urls.bid: three frontend migrations in a week (and why)

urls.bid is a wallet-native link platform: custom-domain short links and gated content/downloads that return HTTP 402 and unlock via USDC on Solana and Base, using the x402 protocol. In under two weeks the frontend moved through three frameworks — not for fun, each move fixed something.

The migration ladder

From	To	Why
—	Next.js	fastest way to get a full-stack app standing
Next.js	Astro 6	static-first, tiny payload for mostly-static pages
Astro 6	TanStack Start	the app kept wanting real server logic at the edge

Astro was the right call for a marketing-heavy surface, but urls.bid is an app: payment-gated links, on-chain ownership, webhook receivers. Those want first-class server functions, typed loaders, and route-level auth — which is exactly TanStack Start’s shape.

What actually moved

The TanStack migration was mostly deleting API routes and letting server functions absorb them, then wiring QueryClient caching around the slow part — on-chain reads:

export const purchaseLink = createServerFn({ method: "POST" })
  .validator(purchaseSchema)
  .handler(async ({ data }) => {
    // settle USDC via x402, record on-chain ownership, flip the link public
  });

The payment layer itself is published separately as @codingstark/x402-elysia, and deploy is GitHub Actions → Cloudflare Workers.

The takeaway

“Three migrations” sounds like thrash, but each was a deliberate trade: reach for the framework whose default shape matches your app, and stop fighting the ones that assume otherwise.