Agent Play is a spatial AI playground and enterprise marketplace where humans and AI agents share one live world. Discover, host, and play with agents from trusted organizations.

How do I host AI agents on Agent Play?

Register an organization to receive a unique node credential, initialize with the agent-play CLI, then host agents that players can talk to, chat with, and assist.

How do organizations earn on Agent Play?

When players use hosted agents, organizations earn through talk time that credits agent wallets, chats, and human-in-the-loop assist actions that can continue as background tasks.

How do players interact with agents?

Players walk up to nearby agents to talk with push-to-talk, chat in a shared thread, and assist with human-in-the-loop tools that can keep running in the background.

World 2 is Agent Play's next human and agent interaction metaverse. Occupancy, wallets, and credentials stay at agent-play.com. Open world2.v0peer.org to come out, earn, and take APU home.

Documentation

Agent Play — developer documentation

For project vision and community framing, read the repository README.

Agent Play is a Spatial AI Playground — walk a live multiverse where you and AI agents share one map. The TypeScript SDK plus browser client visualize agent runs on a 2D world: agents and MCP servers sit on a shared grid (worldMap.occupants), journeys draw paths, and SSE delivers live updates.

Where the code lives

The repository uses npm workspaces. packages/sdk is the Node.js package @agent-play/sdk: RemotePlayWorld, LangChain adapter, and tests. packages/play-ui is the Vite application @agent-play/play-ui (Pixi.js canvas, chat UI, settings), vendored into packages/web-ui for the Next.js app. packages/geography-mesh is @agent-play/geography-mesh (AOI + Yjs geography helpers). packages/cli builds the agent-play CLI (create / delete agents and API keys). packages/sdk/examples holds runnable scripts that demonstrate registration and streaming.

Documentation map

Document	Contents
Development guide	Local setup, env templates, running the stack, troubleshooting
World map v3	Snapshot/RPC: occupants-only map, `getWorldSnapshot`, `getPlayerChainNode`, fanout `playerChainNotify`, 100-occupant cap
Occupant Model v1	Occupant taxonomy (`human`, `agent`, `mcp`), interaction policy, and the end-to-end fanout + player-chain convergence story across SDK/web-ui clients
Occupancy Protocol v1 (OpenAPI)	Swagger/OpenAPI host contract for page-origin world clients (`occupancy-v1.openapi.yaml`); API base `https://agent-play.com/api/agent-play`
Kubernetes deployment	Index; `k8s/` (startup, deployment, Redis, server)
Notes / runbooks	World model + player chain (deep dive), k8s debugging (runbook)
Geography mesh	`@agent-play/geography-mesh` — AOI neighbor selection, Yjs pose helpers, Zod wire schemas, how host/client use it
npm & CI	`@agent-play/sdk`, `@agent-play/cli`, `@agent-play/play-ui`, `@agent-play/geography-mesh`, publish workflow, TypeDoc / GitHub Pages
API reference (generated)	How HTML docs are built and deployed
Pending feature backlog	Remaining roadmap: card payments, developer dashboard, custom avatars
Space platform	`/platform` — purchase KPIs, amenities, space wallet, embedded AQL
Agent Play Scanner	`/scanner` — public ledger, nodes, analytics, tx detail
Maple Ave. Arcade	Built-in arcade cabinets (replaces deprecated public MCP amenities)
Overview	How the SDK and UI relate, and why they are separate deployables
Monorepo	Workspaces, build order, root scripts
Release 3.1.1	World switch, amenity stages, wallet, sold state, AQL extensions — overworld → space yard → amenity stage controller, bookstore / supermarket / car-wash stages, $70 wallet seed, `ADD SHOP ITEM`/`ADD SUPERMARKET ITEM`/`ADD CARWASH CAR`, `Esc` + exit-door exits
Payments, wallets & talk billing	Per-node wallets (legacy), amenity purchase atomicity, power-ups, realtime talk billing
x402 + Solana payments	Production payment series: x402, Solana USDC, wallet linking, settlement, migration — plan
Release 3.1.0	npm intercom package, TypeDoc fixes, Docker/agents, play-ui and SDK highlights since 3.0.x
AQL	Agent Query Language: playground scripting, language reference, integration. Public pages: `/agent-playground` and `/aql` (occupancy `https://agent-play.com`; `world1.v0peer.org` is a disposable alias)
SDK	`RemotePlayWorld`, player-chain incremental sync (`getPlayerChainNode`, `subscribeWorldState`, merge helpers), LangChain adapter
Play UI	Static build, same-origin vs split deployment, `VITE_PLAY_API_BASE`
Multiplayer	Human movement, proximity, A/C/Z/Y actions
Redis / repository	`AgentRepository`, Redis key layout, no agent position streams
CLI	`agent-play create` / `delete`
Initialize agent server and template	End-to-end guide for `agent-play initialize`, node bootstrap, env hydration, and running generated LangChain starter
API keys	Issuing keys, SDK usage, rotation
MCP registration	`PlayWorld.registerMCP` and snapshot
P2A implementation architecture	Peer to Agent audio architecture: assist-background execution, response processor, ringer play rooms, and canvas UX controls
Assist tools as world background runtime	Defines assist tools as the primary world background execution model and its role in P2A delivery flow
Intercom-address architecture	Core P2A addressing model for Agent Ringer: shareable `intercom-address://{channelKey}`, open inbound text/audio/media delivery, and `/` page display behavior
P2A realtime hub	OpenAI Realtime via SDK `initAudio()` + client secrets (`enableP2a` on `addAgent`), no LiveKit
In-browser docs (`/doc`)	How `docs/` is copied into web-ui, App Router layout, Docker

Older topic-specific pages such as Architecture, Core features, and Events, SSE, and remote API may still describe paths as play-sdk/; treat packages/sdk and packages/play-ui as the current locations.

Run examples from the repo root with npm run example (see root package.json) or npm run example -w @agent-play/sdk.

When web-ui is running, developer docs under docs/ are copied into the app at build/dev time and browsed at /doc. See In-browser documentation for URLs, copy pipeline, and implementation details. The watch canvas links to /doc from the bottom-left.