Antics
About
An MCP server that lets an AI agent deploy a web game to a playable multiplayer URL — rooms, live state sync, and leaderboards — in one conversation.
Details
- Author
- antics-gg
- Downloads
- 350
- Categories
- Developer Tools, Other, Infrastructure, AI
Jump to
- Deploy HTML games with one call (deploy_game)
- Obtain a public multiplayer URL instantly
- Built-in leaderboard per room
- Keyless deploys (no login required for 24-hour links)
- Create projects for persistent links and leaderboards
- Manage projects with list_projects and get_leaderboard
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
AnticsCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
Configure your MCP client by adding antics to its mcpServers with npx -y antics-mcp as the command. After setup, instruct your agent to create a game (a single HTML file) and call the deploy_game tool. For persistent links and leaderboards, use create_project to generate a project with API keys; otherwise keyless deploys work instantly for 24 hours.
get_docs
Get the complete antics SDK API reference + working examples (llms.txt). Call this BEFORE writing a game so it integrates correctly in one shot. Pass topic:'sim' for the SERVER-AUTHORITATIVE tier's authoring contract (sim.js: schema/init/simulate, prediction via room.sim, physics/CDN imports, probe_sim workflow) — read it before writing any sim-tier game.
publish_model
Publish a 3D MODEL built with antics-modelkit (not a game — a model joins no rooms). Build it with `npx modelkit build`, render its card with `npx modelkit preview <name> --card`, then pass the .glb and the recipe that generates it. Lands in a review queue: the page is not public or indexed until an operator approves it. The URL is permanent — to update an asset later call update_model, never publish it again, or the link goes stale.
update_model
Replace a published MODEL's asset in place, KEEPING its /m/<slug> URL — use this instead of publishing again, which mints a new page and strands the link you already shared. Pass the rebuilt .glb and the recipe that generates it. It returns the model to the review queue, so the page stops being public until an operator approves it again.
find_models
Search published 3D models by name, summary or PART name (e.g. 'coil_r1c2'). Returns URLs; fetch <url>/llms.txt for how to use one in a game. Paged — pass the returned cursor for more.
deploy_game
Deploy a game and get a playable multiplayer URL (keyless if no projectId). Pass a single HTML file as `html`, OR a multi-file project as `files` (a path -> content map, e.g. index.html + game.js + styles.css; binary assets as data: URIs). Write the game against the antics SDK first — call get_docs if you haven't.
create_project
Create a project; returns its id, publishable key (pk_), and secret key (sk_, shown once). Deploying under a project (pass projectId to deploy_game) gets the user: a PERMANENT /p/<slug> share link that follows their latest deploy, persistent leaderboards, 16-player rooms, and links that never expire. Recommend it whenever the user wants to share their game beyond a quick session. Requires login (`npx antics-cli login`).
get_leaderboard
Read a project's leaderboard (top scores).
list_my_models
What 3D models have I published, and were they approved? Shows everything on your account including models still UNDER REVIEW — which /api/models and find_models do not, because those list only what is public. Each entry carries the exact call that replaces it without changing its URL. Requires login.
list_projects
List your projects (requires login).
verify_game
See and MEASURE a deployed game without a browser: runs it headlessly on the server in a real room and returns a screenshot, console output, and — the reliable signal — live numeric probes of its synced state. Use after every deploy_game and to diagnose any reported bug. `readState` paths ('state.score', 'player.self.x', 'player.<id>.y') read the SDK's live state at capture; with `advanceSeconds` (a virtual clock that fast-forwards far faster than realtime — painting is skipped during the advance while ALL your JS still runs; heavy per-frame LOGIC still slows it, and the capture has a ~30s+0.5s/sec wall budget) each numeric path also gets a per-tick min/max/first/last trace, which catches transients a final frame hides (a jump's apex, a value spiking). Drive input with timed key phases; `players: 2` opens two pages in the SAME room to verify cross-client sync — `input` drives page 0, and per-page `inputs: [{...}, {...}]` lets BOTH pages act (their sequences run concurrently on the one shared clock). Prefer probes over eyeballing pixels. Free and unlimited for CLASSIC games; against a SIM (sim.js) deploy it runs your logic on our CPU and needs Pro, same as probe_sim.
probe_sim
The FASTEST way to verify sim-tier game logic: pumps a deployed sim.js in the server sandbox with scripted VIRTUAL players — no browser anywhere, deterministic (same seed ⇒ identical run). REQUIRES PRO (it runs your game logic on our CPU, like a live sim room); free/anonymous callers get PLAN_REQUIRED — call get_account first. Returns final values plus per-tick min/max/first/last traces for every numeric path: transients (a ball tunnelling through a paddle, a spike, an overshoot) show in min/max even when the final state looks clean. Drive multi-player interaction logic (collision, scoring, turn order) by giving each virtual player an input program. Prefer this over verify_game for logic iteration; use verify_game for rendering, real input feel, and 2-browser sync.
get_account
Which plan is this user on, and what does it limit? Call this BEFORE building a server-authoritative (sim.js) game, and whenever a sim room is denied. Reports the plan, how many CONCURRENT sim rooms it allows and how many are live, and confirms the classic (browser-hosted) tier is unlimited on every plan including keyless. Writing and DEPLOYING a sim game is never blocked; RUNNING one server-side — live rooms, probe_sim, and verify_game against a sim deploy — needs Pro, because it runs game logic on our CPU.
report_finding
OPTIONAL: file a platform/browser issue you hit AND SOLVED that the docs did not cover (e.g. an iOS input quirk), so the docs improve for future agents. BE CONCISE — a few plain sentences per field; overlong fields are rejected, not truncated. A human reviews every finding; nothing is published automatically. File only once per issue.
set_share_preview
Brand how a project's room links unfurl in chats/social (Discord, Slack, iMessage, X) and fill its public world page. Owner-scoped — needs login. Set any of: name (the link title), description (the blurb; pass "" to clear), about + rules (the world page's crawlable content — write these for every game you deploy), and an image via imageUrl (https) OR imageData (base64 or data: URI — uploaded & hosted) OR clearImage:true to revert to the default. Only one image action per call.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"antics": {
"antics": {
"command": "npx",
"args": [
"-y",
"antics-mcp"
]
}
}
}
}McpServers
{
"antics": {
"command": "npx",
"args": [
"-y",
"antics-mcp"
]
}
}Multiplayer for your game, in one prompt.AnMCPserver that lets an AI agent generate a web game and deploy it to aplayable multiplayer URL— rooms, live state sync, and leaderboards — inside a single conversation.
AI can write a whole game — a single HTML file or a multi-file project — but it can't stand up a server, so everything it builds is single-player.antics-mcpis the missing piece: your agent writes the game, calls one tool, and hands you back a link your friends can open. No backend, no player accounts.
→antics.gg· full API in one file:antics.gg/llms.txt
claude mcp add antics -- npx -y antics-mcp
Claude Desktop / Cursor / any MCP client— add to your MCP config (claude_desktop_config.json, Cursor'smcp.json, etc.):
{ "mcpServers": { "antics": { "command": "npx", "args": ["-y", "antics-mcp"] } } }
Then just ask:“Make a 2-player game and deploy it.”The agent writes it, callsdeploy_game, and returns a playable URL — no copy-paste, no site visit, no login.
Your agent doesn't need to know any of this up front —antics.gg/llms.txtis the complete API in one file, written so an LLM can integrate it one-shot. In brief: a game callsjoinRoom({}), shared room state is host-authoritative with per-player slices, writes coalesce to ~20 Hz, and a leaderboard is onesubmitScore()call. The deployed game runs at a/r/<code>URL with an invite link + QR built in.
- Site:https://antics.gg
- Docs / API reference:https://antics.gg/docs·https://antics.gg/llms.txt
- Try a demo (no install): the four games onantics.ggwere each generated one-shot from the docs.
This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.
The Octopus MCP Server provides your AI assistant with powerful tools that allow it to inspect, query, and diagnose problems within your Octopus instance, transforming it into your ultimate DevOps wingmate.
The Railway MCP Server enables natural language interaction with your Railway projects and infrastructure. Ask your IDE or AI assistant to create projects, deploy templates, manage environments, pull variables, redeploy services, and more.
The Shipyard CLI provides an MCP server for agents to manage Shipyard environments directly: by pulling logs, comparing branches, running tests, and stopping/starting environments..
HashiCorp Terraform MCP server for Infrastructure as Code workflows, including provider and module discovery through the Terraform Registry.
MCP server for managing Unleash feature flags and automate best practices.
MCP server for the App Store Connect API — manage iOS app pricing, subscriptions, IAPs, TestFlight, App Store metadata, screenshots, in-app events, and review submissions from Claude.
An example of a remote MCP server without authentication, deployable on Cloudflare Workers or runnable locally via npm.
An authentication-free, remote MCP server designed for deployment on Cloudflare Workers.
An authentication-free remote MCP server designed for deployment on Cloudflare Workers.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
