2 Months Vibe Coding My Own MMORPG — Here’s What Actually Happened

Two months ago I did something slightly crazy: I started building an MMORPG. Not a small mobile game — a real, multiplayer online world that runs in your browser. And here’s the twist: I barely touch the code myself. This is my story of 2 months vibe coding with Claude Code, and honestly, it’s been wilder than I expected.

The game is called Lythravel. It’s a free 3D browser MMORPG with blocky, voxel-style graphics, heavily inspired by old-school MMOs I grew up playing — especially Shaiya. Think of it as a love letter to the games of the 2000s, built almost entirely by talking to an AI.

Wait — What Is “Vibe Coding”?

If you haven’t heard the term, vibe coding means you describe what you want in plain language, and AI writes the actual code. You’re the director, not the typist. Instead of hammering out lines of code, I spend my time explaining how systems should work, testing the results, and giving feedback.

To be clear, I’m not a total beginner — I’ve been a software developer for over 10 years and had built a few small games before AI came along. But an MMORPG? That’s a different beast. Multiple players in one world, servers, combat, inventory, quests. These 2 months vibe have taught me more about scope creep than a decade of regular dev work.

The Tech Stack (In Human Words)

Every project needs tools, and mine are pretty simple. The game runs on Babylon.js, a free engine that draws 3D graphics right in the browser — no downloads needed. For the multiplayer side, I use Colyseus, which is like the referee of the game world: it keeps track of every player and makes sure everyone sees the same thing.

For the AI side, I mostly use Claude Opus for the heavy lifting, with a second model acting as an orchestrator for bigger tasks — basically a project manager that splits huge jobs into smaller chunks. It’s like having a senior dev and a team lead in one subscription.

What Went Well (And What Didn’t)

Here’s the honest part. Adding new features? Shockingly fast. I can describe a new quest system in the morning and be testing it by the afternoon. That speed is exactly why these months vibe coding have felt so productive.

But there’s a catch, and it’s a big one: performance. Features are easy — making them run smoothly as the game grows is the real battle. Think of it like moving into a bigger house: every room you add costs more to heat. My recent weeks have been all about optimization.

• Measuring exactly how long each part of a game frame takes to render (using a Chrome measurement tool)
• Making some systems update less often — not everything needs to refresh 60 times a second
• Merging geometry, so instead of drawing 500 separate blocks, the engine draws one big shape
• Reducing the detail of far-away objects, like how your eyes blur distant things

I also had Claude build custom editors and tools for different parts of the game. Turns out AI is great at building the tools that help you build the game. Very meta, very worth it.

Pro Tips: Things You Need to Start Vibe Coding

Want to try this yourself? Here’s what I’d recommend having before you dive in:

• A Claude Code subscription (Anthropic) — the core of the whole workflow. Budget for the higher tier; big projects eat through usage fast.
• A decent GPU or a machine with good graphics — testing a 3D game locally will make a weak laptop cry. A mid-range NVIDIA RTX card handles it fine.
• Chrome plus a measurement extension (Chrome MCP) — free, and it’s how you find out what’s actually slow instead of guessing.

One more tip: keep your day job’s discipline. Even with AI, you still need to test everything yourself. The AI writes code fast, but it also writes bugs fast.

Was It Worth It? My Honest Take

After 2 months vibe coding, my answer is yes — with an asterisk. If you expect AI to build you a World of Warcraft clone while you sleep, you’ll be disappointed. But if you’re willing to direct, test, and iterate daily, you can build something that would’ve taken a small team a year.

The biggest surprise wasn’t the speed. It was realizing that the hard part of game dev was never typing the code — it’s knowing what to build and making it feel good to play. AI can’t do that part for you. That’s still on you, and honestly, that’s the fun part anyway.

If you’re curious, go small first. Build a tiny prototype, see how the vibe workflow feels, then scale up. Two months from now you might have your own world too.

FAQ

Can a beginner with zero coding experience build an MMORPG with AI?

Honestly, an MMORPG is a rough first project — it’s like learning to drive in a race car. Start with something small like a single-player prototype. That said, vibe coding has genuinely lowered the barrier. You still need patience and problem-solving instincts, because when things break, you need to describe the problem clearly to fix it.

What is vibe coding exactly?

It’s when you describe what you want in normal language and let AI like Claude Code write the actual code. You review, test, and give feedback instead of typing everything yourself. Think director, not typist.

How much does it cost to build a game this way?

The main cost is the AI subscription — expect $20 to $200 a month depending on your plan and how much you build. Everything else I use (Babylon.js, Colyseus, Chrome measurement tools) is free and open source.

What’s the hardest part of AI game development?

Performance, hands down. Adding features is fast, but keeping a growing game running smoothly takes constant measuring and optimizing. That’s where most of my recent effort has gone.

Can you really make a multiplayer game run in a browser?

Yes — modern browsers are surprisingly capable. Babylon.js handles the 3D graphics and Colyseus manages the players on the server. It won’t match a AAA console game, but for an MMO inspired by older titles, it works great.