e2b-sandboxes

E2B Code Interpreter Sandboxes

Secure isolated cloud VMs for executing AI-generated code. Open-source.

Installation

pip install e2b-code-interpreter
npm i @e2b/code-interpreter
export E2B_API_KEY=your_key_here   # get at https://e2b.dev/dashboard

Python Usage

from e2b_code_interpreter import Sandbox

# Basic execution
with Sandbox() as sandbox:
    result = sandbox.run_code("x = 1 + 1; print(x)")
    print(result.logs.stdout)   # ["2"]

# Stateful multi-step session
with Sandbox() as sandbox:
    sandbox.run_code("data = [1, 2, 3]")
    r = sandbox.run_code("print(sum(data))")
    print(r.logs.stdout)        # ["6"]

TypeScript Usage

import { Sandbox } from "@e2b/code-interpreter"

const sandbox = await Sandbox.create()
const result = await sandbox.runCode("print(1 + 1)")
console.log(result.logs.stdout)
await sandbox.kill()

Use With Claude AI Agent

from anthropic import Anthropic
from e2b_code_interpreter import Sandbox

client = Anthropic()
sandbox = Sandbox()

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write Python to compute fibonacci(10)"}]
)
code = response.content[0].text
result = sandbox.run_code(code)
print(result.logs.stdout)
sandbox.kill()

Key Features

Feature	Detail
Isolation	Each sandbox = fresh cloud VM
Stateful	Variables persist within session
Languages	Python, JavaScript/TypeScript
Outputs	stdout, stderr, rich results (plots, tables)
Self-hostable	GCP (prod-ready), AWS (in progress), Azure (planned)

Self-Hosting

# GCP (production-ready)
terraform apply -var-file=gcp.tfvars

---