AgentSandbox MCP

Tencent Cloud AI Agent Sandbox Service - MCP Protocol Implementation

Features • Structure • Quick Start • Guide • Contributing • License

---

Introduction

Agent Sandbox (AGS) is Tencent Cloud's next-generation sandbox execution environment for AI agents, featuring secure isolation, extreme performance, and scalability. It supports multiple sandbox types including browser, code, and custom sandboxes, enabling agents to run in controlled environments and interact safely with the digital world.

This project provides an open-source implementation based on the Model Context Protocol (MCP), including browser sandbox and code sandbox capabilities, supporting seamless integration with any MCP-compatible AI client.

Key Features

🔒 Secure Isolation

• Fully isolated sandbox execution environment
• Prevents malicious code from affecting the host system
• Resource usage limits and monitoring

⚡ Extreme Performance

• Millisecond-level sandbox startup
• High concurrency support, single machine can host numerous sandbox instances
• Intelligent resource scheduling and optimization

🔧 Extensible Architecture

• Supports multiple sandbox types (browser, code, custom)
• Based on standard MCP protocol, easy to integrate
• Flexible configuration and extension mechanisms

🌐 Browser MCP - Browser Sandbox

Playwright-based browser automation MCP server (TypeScript implementation)

• ✅ Fast and lightweight headless browser control
• ✅ Based on structured accessibility snapshots, no vision models needed
• ✅ Supports page navigation, element interaction, form filling
• ✅ Screenshots, PDF generation, JavaScript execution
• ✅ Network monitoring, console capture, tab management
• ✅ Optional browser extension support

🐍 Code MCP - Code Sandbox

Python code execution MCP server (Python implementation)

• ✅ Secure code sandbox execution environment
• ✅ Supports real-time Python code execution
• ✅ File system access and management
• ✅ Fully isolated runtime environment
• ✅ Supports custom dependency installation

Project Structure

agentsandbox-mcp/
├── browser-mcp/                 # TypeScript - Browser MCP Server
│   ├── src/                     # Source code
│   ├── tests/                   # Test files
│   ├── extension/               # Browser extension
│   ├── examples/                # Usage examples
│   ├── package.json             # NPM configuration
│   ├── cli.js                   # CLI entry point
│   └── README.md                # Documentation
├── code-mcp/                    # Python - Code MCP Server
│   ├── src/                     # Source code
│   │   └── server.py           # MCP server implementation
│   ├── main.py                  # Entry point
│   ├── pyproject.toml           # Python project configuration
│   └── README.md                # Documentation
├── CODE_OF_CONDUCT.md           # Code of Conduct
├── CONTRIBUTING.md              # Contributing Guide
├── LICENSE                      # Apache 2.0 License
└── README.md                    # Project README (this file)

Quick Start

Requirements

• Browser MCP: Node.js 18+
• Code MCP: Python 3.8+ and uv

Installation and Usage

Browser MCP Server

cd browser-mcp

# Install dependencies
npm install

# Run in development mode
npm run watch

# Or use npx directly
npx @tencent/agentsandbox-browser-mcp

Configure in MCP client:

{
  "mcpServers": {
    "agentsandbox-browser": {
      "command": "npx",
      "args": ["@tencent/agentsandbox-browser-mcp"]
    }
  }
}

Supported MCP clients: Claude Desktop, VS Code, Cursor, Windsurf, Goose, LM Studio, etc.

For detailed documentation, see browser-mcp/README.md

Code MCP Server

cd code-mcp

# Install dependencies with uv
uv sync

# Configure environment variable (required)
export E2B_API_KEY="your-e2b-api-key"

# Run server
python main.py

Get API Key: Visit Tencent Cloud Agent Sandbox to register and obtain your API key.

For detailed documentation, see code-mcp/README.md

Usage Guide

Use Cases

🔍 Intelligent Web Analysis

Using browser sandbox, AI agents can:

• Automated information scraping and data collection
• Automatic form filling and submission
• Web content understanding and information extraction
• Automated testing and monitoring

📊 Data Analysis Processing

Using code sandbox, AI agents can:

• Execute complex data analysis tasks
• Generate visualization charts and reports
• File processing and format conversion
• Scientific computing and algorithm validation

🤖 Agent Collaboration

Combining both sandboxes enables:

• Web data scraping + data analysis processing
• Dynamic content generation + browser preview
• End-to-end automated workflows

Browser MCP Tools

Browser MCP provides the following MCP tools:

• Navigation: playwright_navigate, playwright_screenshot
• Element Interaction: playwright_click, playwright_fill, playwright_select, playwright_hover
• JavaScript Execution: playwright_evaluate
• Page Management: playwright_snapshot, playwright_new_page, playwright_close_page
• File Operations: playwright_upload_files, playwright_download_files
• Dialog Handling: playwright_accept_dialog, playwright_dismiss_dialog

Code MCP Features

Code MCP provides the following capabilities:

• Code Execution: Safely execute Python code in fully isolated sandboxes
• File System: Support file read/write, directory management, and other file system operations
• Environment Persistence: Sandbox environment can be persisted, supporting multiple interactions
• Dependency Management: Support dynamic installation and use of third-party Python packages

Technical Architecture

Core Tech Stack

Module	Language	Main Dependencies	Sandbox Type
Browser MCP	TypeScript	Playwright 1.56, @modelcontextprotocol/sdk, commander, zod	Browser Sandbox
Code MCP	Python	@e2b/code-interpreter 2.0, @modelcontextprotocol/sdk	Code Sandbox

AGS Core Features

• Standard MCP Protocol: Fully compatible with Model Context Protocol specification, ensuring seamless integration with mainstream AI clients
• Independent Deployment: Two MCP servers can be used independently or combined, flexibly meeting different scenario needs
• Secure Isolation: Both browser and code execution run in strictly isolated sandbox environments, ensuring system security
• High Performance: Uses lightweight container technology, achieving millisecond-level startup and high concurrency support
• Easy to Extend: Open architecture design, supporting custom sandbox types and tool extensions

Sandbox Types

Tencent Cloud AGS supports multiple sandbox types:

1. Browser Sandbox: Enables AI agents to browse web pages, fill forms, and extract information like humans
2. Code Sandbox: Provides AI agents with secure code execution environment, supporting data analysis, file processing, and other tasks
3. Custom Sandbox: Supports customizing specific sandbox environments based on business needs (planned)

Documentation

• Browser MCP Documentation
• Code MCP Documentation
• Contributing Guide
• Code of Conduct

Contributing

We welcome contributions of all kinds! Whether it's bug reports, feature suggestions, documentation improvements, or code contributions.

How to Contribute

1. Fork this project
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

For detailed guidelines, see CONTRIBUTING.md

Code of Conduct

Please note that this project follows the Contributor Covenant Code of Conduct. By participating in this project, you agree to abide by its terms.

License

This project is open-sourced under the Apache License 2.0. See LICENSE file for details.

Copyright 2024-2025 AgentSandbox MCP Contributors

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Related Links

• Tencent Cloud AGS Official Documentation - Learn more about AGS service details
• Model Context Protocol - MCP official protocol specification
• Playwright - Browser automation framework

Contact

• Submit Issue: GitHub Issues
• Tencent Cloud Official Support: Contact Us

About Tencent Cloud AGS

Agent Sandbox (AGS) is Tencent Cloud's enterprise-grade AI agent sandbox service, providing developers with:

• ☁️ Cloud Deployment: No need to build infrastructure, ready to use out of the box
• 🔐 Enterprise Security: Meets enterprise security and compliance requirements
• 📈 Elastic Scaling: Automatically scales based on business needs
• 💼 Professional Support: Provides 24/7 technical support services

This open-source project provides a localized implementation based on the MCP protocol, helping developers quickly experience and integrate AGS capabilities.

---

⭐️ If this project helps you, please give us a Star!