antics alternative for multiplayer games made with Claude Code or Codex
This page is written by the KUMODeck team. It compares KUMODeck with antics for a multiplayer browser game that your AI coding agent builds and puts online, using facts from antics' official pages.
Both let Claude Code or Codex put a browser game online over MCP and add multiplayer rooms to it. antics is built around multiplayer: one deploy_game call returns a playable link with rooms and live sync, without an account at first, and games are listed on antics.gg for its players. With KUMODeck, you sign in once in your browser, your AI puts a test version on your own project, and you press Publish yourself; sign-in, cloud saves and rooms are parts your AI adds when the game needs them.
Both are reasonable choices. This page lists the facts so you can pick by how you want to start and where you want the game to live. antics' facts come from its official pages, checked on 2026-11-05; its own pages are authoritative.
At a glance#
| KUMODeck | antics | |
|---|---|---|
| Connecting your AI | MCP or the KUMODeck CLI, after you sign in once in your browser and press Allow (Use from an AI agent (MCP)) | one line for Claude Code or Codex (claude mcp add antics -- npx -y antics-mcp), also Cursor; or give any agent its llms.txt, the whole API in one file |
| Account to start | yes: sign up, then add funds or enter an invite code (Start here) | none for a keyless deploy; a free account, signed in with Google, keeps links and scores |
| Plans | no free tier: at cost from the first request, from a prepaid balance | Keyless (free) and Free account (free); Pro is usage-based, on a waitlist, with no price published yet |
| Size of one deploy | up to 100 MB per file, with no limit on total size (Hosting & deploy) | one HTML file or a project: up to 3 MB per file and 12 MB in all |
| Test before going live | a test version on a --dev address; the public address waits for you to press Publish, and every version is kept | each deploy has its own link, pinned to those files; under a project, one share link always serves the latest deploy |
| Rooms | quick match, private rooms with a 6-letter code, a public lobby, host migration, a seat kept 20 seconds after a drop (Multiplayer rooms) | rooms with live state sync, invite links and a high-score table; 8 players per room without an account (rooms expire after 24 hours), 16 and up to 25 rooms at once with a free account |
| Game logic on a server | rooms relay and keep shared state; checks that matter go in your own Functions | the classic tier relays; server-authoritative games, where your logic runs on antics' servers, need Pro |
| Players' sign-in and saves | guests first, sign in later without losing progress; saves per slot with a version (Authentication, Cloud saves) | players need no account; the page around the game keeps saves (256 KB per slot, 8 slots per player per game), under the player's antics account when they have one |
| Where the game lives | your app's own address or your own domain ($0.11 a month per domain), with no KUMODeck branding in front of players | antics.gg, where games are listed for the players there; a custom domain and removing the "Made with antics" badge are on Pro |
Sources for antics, checked on 2026-11-05: antics home (making a game with ChatGPT's Codex or Claude Code, listing on antics.gg, worlds and rooms, no server to run), antics pricing (Keyless, Free account and Pro, players per room, concurrent rooms, server-authoritative games, custom domain, badge, Pro waitlist), MCP servers (the install line for Claude Code and Codex, deploy_game), Add multiplayer (adding rooms to a game you already made, Cursor, two-player checks) and the antics llms.txt (deploy sizes, share links, room expiry, saves). Prices and limits can change; antics' pages are authoritative.
Where KUMODeck is different#
You decide when it goes public. Your AI can put up as many test versions as you like, on an address with --dev in it. The address everyone opens waits for you to press Publish on a KUMODeck page, every version is kept, and "roll back to the previous version" switches back. A deploy whose files contain a secret key is stopped before it goes up (Secrets in deployed files).
Multiplayer is one part, not the whole product. The same project can host a big game, such as a Unity or Godot web export with files up to 100 MB each (Unity games, Godot games). Players can start as guests and sign in later with email or quick sign-in, and their saves follow them to every device. You ask your AI for each part in plain words: Add login and cloud saves.
Your AI picks how the game goes online, with prices first. Ask "make it playable online" and your AI asks one short question about who plays, how fast the game is and whether cheating matters, then builds it (Make a Claude Code game multiplayer). Each answer shows its price before anything is built. For a 4-player, 30-minute match, carrying every move costs about 0.31 to 0.47 cents, and letting the players' devices connect directly costs 0 cents for the moves (Play online together).
No secret key in the conversation. Your AI signs in through your browser and never sees a secret key, so there is nothing to copy into a file or paste into the conversation (Deploy a game from Claude Code without API keys).
When antics fits better#
- You want a playable link before making any account. A keyless deploy is instant; you sign in later to keep the link.
- You want players who are already there. Games on antics are listed on antics.gg, next to the other games its players browse.
- You want to start without paying. Keyless and Free account cost nothing, and classic multiplayer stays unlimited on every plan, by antics' pricing page.
- You want a high-score table built in, shared by everyone who plays the game's public link.
- You want game logic on the server without writing a server. On Pro, antics runs server-authoritative games on its own servers: physics, scores that cannot be faked, and checks on the server.
- You like one small API. The whole antics SDK fits in one file that your agent can read in one go.
What it cannot do, and conditions#
- There is no free tier. What your game uses comes out of your prepaid balance at cost. If the balance reaches $0, the game pauses for its players until you top up (Prepaid balance).
- Your AI cannot publish to everyone. It can make test versions; the public address always waits for your press on a KUMODeck page.
- Rooms do not run your game's rules. They relay messages and keep shared state. A competitive game is host-authoritative, or checks every move in your own Functions.
- One browser profile is one player. To test on one computer, open a second tab with
?player=2, or use a private window. - A room keeps nothing after the last player leaves. Scores and progress belong in cloud saves.
Questions and answers#
Can my AI put a game online with KUMODeck before I sign up?#
No. You sign up and connect your AI by signing in once in your browser. Then add funds or enter an invite code, and your AI can deploy test versions from then on.
Do my players need an account?#
No. Players start as guests. If the game turns on sign-in, they can sign in later and keep their progress (Authentication).
How many players can join one room?#
A project without its own modes gets a default mode for 2 to 8 players. Your AI sets the size per mode, for example 2 for a duel or 2 to 4 for a party game (Multiplayer rooms).
Can the game check moves on a server?#
Rooms relay and keep shared state; they do not run your game's rules. For moves that must be checked, your AI writes them in your own Functions, one live object per match.
What does a multiplayer match cost?#
It depends on how the game goes online. For a 4-player, 30-minute match, carrying every move is about 0.31 to 0.47 cents, and direct connections between players cost 0 cents for the moves. The table is on Play online together.
Can I keep a game that is already on antics?#
Yes. The game files your AI wrote are yours. antics' SDK joins rooms with joinRoom and KUMODeck's with kumo.rooms, so ask your AI to swap those calls in the same files, deploy a test version and try it in two tabs. When it plays well, press Publish.