MCP.so
ログイン

Deno Sandbox MCP Server

@bewt85

リポジトリ

Deno Sandbox MCP Server について

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.

基本情報

カテゴリ

推論

ランタイム

node

トランスポート

stdio

公開者

bewt85

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-deno-sandbox": {
      "command": "npx",
      "args": [
        "@modelcontextprotocol/inspector",
        "./node_modules/.bin/ts-node",
        "src/index.ts"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Deno Sandbox MCP Server?

An MCP server that allows you to run TypeScript, JavaScript, and Python code in a local sandbox using the Deno® runtime. It enforces explicit permission controls (e.g., which websites to visit, which files to read/write) to protect your machine from untrusted code, especially when code is generated by LLMs.

How to use Deno Sandbox MCP Server?

Install Node.js or Deno. Add the server to your client’s configuration (e.g., claude_desktop_config.json) using either deno run npm:mcp-deno-sandbox or npx mcp-deno-sandbox, passing permission flags like --allow-net=icanhazip.com,example.com at startup. Permissions are static and require a server restart to change.

Key features of Deno Sandbox MCP Server

  • Runs TypeScript, JavaScript, and Python code locally.
  • Uses Deno’s Chrome-based sandbox technology.
  • Configurable read, write, and network permissions.
  • Deny access to specific files or IP addresses.
  • Python execution via Pyodide inside the Deno sandbox.
  • Minimal codebase designed for auditability.

Use cases of Deno Sandbox MCP Server

  • Let an LLM test generated code without risking your system.
  • Summarise documents that may contain hidden prompt injections.
  • Run untrusted scripts in a restricted file and network sandbox.
  • Process isolated files (e.g., in /tmp) with limited permissions.
  • Experiment with multi-language code without installing runtimes.

FAQ from Deno Sandbox MCP Server

What languages can I run in the sandbox?

TypeScript, JavaScript, and Python. Python runs via Pyodide inside the Deno environment.

How do I set permissions?

Permissions are passed as command-line arguments when starting the server (e.g., --allow-read=/tmp, --allow-net=example.com). They are the same as Deno permissions; you cannot change them without restarting the server.

Is this project affiliated with Deno Land LLC?

No. “Deno” is a registered trademark of Deno Land LLC. This project is an independent fan creation.

What are the known Python limitations?

Writing files with open(PATH, 'w') does not work; use import js; js.fs.writeFileSync(PATH, CONTENT) instead. When using --allow-read broadly, the server guesses which directories to mount into Pyodide (home and /tmp). Windows mounting may have issues.

How does the sandbox protect against malicious code?

It uses Deno’s sandbox, which relies on Chrome’s isolation technology. You control file, network, and execution permissions. Granting write access to the server’s config file or blanket FFI/execution permissions can completely undermine the sandbox.

コメント

「推論」の他のコンテンツ

Agenticstore — The Open Source Standard For Local Mcp Tooling

agenticstore

AgenticStore: The secure toolkit for AI agents. Instantly equip Claude Desktop, Cursor, and Windsurf with 27+ MCP tools, persistent memory, and SearXNG search, all protected by a built-in PII prompt firewall to protect your data from being exposed to AI agents.

4

quarkus-langchain4j-workshop

quarkusio

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

91

NeoCoder: Neo4j-Guided AI Coding Workflow

angrysky56

An MCP server allowing AI assistants to use a Neo4j knowledge graph as their primary, dynamic instruction manual and long term project memory with adaptive templating and autonomous tool development tools.

20

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

27

🚀 Aider-MCP: AI Coding Server with Universal Auto-Detection

jacv888

Aider-MCP-Upgraded is a production-grade multi-agent AI coding system that combines Desktop Commander (DC) investigation capabilities with Aider's implementation power. Features 70%+ token reduction, modular architecture, and intelligent workflow automation through strategic agen

4