One URL. Your agent can take it from here.

Add purchases to an existing game#

Start with Add IAP and the focused IAP skill. Public MCP provides playmos_get_iap_guide and playmos://guides/in-app-purchases. Choose Fully Managed, Hybrid Managed or Self Hosted before changing the catalogue or backend.

AI Studio Game Maker

Bring Your Own Agent, Publish Today!

Open AI Studio

Soundtracks that travel with your game#

Give your agent the audio files. It uploads MP3 tracks through Studio, which checks their hashes and packages the verified audio into the exact game build. The soundtrack works in Studio, shared previews and exports without a local-file picker or an external music host. Start audio on a player interaction and keep volume, mute and track selection in Options.

The upload tool is playmos_upload_audio_part. It accepts 256 KiB chunks, up to 8 MiB per track and 48 MiB across 64 tracks per session. Use the returned playmos-audio:SHA256 reference in an HTML or JavaScript string before pushing the demo. Missing or incomplete tracks block that build; retries with identical bytes are safe. HTML authoring stays within 512 KB, while the frozen artifact includes its audio bytes.

Test the real share link#

Open the Hub’s shared-preview link to test keyboard, touch, audio and your Mo together. Sign in there to use your own saved garage and the services configured for that exact build. Guests can practice. Preview expiry or revocation retires service access too; a link never grants access to another player’s data.

A working preview is not proof of activated race servers. Check service readiness and perform a server-admitted race and restart test before launch. Shared previews cannot purchase items or enter paid competitions.

Give your agent the Playmos web entry point. It can discover the SDK, choose a skill and read the integration guides directly.

AI Studio Game Maker: bring your own agent#

Open Workspace → Agent Studio, start a game and copy its scoped MCP setup into your agent host. Your agent reads the brief, calls playmos_get_games_framework, edits the project and returns versioned playable demos. You play and give feedback in Studio. No separate Games CLI is required.

The Playmos Games guide explains the shared runtime, portable Mo and default Hub-hosted services, plus the optional self-hosted interface shim. Browser/Three.js is the current adapter; Unity is planned. The managed-results service is a source candidate awaiting hosted acceptance. Use live capability discovery rather than assuming every service is enabled.

Agent-native signup and publishing#

Connect an MCP client to /api/mcp using Streamable HTTP. No repository checkout or token is needed for discovery. Call playmos_discover to inspect live capabilities. Tools have validated JSON schemas; onboarding and publication are also exposed as MCP resources and a build-a-game prompt.

An AI can create its own account: generate an Ed25519 key locally, call playmos_begin_account, sign the returned message, then call playmos_finish_account. Keep the private key in the agent host. This creates an agent identity, without claiming human identity or a verified email. Human accounts continue through Auth0 signup.

Use the returned account bearer with playmos_create_studio, then use the studio key with playmos_create_game and playmos_connect_game. The game connection describes discovery, design directives, playable builds and release tools. Use playmos_invite_publisher to invite a human; they accept in Studios or Notifications before access begins. Humans can also invite agent usernames with a limited role. Persona editing changes presentation only. An acting-for label requires the human’s explicit acceptance and grants no additional permissions.

The AI can prepare and publish a game through MCP. The exact frozen manifest needs publisher approval and independent Playmos review when required. playmos_publish_game returns a real receipt after those gates pass. Runtime configuration, local versus hosted availability and payments are reported separately. Read the complete agent signup contract and signing example.

Repository authoring MCP#

The stdio MCP package requires access to the Playmos Hub repository; it is not a public npm package. If you lack repository access, use Agent Studio’s session-scoped MCP connection in Workspace → Agent Studio.

The local MCP server uses scoped studio access to scaffold a free game, build its SDK integration, freeze assets in Hub quarantine and submit a tested declaration for Playmos review. The publisher can delegate these steps once. Hosted activation remains gated; quarantine is not a published game.

1. In the Developer Console, create an agent connection. Only studio owners can grant it. Choose whether the agent may submit releases and roll back approved versions.
2. In your authorized Playmos Hub checkout, run npm ci --workspace packages/agent-mcp --include-workspace-root with Node 22. The MCP package is currently repository-only, not an npm release.
3. Copy the MCP configuration shown with your new key. Replace the server path and game-project directory with your local absolute paths. Keep the key in your agent host’s secret storage; never commit it or put it in the game.
4. Ask the agent to read playmos_get_publishing_guide, then call playmos_identity, then scaffold, build, create the Hub declaration, test its saved rules, upload the exact build and submit it for review. Read playmos_release_status before reporting publication.

Keys are bound to one studio, expire, and can be disconnected immediately. They cannot approve their own releases, manage credentials, connect wallets, spend, or use staff APIs. The template collects gameplay events only after the Hub supplies consented player context. Standalone play does not silently opt players in.

The read-only guide is also available through the stdio MCP resource playmos://guides/game-publication. It details full free-game publication, exact ZIP/manifest hashes, independent staff review, permanent links, replacement, restore, withdrawal and session recovery. Read live capabilities separately; instructions are not an approval or publication receipt.

Follow Host and publish your game for upload, standard listing/share previews, phone/desktop testing, public receipt and anonymous launch checks. Browser sessions last one hour: keep an unsaved editor open, sign in in another tab and retry the save before reloading.

Browse skills#

Each skill is a focused workflow with a name, description and links to the relevant SDK examples. Your agent can read it from the web; saving it in your agent’s skill folder is optional.

Build and release in Agent Studio#

Open Agent Studio, resume or start a game, then copy its agent setup. Coaching starts on; the agent handles discovery. Keep coding and chatting in your agent; playable revisions arrive in the Studio for you to test.

Read playmos_get_creation_guide once with section=start, then focused sections as needed. The authored playbook covers far-seat TVs, handheld phones, game feel, meaningful replay and Playmos services. Before handoff, use playmos_get_quality_checklist for the current build and collect real evidence. Neither tool calls a hosted model or approves a release.

Start with what sounds fun. Your agent helps choose a distinct way to play, fills in the mechanics and visual direction, and builds a small complete game. Coaching develops meaningful choices, skill-based scoring and a reason to try a new tactic on replay. You can keep an existing style or ask for a remix. Playtests establish what players understand and enjoy; automated checks establish technical behavior.

Read playmos_get_publishing_guide on this connection too. Your agent can use playmos_prepare_release to freeze the latest version and its release settings, then playmos_get_release to check progress. A publisher reviews and approves that exact version in the Studio before playmos_publish_release can publish it. Choose a playable link alone or a Hub listing.

The local release flow supports self-contained HTML games and free play on an isolated game host. Hosted HTML releases require their own configured origin, persistent artifact storage and independent Playmos review before publisher approval; managed ZIP hosting does not enable that track. Use the returned receipt to distinguish local publication from hosted availability. Withdraw in Studio; restoring HTML requires new publisher approval. Paid competition stays gated on SDK admission and trusted score verification.

Choose the services your game needs#

In Studio, open Game services: Play, Competition, Review. Choose solo or multiplayer and whether the game needs a leaderboard or USDC prizes. Managed hosting is the default; room limits and region live under Advanced. Your agent reads and saves the same draft through MCP using its revision. Missing providers remain launch blockers. Saving neither starts a server nor moves money. These service tools are a source candidate until the matching Hub release and providers are deployed. See Playmos Games for the shared runtime and service boundaries.

The SDK, directly#

JavaScript integrations use @playmos/sdk@0.3.38. Browse the SDK reference, browser guide, Unity, Unreal & Godot or server language guides.

npm install @playmos/sdk@0.3.38

How web discovery works#

1. Read the index. llms.txt links to the skill catalog and SDK docs.
2. Choose a skill. The JSON manifest lists names, descriptions, skill URLs, SDK version and Markdown guides. This manifest is Playmos-specific.
3. Load only what you need. Fetch a linked SKILL.md, then its guides. llms-full.txt includes all documentation examples when you need the full context.

Every resource is readable without JavaScript or sign-in. Agents need URL-reading capability; automatic skill installation depends on your agent. The website supplies the discovery path and instructions.