❓ Problem

Building a game stopped being the hard part. You describe what you want to Claude, Codex or Cursor, and twenty minutes later you have something playable in a browser tab. The satisfying bit lasts until you try to show someone. You send them a link, they open it, and they are playing their own private copy in a parallel universe. Same game, different reality, no shared score.

Making that link genuinely multiplayer is where the evening dies. You need authoritative game state, a lobby, reconnection handling, latency compensation and somewhere to host it all. This is real distributed-systems work, and it is the one category of problem coding agents are still bad at, because getting it wrong does not throw an error, it just produces a game where your mate's paddle is somewhere it shouldn't be.

The supply side is going vertical anyway. The number of HTML5 games shipped in the first half of 2025 nearly tripled year on year to 15,000, almost five times the H1 2023 figure, per Playgama's research. The demand side is already there: Poki now reports 100 million monthly players, up from 10 million in 2020, and CrazyGames serves roughly 35 million monthly users across 4,500 games. A wave of new creators is arriving, and the only place they can currently monetise a multiplayer game properly is Roblox, which means writing Lua inside someone else's avatar system. Your agent-built Three.js game does not go there. Why browser games are the next billion-dollar bet - MCV/DEVELOP +2

✅ Solution

A hosting and discovery platform for multiplayer web games, with documentation written for coding agents rather than humans.

• The docs are the product. You paste one URL into Claude, Codex or Grok and say "make this multiplayer and publish it". The agent reads an llms.txt file or hits an MCP server, wires up the SDK, and hands back a playable link. No dashboard, no signup, no reading anything yourself.
• The platform owns the hard parts. Authoritative shared state, rollback netcode, lobbies, reconnection, room codes and hosting. These are exactly the things agents get subtly wrong and exactly the things you only need to solve once.
• The link is the unit of distribution. No install, no account, no app. You drop it in a group chat and five people are in a lobby in ten seconds. Accounts only appear when someone wants to save progress or spend money.
• An arcade for the survivors. Every published game lists in a searchable arcade ranked on playtime and repeat sessions, not upload date, so strangers find the small number of games that are actually fun.
• Then a shared currency. Once sessions are regular, introduce a platform currency creators can charge for skins, extra levels, boards and continues. Platform takes a cut, creators keep the rest.

The wedge is deliberately small: get one person's agent-built game playable with four friends tonight. The platform underneath, netcode plus discovery plus a currency, is where the business is.

📊 Key Numbers

Market size

• The global games market closed 2025 at $201.6 billion, up 9.1% and crossing $200 billion for the first time, according to Newzoo. Not the addressable market, but the gravity well everything here sits in. GamesBeat
• The relevant comparable is Roblox. It booked $6.8 billion in 2025, up 55% year on year, on $4.9 billion of revenue, and its 10-K reports an average 127 million daily active users spending 123.9 billion hours on the platform. That is the ceiling for a creator-plus-currency model, documented in a filing. Shane the GamerStocktitan
• The creator payout pool is the real signal. Roblox paid creators $1.503 billion in 2025, up from $922.8 million in 2024, across more than 35,500 registered DevEx creators. Stocktitan
• The browser-game audience is already assembled and unmonetised by anyone doing multiplayer properly: Poki at 100 million monthly players, CrazyGames at 35 million, and GameDistribution syndicating to 4,000-plus portals reaching 350 million monthly users at a 33% developer share. Cinevva