Gbox

gbox is an open-source self-hosted sandbox project that provides a secure execution environment for AI agents, supports various tools such as terminals, browsers, and file operations, and is compatible with the MCP protocol.

Developer tools Virtualization #AI Sandbox #MCP Integration #Self-Hosted #Multi-Tool Local .Go

rating : 2 points

downloads : 7.7K

update time : 2025-04-28

Open Site

What is the MCP Server?

The MCP Server is an open-source project that provides a self-hosted sandbox environment for MCP (Model Context Protocol) integration or other AI agent application scenarios. It allows users to run AI clients locally, such as Claude Desktop or Cursor.

How to use the MCP Server?

First, install the MCP Server, then configure the MCP client to connect to the server. Finally, perform various tasks through simple command-line operations.

Applicable Scenarios

The MCP Server is very suitable for AI task execution that requires localization and security, such as generating charts, PDF files, data analysis, and processing local files.

Main Features

Terminal

You can run any Linux command in the terminal and even directly execute Python scripts.

File Management

You can mount folders on the host to the sandbox, access files in the sandbox, and share them via HTTP links.

Browser

Open any URL, return multi-modal content, and support file downloading and web page operations.

HTTP Server

Start an HTTP service on demand for convenient file sharing.

SDK Support

A Python SDK is provided and can be installed via pip; the Typescript SDK is under development.

MCP Integration

Supports the standard MCP protocol and is compatible with Claude Desktop and Cursor.

Advantages

Supports localization and security

Provides integration of multiple functions

Easy to configure and use

Limitations

Limited to Linux systems

Some advanced functions are still under development

How to Use

Install the MCP Server

Install the MCP Server using Homebrew and initialize the environment.

Configure the MCP Client

Export the MCP configuration and merge it into Claude Desktop or Cursor.

Run Tasks

Create, start, and stop containers via the command line to execute tasks.

Usage Examples

Generate a chart of Tesla's stock prices

Use the MCP Server to generate a chart of Tesla's stock prices.

Analyze the market capitalization comparison between Nvidia and Tesla

Compare the market capitalization data of Nvidia and Tesla.

Frequently Asked Questions

Which operating systems does the MCP Server support?

How to update the MCP Server?

Related Resources

MCP Server GitHub Repository

The open-source code repository of the MCP Server

PyPI Page

The installation page for the Python SDK

🚀 Gbox Usage Documentation

Gbox is a comprehensive toolset that provides a range of services. This documentation will guide you through the installation, usage, and contribution process of Gbox.

🚀 Quick Start

Prerequisites

Go 1.21 or higher
Docker Desktop
Make tool
pnpm (via corepack)
Node.js 16.13 or higher

Build

# Build all components
make build

# Create a distribution package
make dist

Run the Service

# API Server
make -C packages/api-server dev

# MCP Server
cd packages/mcp-server && pnpm dev

# MCP Inspector
cd packages/mcp-server && pnpm inspect

✨ Features

Gbox offers multiple services including API Server, MCP Server, and MCP Inspector. It provides a convenient way to manage and interact with different components in a development environment.

📦 Installation

The installation of Gbox involves several steps including setting up the prerequisites, building the components, and running the services. Please follow the steps in the "Quick Start" section to install and run Gbox.

📚 Documentation

Contribution Guidelines

We welcome your contributions! Please feel free to submit a Pull Request. For major changes, please open an issue for discussion first.

Fork the repository.
Create a feature branch (git checkout -b username/feature-name).
Commit your changes (git commit -m 'Add some feature').
Push to the branch (git push origin username/feature-name).
Open a Pull Request.

Notes (Development and Debugging in Local Environment)

How to run gbox in the development environment instead of the system-installed version

Stop the installed gbox via gbox cleanup so that the API server can run in the development environment.
Execute make api-dev in the project root directory.
Run ./gbox box list, which is a command running from your development environment.

How to connect the MCP client (e.g., Claude Desktop) to the MCP server in the development environment

Execute make mcp-dev in the project root directory.
Run ./gbox mcp export --merge-to claude.

How to open the MCP inspection window

Execute make mcp-inspect in the project root directory.
Click the link returned by the terminal.

How to build and use images in the development environment

Execute make build-image-python in the project root directory to build the Python image, or make build-images to build all images.
Change the image name as needed (e.g., execute make build-image-typescript to get the TypeScript image).
You may need to delete the current sandbox for the new image to take effect: ./gbox box delete --all.

Why does the MCP client still display the old MCP content?

If you have modified the MCP configuration (e.g., tool definitions), you need to run make build to update the dist/index.js file.
You may also need to execute ./gbox mcp export --merge-to claude.