Skills partykit

install
source · Clone the upstream repo
git clone https://github.com/TerminalSkills/skills
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/TerminalSkills/skills "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/partykit" ~/.claude/skills/terminalskills-skills-partykit && rm -rf "$T"
manifest: skills/partykit/SKILL.md
safety · automated scan (low risk)
This is a pattern-based risk scan, not a security review. Our crawler flagged:
  • references .env files
Always read a skill's source content before installing. Patterns alone don't mean the skill is malicious — but they warrant attention.
source content

PartyKit

Overview

PartyKit is a platform for building real-time, multiplayer, and collaborative applications. Each "party" is an isolated room running on Cloudflare's edge network — close to users for low latency. Unlike Socket.IO (which needs a persistent server), PartyKit runs serverless with automatic hibernation — you pay for active connections, not idle servers. Use it for collaborative editing, live cursors, multiplayer games, real-time polls, and any feature where multiple clients need synchronized state.

Instructions

Step 1: Setup

# Create new project
npm create partykit@latest my-party
cd my-party

# Or add to existing project
npm install partykit partysocket

# Start development server
npx partykit dev
# Server running at http://127.0.0.1:1999

Step 2: Define a Party Server

// party/index.ts — A PartyKit server (one instance per room)
import type * as Party from 'partykit/server'

export default class ChatRoom implements Party.Server {
  // Shared state for this room
  messages: Array<{ author: string; text: string; timestamp: number }> = []

  constructor(readonly room: Party.Room) {}

  onConnect(conn: Party.Connection, ctx: Party.ConnectionContext) {
    /** Called when a client connects to this room. */
    // Send existing messages to the new connection
    conn.send(JSON.stringify({ type: 'history', messages: this.messages }))

    // Notify others
    this.room.broadcast(
      JSON.stringify({ type: 'user-joined', connectionId: conn.id }),
      [conn.id]    // exclude the new connection
    )
  }

  onMessage(message: string, sender: Party.Connection) {
    /** Called when a client sends a message. */
    const data = JSON.parse(message)

    if (data.type === 'chat') {
      const chatMessage = {
        author: data.author,
        text: data.text,
        timestamp: Date.now(),
      }
      this.messages.push(chatMessage)

      // Broadcast to all connections in this room
      this.room.broadcast(JSON.stringify({ type: 'new-message', message: chatMessage }))
    }
  }

  onClose(conn: Party.Connection) {
    this.room.broadcast(JSON.stringify({ type: 'user-left', connectionId: conn.id }))
  }
}

Step 3: Client Connection

// hooks/useParty.ts — React hook for PartyKit connection
import usePartySocket from 'partysocket/react'

export function useChatRoom(roomId: string, userName: string) {
  const [messages, setMessages] = useState<ChatMessage[]>([])

  const socket = usePartySocket({
    host: process.env.NEXT_PUBLIC_PARTYKIT_HOST!,
    room: roomId,
    onMessage(event) {
      const data = JSON.parse(event.data)

      if (data.type === 'history') {
        setMessages(data.messages)
      } else if (data.type === 'new-message') {
        setMessages(prev => [...prev, data.message])
      }
    },
  })

  const sendMessage = (text: string) => {
    socket.send(JSON.stringify({ type: 'chat', author: userName, text }))
  }

  return { messages, sendMessage, readyState: socket.readyState }
}

Step 4: Live Cursors

// party/cursors.ts — Real-time cursor sharing (like Figma)
import type * as Party from 'partykit/server'

type Cursor = { x: number; y: number; name: string; color: string }

export default class CursorRoom implements Party.Server {
  cursors = new Map<string, Cursor>()

  constructor(readonly room: Party.Room) {}

  onMessage(message: string, sender: Party.Connection) {
    const cursor: Cursor = JSON.parse(message)
    this.cursors.set(sender.id, cursor)

    // Broadcast cursor position to everyone else
    this.room.broadcast(
      JSON.stringify({ id: sender.id, ...cursor }),
      [sender.id]
    )
  }

  onClose(conn: Party.Connection) {
    this.cursors.delete(conn.id)
    this.room.broadcast(JSON.stringify({ id: conn.id, gone: true }))
  }
}

Step 5: Deploy

# Deploy to Cloudflare's edge network
npx partykit deploy

# Custom domain
npx partykit deploy --domain my-party.example.com

# Environment variables
npx partykit env add OPENAI_API_KEY

Examples

Example 1: Add collaborative editing to a document app

User prompt: "Add real-time collaboration to our note-taking app — multiple users editing the same document with live cursors."

The agent will:

  1. Create a PartyKit server that manages document state and cursor positions.
  2. Use CRDT (Yjs) for conflict-free concurrent editing.
  3. Add cursor presence showing each editor's position and name.
  4. Deploy to Cloudflare's edge for global low-latency access.

Example 2: Build a live polling/voting feature

User prompt: "Add real-time polls to our presentation tool. Audience votes on their phones, results update live on the presenter's screen."

The agent will:

  1. Create a party room per poll with vote state.
  2. Audience connects via phone, casts votes.
  3. Presenter view receives real-time vote count updates.
  4. Results animate as votes come in.

Guidelines

  • Each room is an isolated instance — state is not shared between rooms. Use room IDs to partition data (one room per document, per game session, per chat channel).
  • PartyKit hibernates rooms with no active connections — you're not paying for idle rooms. State persists in the room object while connections exist.
  • For persistent state (survive room hibernation), use PartyKit's built-in storage (
    this.room.storage
    ) or an external database.
  • Use
    partysocket
    (not raw WebSocket) on the client — it handles reconnection, buffering, and the PartyKit protocol.
  • PartyKit deploys to Cloudflare's edge network — rooms run closest to the first connecting user, then all subsequent users connect to that location.