WEPPY Roblox AI Toolkit - Roblox Studio MCP with Sync, Playtest, and UI Studio

Have you ever tried connecting Claude Code, Codex, Gemini, Cursor, Antigravity, or another AI coding agent to Roblox Studio using MCP for development?

If you have, you’ve probably run into moments like these:

  • MCP installation was complicated — you got stuck before even getting started
  • The connection between the plugin and MCP keeps dropping or just won’t connect
  • No way to tell whether the AI agent actually executed an MCP tool correctly
  • No visibility into which instances the AI touched in Studio or in what order
  • No idea whether the AI is actually reading the playtest results properly

Handing work off to an AI feels great — until the whole process becomes a black box. WEPPY was built to solve exactly that.


Introducing WEPPY

WEPPY is a toolkit designed to help developers build on Roblox using AI agents like Claude, Codex, Gemini, and Antigravity.

It’s built around two goals:

Goal 1: Create the best possible environment for AI to work effectively in Roblox Studio
Goal 2: Let developers transparently see and control what the AI is doing


Feature 1 — Easy Installation

MCP installation shouldn’t be a barrier. WEPPY offers two ways to install:

One-line Terminal / PowerShell install — A single command handles MCP registration and Studio plugin installation all at once.

Web-based install guide — Not comfortable with the terminal or PowerShell? Follow the step-by-step UI guide on the website and get set up with just a few clicks.

WEPPY can be installed for any AI client that supports MCP. Common setups include Codex, Claude Code, Antigravity, Cursor, and other MCP-capable AI agents.

→ Install page: Install WEPPY
→ Detailed installation guide: Roblox Studio Plugin Installation Guide | WEPPY


Feature 2 — MCP + Plugin Bidirectional Sync

WEPPY connects an MCP server and a Roblox Studio plugin in both directions, giving AI agents everything they need to work effectively inside Studio.

Here’s what AI can do directly through the MCP tools:

  • Instance Management — Search, create, delete, clone, and move instances
  • Properties & Tags — Read and modify instance properties, manage tags
  • Scripts — Read, write, and edit scripts
  • Selection — Read and set Studio selection state
  • Camera — Query camera info and move focus
  • Spatial Query — Explore instances by position
  • Assets & Environment — Control lighting, terrain, audio, effects, and more
  • Playtest — Run playtests and collect logs
  • Luau Script Execution — When needed, execute Luau scripts directly to call virtually any Roblox Studio API

→ Learn more: Tools & Stats | WEPPY


Feature 3 — Connection Stability: Dashboard & Auto-Reconnect

Dashboard auto-launch: When the MCP server starts, the Dashboard launches automatically alongside it. You can check the connection status between MCP and the plugin at a glance.

Plugin auto-reconnect: The plugin displays the live connection status and automatically attempts to reconnect if the connection drops. No more guessing whether things are connected or not.

→ Check connections guide: Check Connections | WEPPY
→ Dashboard guide: MCP Dashboard Guide | WEPPY
→ Troubleshooting: Troubleshooting | WEPPY


Feature 4 — VS Code Explorer & Agent Activity Tracking

Explorer: View the current state of your Studio workspace as a live tree structure directly inside VS Code.


Dashboard Tool History: See exactly which MCP tools the AI agent called, how many times, and whether each one succeeded or failed. Never wonder again whether the AI actually did what it was supposed to do.

→ Explorer guide: Roblox Explorer Guide | WEPPY
→ Tools & Stats: Tools & Stats | WEPPY


Feature 5 — Playtest Monitoring

Logs, errors, and events during playtesting are visible to both the AI and the developer in real time. The AI can read playtest results directly and use them to inform the next steps — no manual copy-pasting required.

→ Learn more: Playtest | WEPPY


Feature 6 - UI Studio

UI Studio helps AI agents create UI that fits the actual style and context of your Roblox game, instead of generating a generic interface from a simple prompt.

Before starting UI work, the AI Agent analyzes the game’s visual style, genre, existing screen layout, and available assets. If the request is too vague, it first plans the UI direction more clearly so the result is closer to the game you are actually building.

The AI Agent can also review the generated UI and suggest improvements for issues that may affect real gameplay, such as button size, readability, visibility, and layout.

When UI changes are made, WEPPY records screenshots and change details in the dashboard history, so you can review what changed, when it changed, and how the UI evolved over time.

→ Learn more: UI Studio Guide | WEPPY

Links


WEPPY is still a work in progress, and there’s plenty left to improve.

This project started from the frustrations I ran into while building with AI agents — and I suspect some of you have felt the same. If you give it a try and have feedback or ideas, feel free to drop them in GitHub Issues or right here in this thread. I’d genuinely appreciate it.

5 Likes

More AI? Really? Just who funds weppyai.com ?

1 Like

I also find it suspicious that these AI apps keep popping up from nowhere, usually from users who have no public profiles or a GitHub account with less than five repos, or in this case, both. Very odd…

It’s also a huge security risk to be integrating some unknown opaque individuals app into your development pipeline. Please keep that in mind, OP! People need to be able to trust you to use this application of yours.

1 Like

I completely understand the concern. Skepticism around AI tools is a fair and reasonable reaction.
That said, Roblox itself has been actively introducing MCP and AI-powered features, and I believe there are already developers using AI agents as part of their workflow. WEPPY is a tool built for those developers — to provide a more convenient and transparent environment for AI-assisted development. I fully understand it won’t be the right fit for everyone. I’m sorry if the post came across the wrong way.

2 Likes

Security is a completely valid concern, and it’s something I’ve taken seriously throughout the entire development process.

Dependencies are regularly scanned for known vulnerabilities using npm audit and Snyk. Beyond that, with the exception of access statistics collected via Google Analytics, all functionality in WEPPY runs entirely locally — no data is transmitted externally in any form. The source code is publicly available on GitHub, so anyone is welcome to verify that directly. Security feedback is always welcome.

1 Like

Based on the responses above, I guess you can see how much I dislike this type of posts.

2 Likes

My concerns aren’t with your supplychain, but rather with you. You wouldn’t believe how many supplychain attacks stem from compromised accounts. You have absolutely no reputation anywhere on the internet. And it doesn’t exactly help that your responses are likely AI generated, given how you responded in such quick succession (to me and @Tavikron), and the results from an AI detector:


Just to be clear, there’s nothing wrong with using AI to refine your writing, as I’ve stated many many times before (you can check my post history). Though using AI directly to respond to people is not okay, especially regarding your product. If I wanted to talk to an AI chatbot I can do it on my own. Broken/bad English is much more sincere than AI generated corporate nonsense.

2 Likes

These responses were translated into English using Claude. As a non-native English speaker, I’m not fully comfortable writing in English directly. I write my replies in Korean first, then translate them using translation tools and AI assistants like Claude and Codex. The replies appeared back-to-back because I translated them together — I can see how that might look suspicious, and I apologize for any confusion.

4 Likes

I’m glad that you are willing to be transparent about it. It’d be better if you disclosed it on the post too

1 Like

Thank you, that’s a great idea. I’ll add a note about this in the post.

1 Like

You need to let some comments about your product just slide in the one ear & out the other.
In other words - ignore many of them.

Your Plugin looks cool - and does a lot more than other MCP’s floating around.
Some of us like me - code security isn’t that important - as we use Ai anyway.

Keep up the good work - will watch how you update your App. :wink:

5 Likes

Thank you. I honestly thought a lot about posting on DevForum, since there are some critical views on AI. But I wanted to gather a wide range of feedback and use it to improve and add more ways to make creating Roblox games with AI truly easy. I’ll keep updating it consistently, so please stay tuned. Thank you again for your support. @TQNY2

2 Likes

With bidirectional Sync and Roblox Explorer, it is also possible to edit Roblox instance properties directly from VS Code or Antigravity, then sync those changes back into Roblox Studio.

1 Like

Does the property sync handle complex types or nested tables correctly? Syncing basic properties is one thing, but if it breaks on more complicated instance data, it is not much better than what we already have.

2 Likes

@Ionypex23 Thank you for the good question. The current property sync is not limited to only basic primitive properties. For more complicated instance data, we also handle Roblox-supported complex datatypes within the property/attribute sync scope. For example, datatypes such as Vector3, Color3, CFrame, UDim2, Enum, and NumberRange are considered part of the sync target, and we are continuously validating and expanding these scenarios through E2E tests.

That said, if by nested tables you mean custom Lua table data written directly inside Scripts or ModuleScripts, rather than structures used to represent Roblox property datatypes, then that is not handled as part of property sync. In that case, the script source itself can be synced, but we do not parse the internal table values and sync them structurally as separate properties.

Of course, there may still be complex types or sync scenarios that we have not yet thought of, did not include in our test cases, or do not handle well enough yet. If you have any specific cases that are missing or need improvement, please let us know. We will review them and do our best to apply improvements quickly.

1 Like

That makes sense. As long as the standard Roblox types are covered, it is useful enough. The script source sync is fine for custom tables, since parsing arbitrary Lua logic would just be overkill anyway.

2 Likes

You’re right.

In addition to that, I’m also creating rules to handle cases like complex instances with the same name, and checking various edge cases.

In Roblox Studio, the folder and file rules in Explorer allow many things that are different from a normal OS file system, and there is a lot of data that needs to be synced, so there are quite a few constraints.

I’m continuing to improve this so developers who mainly work in VS Code or Antigravity can inspect and work with it more comfortably. Please keep an eye on it.

1 Like

Handling the Explorer hierarchy differences is probably the hardest part. As long as the sync stays reliable with those edge cases, it will be a solid tool.

2 Likes

Thank you. I will continuously look for areas to improve in the synchronization part and keep enhancing it. Please stay tuned.

1 Like

There are already various MCP options, but one reason I built WEPPY was to make installation and connection simpler and more reliable.

Each AI Agent has its own MCP setup method. In many cases, you also need to download the Roblox Studio plugin, copy it manually, and then hope the AI Agent connects correctly. I kept running into cases where the agent was not connected, or the tool did not run for unclear reasons.

So WEPPY provides a one-line install script.

The setup only takes two steps.

  1. Copy and run the one-line script in Terminal on macOS or PowerShell on Windows.

The script automatically handles:

  • finding available AI Agents and registering the MCP server
  • downloading the plugin file
  • registering it as the Roblox MCP plugin in Roblox Studio
  1. Restart the AI Agent and Roblox Studio.

After restart, the MCP server runs automatically and connects to the Roblox Studio plugin.

You can see the full process in the install guide video:

2 Likes