E2B MCP

Overview

E2B MCP is a Model Context Protocol server that provides secure, sandboxed code execution for AI agents. E2B (Environment to Build) offers isolated execution environments for Python, Node.js, and other languages, enabling AI agents to run code safely without risking host system security.

Unlike local code execution, E2B runs code in cloud-based sandboxed containers with network isolation, resource limits, and automatic cleanup. This makes it ideal for executing untrusted code, running data analysis, generating visualizations, and more.

Features

• Sandboxed Execution: Isolated containers for secure code running
• Multi-Language Support: Python, Node.js, and more
• Package Installation: Install packages at runtime
• File System: Read/write files in sandbox
• Network Access: Optional internet access for API calls
• Resource Limits: CPU, memory, and time limits
• Automatic Cleanup: Sandboxes auto-destroy after use
• Real-Time Output: Stream stdout/stderr

Installation

npx -y @modelcontextprotocol/server-e2b

Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "e2b": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-e2b"],
      "env": {
        "E2B_API_KEY": "YOUR_E2B_API_KEY"
      }
    }
  }
}

Getting an E2B API Key

1. Go to E2B Dashboard
2. Sign up or log in
3. Create an API key
4. Add the key to your MCP configuration

Available Tools

Tool	Description
execute_python	Run Python code in sandbox
execute_node	Run Node.js code in sandbox
upload_file	Upload a file to sandbox
download_file	Download a file from sandbox
list_files	List files in sandbox directory
install_package	Install a Python/Node package

Usage Examples

Run Python Code

result = execute_python(code="""
import pandas as pd
import matplotlib.pyplot as plt

# Create sample data
data = {
    'Month': ['Jan', 'Feb', 'Mar', 'Apr', 'May'],
    'Sales': [12000, 15000, 18000, 22000, 25000]
}

df = pd.DataFrame(data)
print(df.to_string())

# Create visualization
plt.figure(figsize=(10, 6))
plt.plot(df['Month'], df['Sales'], marker='o', linewidth=2)
plt.title('Monthly Sales Trend')
plt.xlabel('Month')
plt.ylabel('Sales ($)')
plt.grid(True, alpha=0.3)
plt.savefig('sales_chart.png')
print('Chart saved to sales_chart.png')
""")
print(result.stdout)
print(result.stderr)

Install and Use Packages

# Install a package
install_package(package="requests", language="python")

# Use the installed package
result = execute_python(code="""
import requests

response = requests.get('https://api.github.com')
print(f"GitHub API status: {response.status_code}")
print(f"Rate limit remaining: {response.headers.get('X-RateLimit-Remaining')}")
""")

Data Analysis Workflow

# Upload a CSV file
upload_file(local_path="sales_data.csv", sandbox_path="/home/user/sales_data.csv")

# Analyze the data
result = execute_python(code="""
import pandas as pd

df = pd.read_csv('/home/user/sales_data.csv')
print(f"Shape: {df.shape}")
print(f"Columns: {df.columns.tolist()}")
print(f"\\nSummary statistics:")
print(df.describe())

# Find top products
top_products = df.nlargest(5, 'revenue')[['product', 'revenue']]
print(f"\\nTop 5 products by revenue:")
print(top_products.to_string())
""")
print(result.stdout)

# Download results
download_file(sandbox_path="analysis_results.csv", local_path="results.csv")

Run Node.js Code

result = execute_node(code="""
const fs = require('fs');

// Process JSON data
const data = {
  users: [
    { id: 1, name: 'Alice', score: 85 },
    { id: 2, name: 'Bob', score: 92 },
    { id: 3, name: 'Charlie', score: 78 }
  ]
};

// Calculate average score
const avgScore = data.users.reduce((sum, u) => sum + u.score, 0) / data.users.length;
console.log(`Average score: ${avgScore.toFixed(2)}`);

// Find top performer
const topPerformer = data.users.reduce((max, u) => u.score > max.score ? u : max);
console.log(`Top performer: ${topPerformer.name} (${topPerformer.score})`);

// Write to file
fs.writeFileSync('results.json', JSON.stringify({ avgScore, topPerformer }, null, 2));
console.log('Results written to results.json');
""");
print(result.stdout)

Pros

• ✅ Secure Sandbox: Isolated execution protects host system
• ✅ Multi-Language: Python, Node.js, and more
• ✅ Package Support: Install packages at runtime
• ✅ File System: Read/write files in sandbox
• ✅ Network Access: Optional internet for API calls
• ✅ Automatic Cleanup: Sandboxes auto-destroy
• ✅ Real-Time Output: Stream execution output

Cons

• ❌ API Key Required: Need E2B account and API key
• ❌ Cost: Free tier has limits, paid plans for heavy use
• ❌ Latency: Cloud execution adds latency vs local
• ❌ Network Limits: Some network access may be restricted

When to Use

Use E2B MCP when:

• You need to execute untrusted or user-provided code
• You want to run data analysis safely
• You need to generate visualizations programmatically
• You want to test code in an isolated environment
• You need multi-language code execution

Avoid E2B MCP when:

• You can execute code locally safely
• You need real-time performance (latency matters)
• You're on a tight budget (free tier limits)
• You need access to local hardware/resources

Resources

• E2B Documentation
• GitHub Repository
• E2B Website