Realtime chat on Vercel without Pusher or Ably

Why “just add Pusher on Vercel” is the default

Vercel’s own guides and third-party tutorials often route teams to Pusher Channels or Ably when they need live features. That works, but you add a second realtime vendor, another bill, and another region to reason about.

Builders also hit WebSocket limits and pricing friction when they try to keep everything on serverless functions alone.

Recommended architecture split

• Vercel — Next.js App Router, auth UI, billing pages, marketing.
• Cloudflare Worker + Room DO — WebSocket upgrade, room fan-out, presence.
• D1 — message history and tenant metadata on your CF account.
• Browser — @fluxy-chat/sdk pointed at your Worker URL; mint JWTs from a Vercel Route Handler or API route.

FluxyChat vs Pusher on Vercel

• No Channels app or Pusher keys in the frontend — member JWT from your backend only.
• Room-per-DO isolation instead of a global channel namespace you wire yourself.
• Reconnect, loadMore history, and delivery status in the SDK — not only pub/sub events.
• Optional MIT self-host on your Cloudflare account when hosted beta is not enough.

Migration sketch (Pusher → FluxyChat)

• Deploy FluxyChat Worker (hosted beta or self-host) and create a project + API key.
• Replace Pusher client subscribe with FluxyChatClient + useChat({ roomId, client }).
• Map channel names to roomIds; persist history via REST/D1 instead of Pusher-only caches.
• Mint per-user JWTs server-side on Vercel; never expose fc_... keys to the browser.

Build vs buy on the same stack

You can still DIY a Durable Objects chat repo on GitHub — FluxyChat is for teams that want multi-tenant JWT, RoomDurableObject, D1 history, and operator console without stitching a second vendor. See the DIY comparison on /compare.