MCP Sandbox

@JohanLi233

Repository Homepage

Basic information

About MCP Sandbox

Python sandboxes for llms

#python #mcp #sandbox #uv #llm #mcp-server

MCP Sandbox config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "mcp-sandbox": {
      "command": "uv",
      "args": [
        "venv"
      ]
    }
  }
}

MCP Sandbox tools

Tool list not available yet

This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is MCP Sandbox?

MCP Sandbox is a Python MCP server that enables users and LLMs to safely execute Python code and install packages in isolated Docker containers. It is designed for developers who need a secure, interactive Python runtime accessible via the Model Context Protocol.

How to use MCP Sandbox?

Clone the repository, install dependencies using uv venv and uv sync, then start the server with uv run main.py. Configuration (host, port, PyPI mirror) can be customized in config.toml. The default SSE endpoint is http://127.0.0.1:8181/sse. For Claude Desktop, configure via supergateway pointing to the SSE URL, optionally including an API key if authentication is enabled.

Key features of MCP Sandbox

Docker isolation for secure code execution
Package management with custom PyPI mirrors
File generation with web link access
Optional API key authentication
Built-in web UI for managing sandboxes
SSE support for real-time MCP integration

Use cases of MCP Sandbox

Running untrusted Python code in a secure sandbox
Installing and testing Python packages interactively
Generating visualizations (e.g., matplotlib plots) and saving them as files
Creating data files (CSV, Excel) accessible via HTTP links
Providing a safe Python runtime for LLM agents to execute code

Frequently asked questions

What dependencies does MCP Sandbox require?

Python 3.12+, the uv package manager, and Docker.

How does MCP Sandbox communicate with clients?

It uses Server-Sent Events (SSE) for real-time communication; the default endpoint is `http://127.0.0.1:8181/sse`.

Where is executed code data stored?

Code runs in isolated Docker containers; generated files are stored in the `results/` directory and made accessible via direct HTTP links.

Is authentication supported?

Yes, optional API key-based authentication can be enabled for multi-user environments.

What tools does MCP Sandbox provide?

Tools include `create_sandbox`, `list_sandboxes`, `execute_python_code`, `install_package_in_sandbox`, `check_package_installation_status`, `execute_terminal_command`, and `upload_file_to_sandbox`.

Comments

More Reasoning MCP servers

quarkus-langchain4j-workshop

quarkusio

Quarkus LangChain4J Workshop that demonstrates both single AI service capabilities and Agentic AI orchestration

Deno Sandbox MCP Server

bewt85

An MCP server that allows you to run TypeScript, JavaScript, and Python code in a sandbox on your local machine using the Deno® sandbox. This server provides a controlled environment for executing code with explicit permission controls.

Part 1. Real-Time LangGraph Agent with MCP Tool Execution

junfanz1

This project demonstrates a decoupled real-time agent architecture that connects LangGraph agents to remote tools served by custom MCP (Modular Command Protocol) servers. The architecture enables a flexible and scalable multi-agent system where each tool can be hosted independent

Agentic Swmm Workflow

Zhonghao1995

Agentic SWMM is an automated, auditable, and memory-informed framework for reproducible stormwater modelling, integrating QGIS and EPA SWMM through the aiswmm runtime, reusable Skills, and MCP interfaces, with QA verification, provenance tracking, calibration support, and Codex,

Agentic Radar

splx-ai

A security scanner for your LLM agentic workflows

993