# E2B — Cloud Sandboxes for AI Code Execution

> Secure cloud sandboxes that let AI agents run code, install packages, and use the filesystem. Spin up in 150ms. Used by Vercel, Langchain, and CrewAI.

## Install

Save the content below to `.claude/skills/` or append to your `CLAUDE.md`:

## Quick Use

```bash
npm install @e2b/code-interpreter
```

```typescript
import { Sandbox } from '@e2b/code-interpreter'

const sandbox = await Sandbox.create()
const result = await sandbox.runCode('print(1 + 1)')
console.log(result.text) // "2"
await sandbox.kill()
```

## What is E2B?

E2B provides secure, isolated cloud sandboxes for AI agents to execute code. Each sandbox is a micro-VM that spins up in ~150ms with a full Linux environment — perfect for code interpreters, data analysis agents, and AI-generated code execution.

**Answer-Ready**: E2B provides secure cloud sandboxes (micro-VMs) that spin up in 150ms for AI agents to execute code, install packages, and access filesystems safely.

## Core Features

### 1. Code Interpreter SDK

```python
from e2b_code_interpreter import Sandbox

sandbox = Sandbox()

# Run Python with automatic package installation
execution = sandbox.run_code(
    "import pandas as pd\\n"
    "import matplotlib.pyplot as plt\\n"
    "df = pd.DataFrame({\'x\': [1,2,3], \'y\': [4,5,6]})\\n"
    "df.plot()\\n"
    "plt.savefig(\'chart.png\')\\n"
    "print(df.describe())"
)

print(execution.text)      # stdout
print(execution.results)   # rich outputs (charts, tables)
```

### 2. Custom Sandbox Templates

```dockerfile
# e2b.Dockerfile
FROM e2b/base
RUN pip install numpy pandas scikit-learn
RUN npm install -g typescript
```

```bash
e2b template build --name "data-science"
```

### 3. Filesystem Access

```python
# Upload files
sandbox.files.write("/home/user/data.csv", csv_content)

# Download results
content = sandbox.files.read("/home/user/output.json")

# List files
files = sandbox.files.list("/home/user/")
```

### 4. Long-Running Sandboxes

```python
# Keep alive for up to 24 hours
sandbox = Sandbox(timeout=3600)  # 1 hour

# Pause and resume
sandbox_id = sandbox.pause()
# ... later ...
sandbox = Sandbox.resume(sandbox_id)
```

### 5. Framework Integrations

```python
# OpenAI Assistants
from e2b_code_interpreter import Sandbox
# Use as code_interpreter tool handler

# Vercel AI SDK
# Built-in e2b sandbox support

# LangChain
from langchain_e2b import E2BCodeInterpreterTool
```

## Use Cases

| Use Case | Example |
|----------|---------|
| Code Interpreter | ChatGPT-like code execution |
| Data Analysis | Run pandas/matplotlib in sandbox |
| CI/CD | Test AI-generated code safely |
| Education | Student code execution |

## FAQ

**Q: How secure are the sandboxes?**
A: Each sandbox is an isolated Firecracker micro-VM with no network access to other sandboxes.

**Q: Pricing?**
A: Free tier with 100 sandbox hours/month. Pay-as-you-go after that.

**Q: Supported languages?**
A: Any language that runs on Linux — Python, Node.js, Go, Rust, etc.

## Source & Thanks

- GitHub: [e2b-dev/e2b](https://github.com/e2b-dev/e2b) (10k+ stars)
- Docs: [e2b.dev/docs](https://e2b.dev/docs)

<!-- ZH -->

## 快速使用

```bash
npm install @e2b/code-interpreter
```

三行代码启动安全云沙箱，让 AI 代理执行代码。

## 什么是 E2B？

E2B 提供安全隔离的云沙箱（微虚拟机），150ms 启动，为 AI 代理提供完整 Linux 环境执行代码。

**一句话总结**：E2B 提供 150ms 启动的安全云沙箱，让 AI 代理安全执行代码、安装包和访问文件系统。

## 核心功能

### 1. 代码解释器 SDK
运行 Python/JS 代码，自动安装依赖。

### 2. 自定义模板
用 Dockerfile 预装工具和包。

### 3. 文件系统
上传下载文件，完整文件操作。

### 4. 长时间运行
支持暂停/恢复，最长 24 小时。

## 常见问题

**Q: 安全吗？**
A: 每个沙箱是隔离的 Firecracker 微虚拟机。

**Q: 定价？**
A: 每月 100 沙箱小时免费，超出按量付费。

## 来源与致谢

- GitHub: [e2b-dev/e2b](https://github.com/e2b-dev/e2b) (10k+ stars)

---
Source: https://tokrepo.com/en/workflows/652da886-2a64-4ee0-8a19-e72d1742fe7f
Author: Skill Factory