Connect an existing agent.
Keep your agent in your app or on your laptop. It sends commands to a VMNest computer and gets the results back.
Give your AI agent
a computer to work on.
Your agent decides what to do. VMNest gives it an isolated cloud computer to run commands, install tools, and work with files.
Agents at these companies use VMNest
One agent. Multiple computers.
Give your agent a separate computer for each task, each with its own files, tools, and running processes.
Two ways to run your agent’s work.
Run the agent inside VMNest.
Run your agent on the cloud computer itself, alongside the files and tools it needs.
Do several jobs at once.
Give each part of a code review its own computer. Check security, run tests, and review dependencies at the same time.
Turn incoming events into tasks.
Planned: connect events from your app to an agent. A new bug report could start an investigation and a proposed code fix.
Restart one task. Keep the others running.
Each task runs on its own isolated computer. Restart one without stopping the others. Select a computer to see how it works.
Spin up agent workspaces as part of your workflows.
Give your agents a place to work whenever a task starts. Create a workspace from Python, run the task, and shut it down when the work is done.
Explore the docsPython SDK
import asyncio
from agent_sandbox import Sandbox
async def main():
vm = await Sandbox.create("my-agent")
try:
result = await vm.exec("python", ["-c", "print('Hello')"])
print(result.stdout_text)
finally:
await vm.stop()
asyncio.run(main())Using the SDK directly? Set up the backend first. Setup guide