Overview

This MCP (Model Context Protocol) server is a reference implementation that connects Roblox Studio to external MCP-capable AI clients (for example, Claude Desktop or Cursor). It acts as a bridge between a Studio plugin and an LLM client, letting the model request actions in the open Roblox place (run code, insert models, collect console output, control play mode) and receive structured responses. The project is implemented in Rust and includes both an HTTP long-polling server (for the Studio plugin) and an rmcp/stdio component (for communicating with MCP clients).

Important: this repository is no longer under active development. Roblox recommends using the built-in MCP Server bundled with Roblox Studio for newer integrations. The source and releases in this repo remain available for reference or to support existing workflows.

Features

Long-polling web endpoint that a Roblox Studio plugin uses to receive tool requests
stdio rmcp transport for connecting widely used desktop MCP clients
A set of pre-built tools that let an LLM read and modify the current place, run code, and control play mode
Prepackaged installers for Windows and macOS, plus a build-from-source flow
Simple configuration snippets for common MCP clients (e.g., Claude Desktop)

Installation / Configuration

Quick automated install (release binaries):

Download the latest release for your platform from the GitHub releases page: https://github.com/Roblox/studio-rust-mcp-server/releases
Run the downloaded executable (Windows) or unzip and run the macOS app.
Restart your MCP client (Claude, Cursor, etc.) and Roblox Studio.

Manual MCP client configuration (example JSON):

{
  "mcpServers": {
    "Roblox_Studio": {
      "args": ["--stdio"],
      "command": "C:\\Path\\to\\rbx-studio-mcp.exe"
    }
  }
}

macOS example path:

"command": "/Applications/RobloxStudioMCP.app/Contents/MacOS/rbx-studio-mcp"

Add via Claude CLI example:

claude mcp add --transport stdio Roblox_Studio -- '/Applications/RobloxStudioMCP.app/Contents/MacOS/rbx-studio-mcp' --stdio

Build from source:

# Install Rust (https://www.rust-lang.org/tools/install)
git clone https://github.com/Roblox/studio-rust-mcp-server.git
cd studio-rust-mcp-server
cargo run

The cargo run workflow builds the server binary, configures the MCP client (when supported), and installs the Studio plugin used for long-polling.

Available Tools

The server exposes a set of tools the Studio plugin can execute on behalf of an LLM. Below is a summary.

Tool name	Purpose	Typical output
run_code	Execute arbitrary Studio commands (e.g., Lua code or plugin actions) and capture printed output	Standard output text and success/failure
insert_model	Insert a model from the Roblox Creator Marketplace into the workspace	Name/ID of inserted model
get_console_output	Retrieve current Roblox Studio console logs	Log entries, timestamps
start_stop_play	Start or stop play mode or run a server instance	New Studio mode or confirmation
run_script_in_play_mode	Run a script while in play mode and auto-stop after completion/timeout	Structured result with logs, errors, duration
get_studio_mode	Query current Studio mode (start_play, run_server, stop)	Mode string

Use the MCP client UI (hammer/tools icon in clients like Claude Desktop) to view and invoke these tools after installation.

Use Cases

Automated debugging: Ask an LLM to run diagnostic code, collect console output, and summarize errors without leaving Studio.
Rapid iteration: Insert example models from the Creator Store, tweak properties via run_code, and validate behavior in play/test mode.
Integration testing: Execute deterministic scripts in play mode using run_script_in_play_mode to reproduce edge cases and capture logs.
Developer-assist workflows: Have an assistant generate and run setup scripts to scaffold scenes and verify expected objects exist.

Concrete example — run a setup snippet and collect logs:

From your MCP client, invoke run_code with a Lua snippet that creates parts and prints status.
Use get_console_output to capture prints and errors.
If you need runtime behavior, call start_stop_play to enter play mode and run run_script_in_play_mode to exercise runtime logic.

Verification & Troubleshooting

Confirm the Studio plugin appears under Plugins and that the Studio console prints: “The MCP Studio plugin is ready for prompts.”
In your MCP client, open the MCP tools list (hammer icon) and verify Roblox tools are listed.
If tools don’t appear, restart the MCP client and Roblox Studio. Some clients run in the system tray—ensure the app is fully open.

Security & Privacy

This MCP server allows third-party tools to read and modify your open place. Only enable MCP connections for trusted clients. Third-party privacy and data-handling policies apply — review them before use.

Repository and releases: https://github.com/Roblox/studio-rust-mcp-server

For the most up-to-date and maintained experience, consider the built-in MCP Server provided within Roblox Studio.

Roblox Studio MCP Server: Scenes and Scripts