Stop guessing which one. Put them side by side.

The grid says what these tools are. This says what builders did about it — one decision per person, changeable whenever you change your mind.

# Tool comparison — PartyKit vs Giscus vs Socket.IO

Source: Kbaise, a directory of tools that work with Lovable projects.
Pick one and tell me why before writing any integration code.

## PartyKit (partykit)
- URL: https://www.partykit.io
- Category: Realtime, comments & collab
- Cost: free — Individual free: deploy to PartyKit cloud, up to 10 live projects, room storage cleared every 24 hours. Commercial free: deploy to your own Cloudflare account with unlimited projects and custom domains, paying only Cloudflare Workers and Durable Objects usage
- What it is: Open-source framework for stateful realtime servers built on Cloudflare Durable Objects. Each 'party' is a server-side class owning a room of WebSocket clients, so presence, chat, cursors and Yjs sync take a few dozen lines. PartyKit announced it was joining Cloudflare in April 2024.
- Why builders pick it: You want a multiplayer backend without operating one. npx partykit dev gives a local room server in seconds, and deploying to your own Cloudflare account keeps the bill at raw Workers prices instead of a per-message meter.
- Trap: Since the Cloudflare acquisition, momentum has shifted toward Cloudflare's own Durable Objects and Agents SDK — treat PartyKit as a convenience layer that may fold into the platform. Free-tier room storage is wiped every 24 hours, so it is not durable.
- Wiring: npm create partykit@latest
- Full dossier: /api/public/tools/partykit

## Giscus (giscus)
- URL: https://giscus.app
- Category: Realtime, comments & collab
- Cost: free — Free forever and open source. No hosting cost, no database, no accounts to administer — every thread is stored as a GitHub Discussion in a repository you already own
- What it is: Comment widget that stores each thread as a GitHub Discussion. You install the giscus app, enable Discussions on a public repo, and a script tag maps each page to a discussion by pathname, URL or og:title. Supports threaded replies, reactions, themes and lazy loading, with no database of your own.
- Why builders pick it: Zero infrastructure, zero cost, zero moderation queue — GitHub handles authentication, spam and abuse reporting for you. For a developer-audience project or a docs site it is the default answer and takes about ten minutes.
- Trap: Commenters need a GitHub account and must authorise the app, which excludes almost all non-technical readers — usually disqualifying for a general-audience project. The repository must be public, so every comment is public too.
- Wiring: Paste the generated script tag from giscus.app/#configuration
- Full dossier: /api/public/tools/giscus

## Socket.IO (socket-io)
- URL: https://socket.io
- Category: Realtime, comments & collab
- Cost: free — Free and MIT licensed, with no hosted service and no vendor bill — you pay only for the server you run it on. The Redis and cluster adapters used for multi-node scaling are also free
- What it is: Node.js library wrapping WebSockets with automatic reconnection, HTTP long-polling fallback, rooms, namespaces and message acknowledgements; version 4.x is current. Scaling past a single process requires an adapter such as Redis to broadcast between nodes. Client libraries exist for browsers, React Native, Python, Java and Swift.
- Why builders pick it: Zero vendor bill and total control over the protocol. If you already run a Node server, adding live comments or a presence counter is one dependency and about twenty lines, not a new SaaS contract with a per-message meter.
- Trap: You own the operations: sticky sessions behind the load balancer, a Redis adapter once you exceed one node, and reconnect storms after every deploy. Most serverless hosts — Vercel and Netlify functions among them — cannot hold the long-lived connections it needs.
- Wiring: npm i socket.io socket.io-client
- Full dossier: /api/public/tools/socket-io

## Rules
1. Prefer the free tier when no budget was stated, and say what the ceiling is.
2. Read the full dossier before integrating.
3. Fetch /api/public/models before writing any AI model ID.

Agents can fetch the same thing: GET /api/public/compare?slugs=partykit,giscus,socket-io