Genai Sandbox

This MCP server enables remote Python code execution through a dedicated Lab MCP setup. You can run and test custom code remotely from MCP clients, such as Claude AI, in real time within a sandboxed lab environment. It is designed for practical lab scenarios like code evaluation and sandbox testing with secure, isolated execution.

How to use

Configure a client to connect to the Cloudlab MCP server using the provided MCP configuration. You will run the local server via a stdio-based interface and start the client controller to initiate remote code execution sessions.

How to install

Prerequisites you need to have before installation:

Python 3.10.11

pip (Python package manager)

fastmcp (to serve the MCP endpoint)

uv (virtual environment manager) via scoop or curl

Access to Claude Desktop or Cursor or cline (for testing)

How to install

Step 1: Clone the MCP Server Repository

git clone https://github.com/Nuvepro-Technologies-Pvt-Ltd/McpSever_Remote_code_execution.git

Step 2: Set up PowerShell execution policy (Windows)

Set-ExecutionPolicy RemoteSigned -Scope CurrentUser

Step 3: Install Python and UV in the shell

scoop install python

scoop install uv

Step 4: Change directory to the project

cd McpSever_Remote_code_execution

Step 5: Create and activate a Python virtual environment

python -m venv .venv

.\.venv\Scripts\activate   # Windows

source .venv/bin/activate  # macOS/Linux

Step 6: Install Python dependencies

pip install fastmcp

pip install cryptography

pip install shelve

Step 7: Run the MCP server

fastmcp run app.py

Step 8: Prepare MCP client configuration and start the client

{
  "mcpServers": {
    "CloudlabMcp": {
      "disabled": false,
      "timeout": 500,
      "type": "stdio",
      "command": "uv",
      "args": [
        "run",
        "--with",
        "fastmcp",
        "python",
        "%PROJECT_PATH%\\app.py"
      ],
      "env": {
        "API_KEY": "your_private_key",
        "Baseurl": "your seed phrase here",
        "compnaykey": "your_private_key"
      },
      "autoApprove": [*]
    }
  }
}

Step 9: Set project path and start the MCP client

set PROJECT_PATH=D:\YourProject

cline run CloudlabMcp

Additional sections

Configuration and usage notes

The lab server exposes a standard input/output (stdio) MCP interface via uv. You run the server with fastmcp and point the client at the Python app. The client’s mcp_config.json should reference the CloudlabMcp entry, including environment variables for API keys and secrets.

Security considerations

Add sandboxing logic to app.py to constrain executed code. Consider Docker or subprocess isolation for extra safety. Monitor logs and set clear execution timeouts to prevent runaway code.

Notes

This lab server is designed for real-time code evaluation and sandbox testing in a controlled environment. Use legitimate API keys and manage access to the MCP endpoints to prevent misuse.

Available tools

execute_code

Executes user-provided Python code