKEN"' | jq '{id, status}'
See references/runs.md for events streaming, logs filtering, and metrics.

Volumes (Input Storage)

List volumes:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/volumes" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data[] | {id, name}'
Download volume as tar.gz archive (follows 302 redirect):
bash
curl -L -o volume.tar.gz "https://api.vm0.ai/v1/volumes/<volume-id>/download" \
  -H "Authorization: Bearer $VM0_TOKEN"
See references/volumes.md for version listing and download options.

Artifacts (Output Storage)

List artifacts:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/artifacts" -H "Authorization: Bearer $VM0_TOKEN"' | jq '.data[] | {id, name}'
Download artifact as tar.gz archive (follows 302 redirect):
bash
curl -L -o artifact.tar.gz "https://api.vm0.ai/v1/artifacts/<artifact-id>/download" \
  -H "Authorization: Bearer $VM0_TOKEN"
Extract downloaded archive:
bash
tar -xzf artifact.tar.gz -C ./output/
See references/artifacts.md for version listing and download options.

Common Patterns

Pagination

List endpoints support cursor-based pagination:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/runs?limit=10" -H "Authorization: Bearer $VM0_TOKEN"' | jq '{data, pagination}'
Response includes:
json
{
  "pagination": {
    "has_more": true,
    "next_cursor": "abc123"
  }
}
Fetch next page:
bash
bash -c 'curl -s "https://api.vm0.ai/v1/runs?limit=10&cursor=abc123" -H "Authorization: Bearer $VM0_TOKEN"' | jq '{data, pagination}'

Error Handling

All errors return a consistent format:
json
{
  "error": {
    "type": "invalid_request_error",
    "code": "resource_not_found",
    "message": "No such agent: 'my-agent'",
    "param": "agent"
  }
}
Error TypeStatusDescription
authentication_error
401Invalid or missing API key
invalid_request_error
400Invalid parameters
not_found_error
404Resource doesn't exist
api_error
500Internal server error

Detailed References

  • Agents API - List agents and versions
  • Runs API - Execute agents, stream events, get logs and metrics
  • Artifacts API - List and download agent outputs
  • Volumes API - List and download input files

API Reference