AIO Sandbox - All-in-One Agent Sandbox Environment

🌐 Browser | 💻 Terminal | 📁 File | 🔧 VSCode | 📊 Jupyter | 🤖 MCP

🌐 Website   |   🔌 API   |   📑 Paper   |   🌟 Examples   |   📊 Evaluation   

🚀 Quick Start

Get up and running in 30 seconds:

# Recommended: Enable API Key authentication (protects all services: API, JupyterLab, VNC)
# - Supports three methods: X-AIO-API-Key header, Authorization: Bearer header, ?api_key= query parameter
# - Without SANDBOX_API_KEY, services remain open (backward compatible)
docker run --security-opt seccomp=unconfined --rm -it \
  -e SANDBOX_API_KEY=your-secret-key \
  -p 127.0.0.1:8080:8080 ghcr.io/agent-infra/sandbox:latest

For users in mainland China:

docker run --security-opt seccomp=unconfined --rm -it \
  -e SANDBOX_API_KEY=your-secret-key \
  -p 127.0.0.1:8080:8080 enterprise-public-cn-beijing.cr.volces.com/vefaas-public/all-in-one-sandbox:1.11.0

For reproducible deployments, pin a release tag. Replace 1.11.0 with the release you want:

docker run --security-opt seccomp=unconfined --rm -it \
  -p 127.0.0.1:8080:8080 ghcr.io/agent-infra/sandbox:1.11.0
# or use the pinned mainland China mirror
docker run --security-opt seccomp=unconfined --rm -it \
  -p 127.0.0.1:8080:8080 enterprise-public-cn-beijing.cr.volces.com/vefaas-public/all-in-one-sandbox:1.11.0

These examples intentionally bind the host side to 127.0.0.1 because the sandbox listens on 0.0.0.0 inside the container. For cloud deployment, keep port 8080 private and publish it through a reverse proxy or Ingress: Cloud Deployment Guide.

Once running, access the environment at:

• 📖 Documentation: http://localhost:8080/v1/docs
• 🌐 VNC Browser: http://localhost:8080/vnc/index.html?autoconnect=true
• 💻 VSCode Server: http://localhost:8080/code-server/
• 🤖 MCP Services: http://localhost:8080/mcp

🎯 What is AIO Sandbox?

AIO Sandbox is an all-in-one agent sandbox environment that combines Browser, Shell, File, MCP operations, and VSCode Server in a single Docker container. Built on cloud-native lightweight sandbox technology, it provides a unified, secure execution environment for AI agents and developers.

Why Choose AIO Sandbox?

Traditional sandboxes are single-purpose (browser, code, or shell), making file sharing and functional coordination extremely challenging. AIO Sandbox solves this by providing:

• ✅ Unified File System - Files downloaded in browser are instantly available in Shell/File operations
• ✅ Multiple Interfaces - VNC, VSCode, Jupyter, and Terminal in one unified environment
• ✅ Secure Execution - Sandboxed Python and Node.js execution with safety guarantees
• ✅ Zero Configuration - Pre-configured MCP servers and development tools ready to use
• ✅ Agent-Ready - MCP-compatible APIs for seamless AI agent integration

📦 Installation

SDK Installation

pip install agent-sandbox

npm install @agent-infra/sandbox

go get github.com/agent-infra/sandbox-sdk-go

Basic Usage

from agent_sandbox import Sandbox

# Initialize client
client = Sandbox(base_url="http://localhost:8080")
home_dir = client.sandbox.get_context().home_dir

# Execute shell commands
result = client.shell.exec_command(command="ls -la")
print(result.data.output)

# File operations
content = client.file.read_file(file=f"{home_dir}/.bashrc")
print(content.data.content)

# Browser automation
screenshot = client.browser.screenshot()

import { Sandbox } from '@agent-infra/sandbox';

// Initialize client
const sandbox = new Sandbox({ baseURL: 'http://localhost:8080' });

// Execute shell commands
const result = await sandbox.shell.exec({ command: 'ls -la' });
console.log(result.output);

// File operations
const content = await sandbox.file.read({ path: '/home/gem/.bashrc' });
console.log(content);

// Browser automation
const screenshot = await sandbox.browser.screenshot();

🌟 Key Features

🔗 Unified Environment

All components run in the same container with a shared filesystem, enabling seamless workflows:

🌐 Browser Automation

Full browser control through multiple interfaces:

• VNC - Visual browser interaction through remote desktop
• CDP - Chrome DevTools Protocol for programmatic control
• MCP - High-level browser automation tools

💻 Development Tools

Integrated development environment with:

• VSCode Server - Full IDE experience in browser
• Jupyter Notebook - Interactive Python environment
• Terminal - WebSocket-based terminal access
• Port Forwarding - Smart preview for web applications

🤖 MCP Integration

Pre-configured Model Context Protocol servers:

• Browser - Web automation and scraping
• File - File system operations
• Shell - Command execution
• Markitdown - Document processing

📚 Complete Example

Convert a webpage to Markdown with embedded screenshot:

import asyncio
import base64
from playwright.async_api import async_playwright
from agent_sandbox import Sandbox

async def site_to_markdown():
    # Initialize sandbox client
    c = Sandbox(base_url="http://localhost:8080")
    home_dir = c.sandbox.get_context().home_dir

    # Browser: Automation to download HTML
    async with async_playwright() as p:
        browser_info = c.browser.get_info().data
        page = await (await p.chromium.connect_over_cdp(browser_info.cdp_url)).new_page()
        await page.goto("https://example.com", wait_until="networkidle")
        html = await page.content()
        screenshot_b64 = base64.b64encode(await page.screenshot()).decode('utf-8')

    # Jupyter: Convert HTML to markdown in sandbox
    c.jupyter.execute_code(code=f"""
from markdownify import markdownify
html = '''{html}'''
screenshot_b64 = "{screenshot_b64}"

md = f"{{markdownify(html)}}\\n\\n![Screenshot](data:image/png;base64,{{screenshot_b64}})"
with open('{home_dir}/site.md', 'w') as f:
    f.write(md)
print("Done!")
""")

    # Shell: List files in sandbox
    list_result = c.shell.exec_command(command=f"ls -lh {home_dir}")
    print(f"Files in sandbox: {list_result.data.output}")

    # File: Read the generated markdown
    return c.file.read_file(file=f"{home_dir}/site.md").data.content

if __name__ == "__main__":
    result = asyncio.run(site_to_markdown())
    print(f"Markdown saved successfully!")

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                    🌐 Browser + VNC                        │
├─────────────────────────────────────────────────────────────┤
│  💻 VSCode Server  │  🐚 Shell Terminal  │  📁 File Ops   │
├─────────────────────────────────────────────────────────────┤
│              🔗 MCP Hub + 🔒 Sandbox Fusion               │
├─────────────────────────────────────────────────────────────┤
│         🚀 Preview Proxy + 📊 Service Monitoring          │
└─────────────────────────────────────────────────────────────┘

🛠️ API Reference

Core APIs

MCP Servers

🚢 Deployment

Docker Compose

services:
  sandbox:
    container_name: aio-sandbox
    image: ghcr.io/agent-infra/sandbox:latest
    security_opt:
      - seccomp:unconfined
    ports:
      - "127.0.0.1:${HOST_PORT:-8080}:8080"
    volumes:
      - sandbox_data:/home/gem/workspace
    extra_hosts:
      - "host.docker.internal:host-gateway"
    restart: "unless-stopped"
    shm_size: "2gb"
    environment:
      SANDBOX_API_KEY: ${SANDBOX_API_KEY:-}
      PROXY_SERVER: ${PROXY_SERVER:-}
      WORKSPACE: ${WORKSPACE:-/home/gem/workspace}
      TZ: ${TZ:-Asia/Singapore}

volumes:
  sandbox_data:

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata:
  name: aio-sandbox
spec:
  replicas: 2
  selector:
    matchLabels:
      app: aio-sandbox
  template:
    metadata:
      labels:
        app: aio-sandbox
    spec:
      containers:
      - name: aio-sandbox
        image: ghcr.io/agent-infra/sandbox:latest
        ports:
        - containerPort: 8080
        resources:
          limits:
            memory: "2Gi"
            cpu: "1000m"

🤝 Integration Examples

Browser Use Integration

import asyncio

from agent_sandbox import Sandbox
from browser_use import Agent, Tools
from browser_use.browser import BrowserProfile, BrowserSession
from browser_use.llm import ChatOpenAI

sandbox = Sandbox(base_url="http://localhost:8080")
print("sandbox", sandbox.browser)
cdp_url = sandbox.browser.get_info().data.cdp_url

browser_session = BrowserSession(
    browser_profile=BrowserProfile(cdp_url=cdp_url, is_local=True)
)
tools = Tools()


async def main():
    agent = Agent(
        task='Visit https://duckduckgo.com and search for "browser-use founders"',
        llm=ChatOpenAI(model="gcp-claude4.1-opus"),
        tools=tools,
        browser_session=browser_session,
    )

    await agent.run()
    await browser_session.kill()

    input("Press Enter to close...")


if __name__ == "__main__":
    asyncio.run(main())

LangChain Integration

from langchain.tools import BaseTool
from agent_sandbox import Sandbox

class SandboxTool(BaseTool):
    name = "sandbox_execute"
    description = "Execute commands in AIO Sandbox"

    def _run(self, command: str) -> str:
        client = Sandbox(base_url="http://localhost:8080")
        result = client.shell.exec_command(command=command)
        return result.data.output

OpenAI Assistant Integration

from openai import OpenAI
from agent_sandbox import Sandbox
import json

client = OpenAI(
    api_key="your_api_key",
)
sandbox = Sandbox(base_url="http://localhost:8080")


# define a tool to run code in the sandbox
def run_code(code, lang="python"):
    if lang == "python":
        return sandbox.jupyter.execute_code(code=code).data
    return sandbox.nodejs.execute_nodejs_code(code=code).data


# Use OpenAI
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "calculate 1+1"}],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "run_code",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "code": {"type": "string"},
                        "lang": {"type": "string"},
                    },
                },
            },
        }
    ],
)


if response.choices[0].message.tool_calls:
    args = json.loads(response.choices[0].message.tool_calls[0].function.arguments)
    print("args", args)
    result = run_code(**args)
    print(result['outputs'][0]['text'])

MiniMax Integration

MiniMax provides an OpenAI-compatible API, so you can use the same openai SDK with a different base_url:

from openai import OpenAI
from agent_sandbox import Sandbox
import json

client = OpenAI(
    api_key="your_minimax_api_key",
    base_url="https://api.minimax.io/v1",
)
sandbox = Sandbox(base_url="http://localhost:8080")


def run_code(code, lang="python"):
    if lang == "python":
        return sandbox.jupyter.execute_code(code=code).data
    return sandbox.nodejs.execute_code(code=code).data


response = client.chat.completions.create(
    model="MiniMax-M2.7",
    messages=[{"role": "user", "content": "calculate 1+1"}],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "run_code",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "code": {"type": "string"},
                        "lang": {"type": "string"},
                    },
                },
            },
        }
    ],
    temperature=0.01,  # MiniMax requires temperature > 0
)


if response.choices[0].message.tool_calls:
    args = json.loads(response.choices[0].message.tool_calls[0].function.arguments)
    result = run_code(**args)
    print(result.outputs[0].text)

See the full minimax-integration example for more details.

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📄 License

AIO Sandbox is released under the Apache License 2.0.

🙏 Acknowledgments

Built with ❤️ by the Agent Infra team. Special thanks to all contributors and the open-source community.

📞 Support

• 📖 Documentation
• 💬 GitHub Discussions
• 🐛 Issue Tracker

Ready to revolutionize your AI development workflow?
⭐ Star us on GitHub • 📚 Read the Docs • 🐛 Report Issues