code-sandbox-mcp

Stale
GitHub Go MIT

Description

An MCP server that creates secure Docker-based sandboxes for executing untrusted code from AI applications.

Key Features

  • Docker isolation - Each execution runs in a fresh Docker container isolated from host network and filesystem
  • MCP-compliant API - Exposes create, execute, and cleanup tools over the Model Context Protocol
  • Multi-language - Bundles Python, Node, and Bash runtimes that can be invoked inside the sandbox
  • Resource limits - Configurable CPU, memory, and timeout caps protect against runaway agent loops
  • Auto cleanup - Containers are torn down automatically after execution

Use Cases

💡 Let Claude Desktop or Cursor execute arbitrary Python/JS via MCP
💡 Provide per-run isolated environments for code-generation agents
💡 Evaluate third-party code snippets under controlled constraints
💡 Act as a generic code-interpreter node in multi-agent pipelines

Tags

mcp sandbox docker code-execution

Categories

🏗️ Sandbox & Execution 🔌 MCP Tools

Quick Start

# Start code-sandbox-mcp
docker run -d --rm -p 8080:8080 automata-labs/code-sandbox-mcp

# Configure in your MCP client (e.g. Claude Desktop)
# claude_desktop_config.json
{ "mcpServers": { "sandbox": { "url": "http://localhost:8080/mcp" } } }
Visit GitHub

Related Projects