Skip to content

MCP.md

Latest commit

 

History

History
80 lines (52 loc) · 2.12 KB

MCP.md

File metadata and controls

80 lines (52 loc) · 2.12 KB

  MODEL CONTEXT PROTOCOL

microsandbox server is also a Model Context Protocol (MCP) server, enabling seamless integration with AI tools and agents that support MCP.

Claude.Desktop.Demo.mp4

What is MCP?

The Model Context Protocol (MCP) is an open standard that allows AI models to securely connect to external data sources and tools. It provides a standardized way for AI assistants to access and interact with various services through a unified interface.

With MCP, your AI can:

  • Execute code in secure sandboxes
  • Access real-time data and services
  • Perform complex operations safely
  • Maintain context across interactions

Connection Details

microsandbox server supports MCP connections via Streamable HTTP transport only.

  • Transport: HTTP
  • URL: http://localhost:5555/mcp
  • Method: Streamable HTTP
Server Configuration

When running your microsandbox server, it automatically exposes MCP endpoints at:

http://localhost:5555/mcp

Note

The MCP endpoint uses the same host and port as your main microsandbox server.

Getting Started

  1. Start your microsandbox server:

    msb server start --dev

  2. Configure your MCP client with the connection details above

  3. Begin using sandbox tools through your AI assistant

    Try these example prompts:

    "Create a Python sandbox and run a simple hello world program"

    "Start a Node.js sandbox, install express, and create a basic web server"

    "Execute this shell command in a sandbox: curl -s https://api.github.com/users/octocat"

Available Tools

The microsandbox MCP server provides these tools:

  • sandbox_start - Create and start new sandboxes
  • sandbox_stop - Stop running sandboxes
  • sandbox_run_code - Execute code in sandboxes
  • sandbox_run_command - Run shell commands
  • sandbox_get_metrics - Monitor sandbox status