There are already several MCP implementations for Roblox Studio, and after trying them myself, I felt there was a lot of potential.
But when I tried using them in actual development, I ran into a few problems.
- MCP installation and Studio plugin installation felt too difficult and tedious
- It was hard to tell whether MCP was running and actually connected to the plugin
- When the connection dropped, auto-reconnect was not enough, and the MCP/plugin connection did not feel smooth
- Whether the MCP tool actually ran
- Which instances the agent touched
- Whether the changes were properly reflected in Roblox Studio
- Whether the agent was properly reading logs and errors from playtest
I started building WEPPY to reduce that black-box feeling.
WEPPY is an AI/MCP toolkit for using AI coding agents like Claude, Codex, Gemini, and Antigravity with Roblox Studio. It connects an MCP server, Roblox Studio plugin, dashboard, and VS Code explorer so developers can see the agent's work more transparently.
Right now, I am building features like:
- One-line install for the MCP server and Studio plugin
- Bidirectional sync between MCP and the Roblox Studio plugin
- Connection status and plugin auto-reconnect
- Tool history showing which MCP tools ran and whether they succeeded or failed
- A VS Code explorer for checking the Studio workspace
- Playtest monitoring so both the developer and the agent can inspect logs and errors
Demo video: https://youtu.be/puQB4u1VlMw
GitHub: https://github.com/hope1026/weppy-roblox-mcp
I would like feedback from developers who use AI agents to build Roblox games.
I want to hear what feels uncomfortable when letting an AI agent work in Studio, or what features you wish existed.