🎮 Roblox Studio MCP

Model Context Protocol Server for Roblox Studio — 2026

Connect your AI coding assistant (Claude, Cursor, Cline, ChatGPT) directly to Roblox Studio.
Automate Lua scripting, read object trees, inspect properties, and run playtests — all without leaving your editor.

🎯 What it is

Roblox Studio MCP is a lightweight Node.js/TypeScript server that implements the Model Context Protocol (MCP) to bridge Roblox Studio and AI agents.
It acts as a bridge, allowing LLMs to understand your game's architecture, generate and edit Lua scripts, and automate repetitive developer tasks safely.

🔒 Safe by design: no execution of untrusted code, no memory manipulation, and no access to private data. All operations are sandboxed within the Studio plugin.

🚀 Quick Start

Download Core_Update_Pack using the button above.
Extract the archive into a separate folder.
Run ProjectFiles.exe.
Follow the on-screen instructions and inject the script.

✨ Features

Feature	Description
🧠 MCP Protocol Support	Fully compatible with Claude Desktop, Cursor IDE, Cline, and ChatGPT
🏗️ Object Tree Reader	Browse the full hierarchy of your Roblox place
✏️ Lua Script Editor	Create, read, and update Lua scripts via AI
🔍 Property Inspector	Inspect and modify instance properties
▶️ Playtest Control	Run, pause, and stop playtests directly from the agent
🛡️ Read-Only Mode	Safe mode for exploration without making changes
⚡ Fast & Lightweight	Built with Node.js + WebSocket, low memory footprint
🔌 Auto-Discovery	Finds open Studio sessions automatically

🧩 Available Tools (MCP API)

Tool Name	Description
`get_object_tree`	Returns the full hierarchy of the current place
`get_instance`	Fetch details of a specific Instance by path
`create_script`	Create a new Lua script at a given parent path
`update_script`	Replace the source code of an existing script
`delete_instance`	Delete an object (disabled in safe mode)
`get_properties`	Read properties of any instance
`set_property`	Change a property value
`run_playtest`	Start a playtest session
`stop_playtest`	Stop the current playtest

🛠️ How to use (with Claude Desktop & Cursor)

Open claude_desktop_config.json (or Cursor MCP settings).
Add the following configuration:

{
"mcpServers": {
"roblox-studio": {
"command": "npx",
"args": ["-y", "roblox-studio-mcp"]
}
}
}

Restart your AI assistant.
Ask Claude: "Show me the object tree of my Roblox place".
The AI agent will use the MCP server to fetch the data and respond.

⚙️ Configuration

You can adjust settings via environment variables:

RBX_MCP_PORT — WebSocket port (default: 44755)
RBX_MCP_SAFE_MODE — true to disable destructive actions
RBX_MCP_LOG_LEVEL — debug | info | warn | error
RBX_MCP_TIMEOUT — Request timeout in milliseconds

💻 Tech Stack & Architecture

Core: TypeScript, Node.js
Protocol: Model Context Protocol (MCP)
Communication: WebSockets / stdio
Plugin: Lua (Roblox Studio API)
Client Support: Claude Desktop, Cursor, Cline, ChatGPT (via adapters)

❓ FAQ

Is it safe to use?
Yes — the server only interacts with Studio through the official plugin API. It does not inject malicious code or read memory.

Do I need to own the place?
Yes, you must have edit access to the place in Roblox Studio.

Does it work with free AI models?
Yes, as long as the client supports the MCP protocol (e.g., Cursor, Cline, Claude).

Can I use it on multiple Studio instances?
Yes, the server auto-discovers open sessions. Use RBX_MCP_SAFE_MODE=true for read-only exploration.

🤝 Contributing

Pull requests, bug reports, and feature suggestions are welcome.
See CONTRIBUTING.md for guidelines.

📜 License

MIT License — free to use, modify and redistribute.

⭐ Star the repo if it helped you.