How to beginner · 3 min read

How to run AI-generated code safely with E2B

Q: How to run AI-generated code safely with E2B

Use the e2b_code_interpreter package to run AI-generated code safely inside a secure sandbox environment. Instantiate Sandbox with your API key, run code via run_code(), and manage files securely without risking your host system.

Quick answer

Use the e2b_code_interpreter package to run AI-generated code safely inside a secure sandbox environment. Instantiate Sandbox with your API key, run code via run_code(), and manage files securely without risking your host system.

PREREQUISITES

Python 3.8+
E2B API key (set as environment variable E2B_API_KEY)
pip install e2b-code-interpreter

Setup

Install the e2b-code-interpreter package and set your E2B API key as an environment variable for secure authentication.

bash

pip install e2b-code-interpreter

output

Collecting e2b-code-interpreter
  Downloading e2b_code_interpreter-1.0.0-py3-none-any.whl (15 kB)
Installing collected packages: e2b-code-interpreter
Successfully installed e2b-code-interpreter-1.0.0

Step by step

This example shows how to create a sandbox, run AI-generated Python code safely, and retrieve the output.

python

import os
from e2b_code_interpreter import Sandbox

# Initialize sandbox with API key from environment
sandbox = Sandbox(api_key=os.environ["E2B_API_KEY"])

# AI-generated code to run
code = """
print('Hello from sandbox!')
result = 2 + 2
print(f'Result: {result}')
"""

# Run code safely inside sandbox
execution = sandbox.run_code(code)

# Print output
print(execution.text)

# Close sandbox session
sandbox.close()

output

Hello from sandbox!
Result: 4

Common variations

You can write files into the sandbox before running code, install Python packages dynamically, or run asynchronous code if supported.

Write files: sandbox.files.write(filename, data)
Install packages: sandbox.run_code("import subprocess; subprocess.run(['pip','install','pandas'])")
Run multiple code snippets sequentially

python

import os
from e2b_code_interpreter import Sandbox

sandbox = Sandbox(api_key=os.environ["E2B_API_KEY"])

# Write a data file into sandbox
sandbox.files.write("data.csv", b"col1,col2\n1,2\n3,4")

# Run code that reads the file
code = """
with open('data.csv') as f:
    print(f.read())
"""
execution = sandbox.run_code(code)
print(execution.text)

# Install a package dynamically
sandbox.run_code("import subprocess; subprocess.run(['pip','install','numpy'])")

sandbox.close()

output

col1,col2
1,2
3,4

Troubleshooting

If you see authentication errors, verify your E2B_API_KEY environment variable is set correctly.
For code execution errors, check the execution.text output for Python exceptions.
Sandbox sessions should be closed with sandbox.close() to free resources.
File writes must be in bytes; encode strings before writing.

✅

Key Takeaways

Use Sandbox from e2b_code_interpreter to isolate AI-generated code safely.
Always set your API key via the E2B_API_KEY environment variable for secure access.
Manage files and dependencies inside the sandbox to avoid host system risks.
Close sandbox sessions after use to release resources and avoid leaks.

Verified 2026-04

Verify ↗