Gru-sandbox(gbox) is an open source project that provides a self-hostable sandbox for MCP integration or other AI agent usecases.
gbox is a self-hostable sandbox for AI Agents to execute commands, surf web and use desktop/mobile. See "Features" section for details.
This project is based on the technology behind gru.ai. It has been tested over 100000 Agent jobs.
As MCP is getting more and more popular, we also implemented a MCP server to make it easy to be directly integrated into MCP client such as Claude Desktop/Cursor.
npm install gbox-sdk
, see Gbox SDK Reference for details.Note: Support for other platforms (Linux, Windows) is coming soon.
# Install via Homebrew
brew tap babelcloud/gru && brew install gbox
# Initialize environment
gbox setup
# Export MCP config and merge into Claude Desktop
gbox mcp export --merge-to claude
# or gbox mcp export --merge-to cursor
# Restart Claude Desktop
# Update gbox to the latest version
brew update && brew upgrade gbox
# Update the environment
gbox setup
# Export and merge latest MCP config into Claude Desktop
gbox mcp export --merge-to claude
# or gbox mcp export --merge-to cursor
# Restart Claude Desktop
The project provides a command-line tool gbox
for managing sandbox containers:
# Login to gbox.cloud
gbox login
# Profile management
gbox profile add --key <key> --name <name> # add profile
gbox profile list # list all profiles
gbox profile use [index] # switch current profile (interactive if index omitted)
gbox profile delete <index> # delete profile
gbox profile current # show current profile
# Local environment
gbox setup # initialize local runtime environment (alias: cluster setup)
gbox cleanup # clean up local runtime environment (alias: cluster cleanup)
# Container (Box) management
gbox box create --image python:3.9 --env DEBUG=true -- python -c "print('hello')" # create a box
gbox box list # list boxes
gbox box terminate <box-id> # terminate box
gbox box exec <box-id> -- ls / # execute command inside box
gbox box cp <box-id>:<container-path> <local-path> # file copy
gbox box inspect <box-id> # inspect box
# Android CUA (requires OPENAI_API_KEY)
gbox cua android "Open Uber and order a ride to CUHK"
# MCP configuration
gbox mcp export # export config only
gbox mcp export --merge-to claude # integrate linux mcp server into Claude Desktop
gbox mcp export --merge-to cursor # integrate linux mcp server into Cursor
gbox mcp export --merge-to claude-code --type android # integrate android mcp server into Claude Code
gbox mcp export --dry-run # preview merge result
ADB (Android Debug Bridge) is required for gbox to communicate with your Android device:
Install ADB:
brew install android-platform-tools
sudo apt install adb
(Ubuntu/Debian) or sudo pacman -S android-tools
(Arch)Add ADB to your PATH:
export PATH=$PATH:/path/to/platform-tools
Verify ADB installation:
adb version
Enable USB debugging on your Android device:
Connect Android device via USB and confirm ADB connected
adb devices -l
List of devices attached
897X0691U Pixel_3 device
Set the OPENAI_API_KEY environment variable and enjoy gbox Android CUA
export OPENAI_API_KEY=YOUR_KEY
gbox cua android "Open Uber and book a ride to The Chinese University of Hong Kong now."
The video demonstration for this task can be found in the first use case below under "Use Cases" section.
Prompt: Open Uber and book a ride to The Chinese University of Hong Kong now.
Your AI client such as Claude Desktop can use gbox MCP to deliver better results, such as
Generate diagrams of Tesla stock prices:
https://claude.ai/share/34de8ca3-4e04-441b-9e79-5875fa9fc97a
Generate PDF of latest AI news:
https://claude.ai/share/84600933-dcf2-44be-a2fd-7f49540db57a
Analyze and compare Nvidia/Tesla market cap:
https://claude.ai/share/70c335b7-9fff-4ee7-8459-e6b7462d8994
Download youtube video:
https://claude.ai/share/c2ab6bcb-7032-489f-87d5-cc38f72c2ca9
Test geoquiz apk:
https://claude.ai/share/78242bf9-201b-40cc-9af8-7f2cdca36e56
# Build all components
make build
# Create distribution package
make dist
# API Server
make -C packages/api-server dev
# MCP Server
cd packages/mcp-server && pnpm dev
# MCP Inspector
cd packages/mcp-server && pnpm inspect
We welcome contributions! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
git checkout -b username/feature-name
)git commit -m 'Add some feature'
)git push origin username/feature-name
)gbox cleanup
. It will stop the api server so that you can run the api server in dev env.make api-dev
in project root../gbox box list
, this is the command run from your dev env.make mcp-dev
in project root../gbox mcp export --merge-to claude
make mcp-inspect
in project root.make build-image-python
in project root to build Python image, or make build-images
to build all images.make build-image-typescript
for TypeScript image)../gbox box delete --all
make build
to update the dist/index.js
file../gbox mcp export --merge-to claude
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
A goal-agnostic parallel orchestration framework implementing Infinite Agentic Loop patterns as a Model Context Protocol (MCP) server.
An MCP server for AI-assisted frontend development using Chrome DevTools. Requires Google Chrome.
Advanced text-to-image generation using the fal.ai Recraft v3 API.
Access prompt templates managed in an MLflow Prompt Registry. Requires a running MLflow server configured via the MLFLOW_TRACKING_URI environment variable.
Fetches comprehensive information about NuGet packages from the NuGet Gallery, including READMEs, metadata, and search functionality.
Access comprehensive documentation for Ant Design components, including examples, API references, and best practices.
Integrates with Microsoft's AutoGen framework to enable sophisticated multi-agent conversations via the Model Context Protocol.
Provides access to the Lucide icon library for use in LLM and agentic applications.
Parses HAR (HTTP Archive) files and displays requests in a simplified format for AI assistants.
Performs complementary code analysis by combining Claude Code and Google's Gemini AI.