Sandbox
All-in-One Sandbox for AI Agents that combines Browser, Shell, File, MCP and VSCode Server in a single Docker container.
What is Sandbox?
Sandbox is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to all-in-one sandbox for ai agents that combines browser, shell, file, mcp and vscode server in a single docker container.
All-in-One Sandbox for AI Agents that combines Browser, Shell, File, MCP and VSCode Server in a single Docker container.
This server falls under the Cloud Services category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- All-in-One Sandbox for AI Agents that combines Browser, Shel
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx sandboxConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use Sandbox
The AIO Sandbox is an all-in-one containerized environment for AI agents that bundles a full browser, a shell, a file system, a VSCode Server, and an MCP server inside a single Docker container. It exposes browser automation, shell command execution, file read/write, Jupyter notebook execution, and Node.js code evaluation all through a unified REST API and built-in MCP interface. AI agents use it as a secure, isolated execution environment where they can browse the web, run scripts, edit files, and perform end-to-end development tasks without touching the host machine.
Prerequisites
- Docker 20.10+ installed and running on the host machine
- Sufficient system resources: at least 4 GB RAM and 2 CPU cores recommended
- An MCP-compatible client such as Claude Desktop or a custom agent
- Python 3.9+ or Node.js 18+ if using the SDK (optional; the REST API works without an SDK)
Pull and run the sandbox Docker container
Run the all-in-one sandbox container, exposing port 8080 for the REST API, documentation, VSCode, and browser VNC. The --security-opt flag is required for browser sandbox processes inside Docker.
docker run --security-opt seccomp=unconfined --rm -it -p 8080:8080 ghcr.io/agent-infra/sandbox:latestVerify the sandbox is running
Open the built-in API documentation in your browser to confirm the container is healthy and the API is accessible.
open http://localhost:8080/v1/docsInstall the Python SDK (optional)
For Python-based agents, install the agent-sandbox SDK to interact with the sandbox programmatically.
pip install agent-sandboxExecute a shell command via the SDK
Connect the Python SDK to the running container and run a shell command, read a file, or capture a browser screenshot.
from agent_sandbox import Sandbox
client = Sandbox(base_url="http://localhost:8080")
result = client.shell.exec_command(command="ls -la /home/gem")
content = client.file.read_file(file="/home/gem/.bashrc")
screenshot = client.browser.screenshot()Configure your MCP client to use the sandbox
Add the sandbox MCP server endpoint to your MCP client configuration. The sandbox exposes MCP-compatible tools for browser, shell, file, and document operations.
{
"mcpServers": {
"sandbox": {
"url": "http://localhost:8080/mcp",
"transportType": "streamable-http"
}
}
}Configure environment variables (optional)
Pass environment variables to the Docker container to customize behaviour such as proxy settings, authentication, and the workspace directory.
docker run --security-opt seccomp=unconfined --rm -it -p 8080:8080 \
-e PROXY_SERVER=host.docker.internal:7890 \
-e WORKSPACE=/home/gem/project \
-e TZ=America/New_York \
ghcr.io/agent-infra/sandbox:latestSandbox Examples
Client configuration
Connect Claude Desktop to the locally running AIO Sandbox via its MCP HTTP endpoint.
{
"mcpServers": {
"sandbox": {
"url": "http://localhost:8080/mcp",
"transportType": "streamable-http"
}
}
}Prompts to try
Use the sandbox to have your AI agent perform real browser, shell, and file operations.
- "Open the browser and take a screenshot of https://example.com"
- "Run 'pip install requests && python3 -c "import requests; print(requests.__version__)"' in the shell"
- "Read the file at /home/gem/.bashrc and show me its contents"
- "Execute this Python script in Jupyter and show the output: import pandas as pd; df = pd.DataFrame({'a': [1,2,3]}); print(df)"Troubleshooting Sandbox
Browser automation fails with a sandbox security error inside Docker
Ensure you are passing --security-opt seccomp=unconfined to the docker run command. Without this flag, the browser process cannot create its own sandbox inside the container.
Port 8080 is already in use when starting the container
Map to a different host port with -p 9090:8080 and update your MCP client config and SDK base_url to http://localhost:9090 accordingly.
JWT_PUBLIC_KEY authentication errors when calling the API
If you have deployed the sandbox with a JWT_PUBLIC_KEY environment variable for security, include a valid signed Bearer token in the Authorization header of all API requests or MCP client configuration.
Frequently Asked Questions about Sandbox
What is Sandbox?
Sandbox is a Model Context Protocol (MCP) server that all-in-one sandbox for ai agents that combines browser, shell, file, mcp and vscode server in a single docker container. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install Sandbox?
Follow the installation instructions on the Sandbox GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with Sandbox?
Sandbox works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is Sandbox free to use?
Yes, Sandbox is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.
Sandbox Alternatives — Similar Cloud Services Servers
Looking for alternatives to Sandbox? Here are other popular cloud services servers you can use with Claude, Cursor, and VS Code.
Open WebUI
★ 138.2kUser-friendly AI Interface (Supports Ollama, OpenAI API, ...)
Anything LLM
★ 60.4kThe all-in-one AI productivity accelerator. On device and privacy first with no annoying setup or configuration.
LocalAI
★ 46.4kLocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.
Nacos
★ 33.0kan easy-to-use dynamic service discovery, configuration and service management platform for building AI cloud native applications.
Xiaozhi ESP32
★ 26.7k本项目为xiaozhi-esp32提供后端服务,帮助您快速搭建ESP32设备控制服务器。Backend service for xiaozhi-esp32, helps you quickly build an ESP32 device control server.
Gateway
★ 11.8kA blazing fast AI Gateway with integrated guardrails. Route to 1,600+ LLMs, 50+ AI Guardrails with 1 fast & friendly API.
Browse More Cloud Services MCP Servers
Explore all cloud services servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up Sandbox in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use Sandbox?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.