MCP Sandbox

MCP Sandbox

Automatically converts JavaScript modules into MCP-compatible servers, making any JavaScript function accessible to AI systems through secure sandboxing with automatic type inference.

Category
Visit Server

README

MCP Sandbox

npm version License: MIT

<p align=center> <img width="80%" src="https://raw.githubusercontent.com/danstarns/mcp-sandbox/main/docs/banner.png#gh-dark-mode-only" alt="MCP Sandbox"/> </p>

Turn any JavaScript module into a sandboxed MCP (Model Context Protocol) server with automatic reflection and type inference.

🎯 What is MCP Sandbox?

MCP Sandbox automatically converts JavaScript modules into MCP (Model Context Protocol) compatible servers, making any JavaScript function accessible to AI systems. It uses VM sandboxing for security, automatic type inference, and generates proper MCP configurations.

✨ Features

  • 🔍 Automatic Reflection - Analyzes JS modules and extracts function signatures
  • 🛡️ Secure Sandboxing - Executes code in isolated VM contexts with timeouts
  • 🧠 Smart Type Inference - Detects parameter types from defaults and naming patterns
  • 📚 JSDoc Integration - Extracts documentation from function comments
  • 📡 MCP Protocol - Full JSON-RPC 2.0 and SSE support
  • 🌐 REST API - Legacy REST endpoints for easy testing
  • ⚙️ TypeScript - Full type safety and IntelliSense support

🚀 Quick Start

Installation

# Install globally for CLI usage
npm install -g @mcp-sandbox/cli

# Or use in a project
npm install @mcp-sandbox/core @mcp-sandbox/cli

Basic Usage

# Start MCP server for a JavaScript module
$ mcp-sandbox start ./math-utils.js

🏗️  Initializing MCP Sandbox...
🔍 Reflecting module
📊 Discovered 2 tools:
  - circleArea: Calculate area of a circle
  - fibonacci: Generate Fibonacci sequence
🚀 MCP Sandbox server running at http://localhost:3000
📋 MCP Tools: http://localhost:3000/mcp/tools
⚡ MCP Execute: http://localhost:3000/mcp/execute
🔄 MCP SSE: http://localhost:3000/sse
📡 MCP JSON-RPC: http://localhost:3000/mcp/jsonrpc
⚙️  MCP Config: http://localhost:3000/mcp-config
💡 For MCP Inspector, use: http://localhost:3000/sse

Example Module

/**
 * Calculate the area of a circle
 * @param radius The radius of the circle
 */
function circleArea(radius = 1) {
  return Math.PI * radius * radius;
}

/**
 * Generate fibonacci sequence
 * @param count Number of fibonacci numbers to generate
 */
function fibonacci(count = 10) {
  const seq = [0, 1];
  for (let i = 2; i < count; i++) {
    seq[i] = seq[i - 1] + seq[i - 2];
  }
  return seq.slice(0, count);
}

module.exports = { circleArea, fibonacci };

Running mcp-sandbox start math-utils.js automatically:

  1. 🔍 Reflects the module and discovers functions
  2. 📊 Generates type schemas from parameters
  3. 🚀 Starts MCP server at http://localhost:3000
  4. 💾 Creates mcp-config.json for MCP clients

📡 API Endpoints

The server exposes both MCP and REST endpoints:

MCP Protocol (JSON-RPC 2.0)

  • POST /mcp/jsonrpc - Main MCP endpoint
  • GET /sse - Server-Sent Events for real-time updates

REST API (for testing)

  • GET /tools - List available tools
  • POST /execute/:toolName - Execute a specific tool
  • GET /mcp-config - Get MCP server configuration
  • GET /health - Health check

Example Usage

# List tools
curl http://localhost:3000/tools

# Execute function via REST
curl -X POST http://localhost:3000/execute/circleArea \
  -H "Content-Type: application/json" \
  -d '{"args": {"radius": 5}}'

# MCP JSON-RPC call
curl -X POST http://localhost:3000/mcp/jsonrpc \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "fibonacci", "arguments": {"count": 8}}}'

🏗️ Programmatic Usage

import { MCPSandbox } from '@mcp-sandbox/core';

const sandbox = new MCPSandbox({
  port: 3000,
  timeout: 5000,
});

// Load and analyze module
await sandbox.loadModule('./my-module.js');

// Start MCP server
await sandbox.start();

// Execute tools directly
const result = await sandbox.executeTool('myFunction', {
  param1: 'value1',
});

📚 Example Modules Included

The repository includes several example modules demonstrating different use cases:

Mathematical Operations (examples/math-utils.js)

  • Circle area calculation
  • Fibonacci sequence generation
  • Compound interest calculation
  • Prime number checking
  • Degree/radian conversion
  • Factorial calculation

String Manipulation (examples/string-utils.js)

  • Title case conversion
  • Random string generation
  • Word counting
  • Palindrome detection
  • String reversal
  • Capitalization

Array Operations (examples/array-utils.js)

  • Array shuffling (Fisher-Yates)
  • Unique value extraction
  • Array chunking
  • Set operations (intersection, difference)
  • Array flattening

Filesystem Operations (examples/filesystem-utils.js)

  • File reading/writing (async)
  • Directory listing and creation
  • File searching with patterns
  • Disk usage calculation
  • File copying and deletion
  • Line-by-line file reading

🛠️ Development

# Clone repository
git clone https://github.com/danstarns/mcp-sandbox.git
cd mcp-sandbox

# Install dependencies
pnpm install

# Build all packages
pnpm build

# Run examples
pnpm example:math          # Math utilities
pnpm example:filesystem    # File operations
pnpm example:string        # String manipulation

# Lint and format
pnpm lint && pnpm format

🔧 Configuration Options

CLI Options

mcp-sandbox start <module> [options]

Options:
  -p, --port <port>      Server port (default: 3000)
  -h, --host <host>      Server host (default: localhost)
  -t, --timeout <ms>     Execution timeout (default: 5000ms)
  -o, --output <file>    Output MCP configuration to file

Programmatic Options

interface SandboxOptions {
  port?: number; // Server port (default: 3000)
  host?: string; // Server host (default: 'localhost')
  timeout?: number; // Execution timeout (default: 5000ms)
  maxMemory?: number; // Memory limit (default: 64MB)
}

🔒 Security Features

  • VM Isolation - Code runs in separate V8 contexts
  • Execution Timeouts - Configurable time limits prevent infinite loops
  • Memory Limits - Prevent memory exhaustion attacks
  • Controlled Requires - Limited module access in sandbox
  • Input Validation - Parameter type checking and validation

🎮 Testing with MCP Inspector

  1. Start your MCP server: mcp-sandbox start examples/math-utils.js
  2. Open MCP Inspector
  3. Set Transport Type to "Streamable HTTP"
  4. Enter URL: http://localhost:3000/mcp/jsonrpc
  5. Connect and test your tools!

📦 Packages

This is a monorepo containing multiple packages:

🤝 Contributing

Contributions are welcome! Please read our Contributing Guide and check out the open issues.

📄 License

MIT License - see LICENSE for details.

🔗 Links

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured