Skip to main content

Quickstart

InstaVM provides infrastructure for AI agents to execute code in isolated sandboxes. This guide gets you running code in under 5 minutes.

Get your API key​

  1. Sign up at the InstaVM Dashboard
  2. Navigate to API Keys
  3. Create a new key (starts with sk_instavm_...)

Install the SDK​

pip install instavm

Run your first code​

from instavm import InstaVM

with InstaVM('your_api_key') as vm:
    result = vm.execute("print('Hello from InstaVM')")
    print(result['output'])

That is it. InstaVM boots a microVM, runs your code inside it, and returns the output. The VM is destroyed when the session ends.

Choose your path​

AI agent builders​

Give your LLM a code execution tool in 5 lines:

from instavm import InstaVM
import json

vm = InstaVM("your_api_key")

def execute_code(code: str) -> str:
    """Tool function for OpenAI function calling."""
    result = vm.execute(code)
    return result.get("output", "")

Register execute_code as a tool with OpenAI, Anthropic, or any LLM that supports function calling. The sandbox isolates untrusted LLM-generated code from your infrastructure.

Next: Build an AI Agent Code Execution Tool

Code interpreter builders​

Build a ChatGPT-style code interpreter with persistent sessions:

from instavm import InstaVM

vm = InstaVM("your_api_key")

# State persists between execute() calls
vm.execute("import pandas as pd; df = pd.DataFrame({'x': [1,2,3]})")
result = vm.execute("print(df.describe())")
print(result['output'])

Variables, files, and installed packages carry over between calls within the same session.

Next: Build a Code Interpreter

Deploy and share​

Spin up a web server and get a public URL:

from instavm import InstaVM

vm = InstaVM("your_api_key")

vm.execute("""
import subprocess, sys
subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'flask'])
""")

vm.execute_async("""
from flask import Flask
app = Flask(__name__)

@app.route('/')
def index():
    return '<h1>Running on InstaVM</h1>'

app.run(host='0.0.0.0', port=8080)
""")

share = vm.shares.create(port=8080, is_public=True)
print(f"Live at: {share.get('public_url')}")

Next: Deploy a Web App

Next steps​