This source did not publish a separate summary. Review SKILL.md before using the skill.
SKILL.md
TanStack Start (React) — RC-Ready Playbook
Full-stack React on TanStack Router with per-route SSR/CSR, file-based routing, server functions, and first-class Cloudflare Workers support.
Use this skill when
Building a greenfield React app that needs route-level SSR/CSR/SSG switches.
Migrating from Next.js/React Router while keeping file-based routing + API routes.
Shipping to edge runtimes (Workers) with typed server functions and bindings.
You want predictable routing with type-safe params/search + built-in preloading.
What’s inside
References: quickstart/layout, rendering modes, server functions, Cloudflare hosting, execution/auth, plus new routing/data/navigation/devtools guides.
Troubleshooting: hydration, API routing, bindings, navigation/preloading failures.
Quick Start (React)
npm create @tanstack/start@latest my-app
cd my-app
npm run dev
Manual installs (all bundle targets are supported): add @tanstack/react-router + @tanstack/react-start with your bundler plugin (vite, webpack, or esbuild) per the official install guides.
app/config.ts or app/start.ts sets defaultSsr, spaMode, middleware, and context.
Routing + Data Best Practices
Type-safe params & search: createFileRoute() infers path params; add validateSearch (zod) to parse and coerce search params.
Route matching order is deterministic (index → static → dynamic → splat); rely on this when adding catch-alls.
Loaders run once per location change; return plain data, throw redirect()/notFound() for control flow.
Data mutations: colocate action/server functions; keep loaders read-only and invalidate via router.invalidate() after mutation.
TanStack Query bridge: create a QueryClient in router context and ensureQueryData inside loaders to dedupe fetches.
Deferred/external data: stream partial data or read from external loaders; prefer suspense-friendly responses.
Head management: set head per route for <title>/meta; derive from loader data to keep SEO consistent.
Not-found/auth: throw notFound() or redirect() in loaders/middleware; use error boundaries for UX.
Example route (typed search + data-only SSR):
// app/routes/posts.$postId.tsx
import { createFileRoute, redirect } from '@tanstack/react-router'
import { z } from 'zod'
export const Route = createFileRoute('/posts/$postId')({
validateSearch: z.object({ preview: z.boolean().optional() }),
ssr: 'data-only',
loader: async ({ params, search, context }) => {
const post = await context.queryClient.ensureQueryData(['post', params.postId], () =>
fetch(`/api/posts/${params.postId}?preview=${!!search.preview}`).then(r => r.json())
)
if (!post.published && !search.preview) throw redirect({ to: '/drafts' })
return { post }
},
})
Navigation, Preloading, and UX
Link prefetch defaults: <Link preload="intent"> (hover/focus) preloads route data/code; use preload="render" for above-the-fold routes.
Programmatic preloading: router.preloadRoute({ to, search }) to warm caches before navigation (e.g., on visibility).
Route masking: keep canonical URLs while showing user-friendly masks (e.g., /products?slug=abc masked as /p/abc).
Navigation blocking: protect unsaved forms with router.navigate({ to, replace, from }) blockers or useBlocker.
Scroll restoration: enable scrollRestoration to restore positions on back/forward; customize per route when using long lists.
Search param serialization: customize parse/stringify to keep numbers/dates stable and avoid stringified booleans.
Rendering & Performance
Per-route SSR: set ssr: true | false | 'data-only' on routes; defaultSsr config sets the baseline.
Code-splitting: file-based routes auto-split; add lazy/load for manual chunks on code-based routes.
Preloading strategy: pair preload="intent" links with defaultPreloadStaleTime to avoid over-fetching.
Render optimizations: keep loaders pure, memoize heavy components, and use pendingComponent for CSR routes to avoid layout shift.
Devtools, Linting, and LLM Support
Add <RouterDevtools /> during development to inspect matches, loader states, and preloading.
Enable the ESLint plugin @tanstack/eslint-plugin-router with the recommended config to enforce inference-sensitive property order (e.g., beforeLoad before loader).
LLM-aware routing: the Router exposes structured route metadata to LLM agents; keep descriptions concise in Route meta for better AI navigation.
Deployment Notes (Cloudflare-friendly)
Keep cloudflare({ viteEnvironment: { name: 'ssr' } }) first in Vite plugins so bindings reach server entry.
Regenerate bindings after changes: npm run cf-typegen.
For static-heavy sites, enable prerender to ship HTML to Workers Assets/Pages; exclude param routes or add explicit pages.
Ship Checklist
Routes load without hydration warnings (prefer ssr: 'data-only' for non-deterministic UI).
Search params validated with validateSearch and custom serializer where needed.
Link preloading configured for high-traffic routes; blockers added for unsaved forms.
ESLint plugin enabled (create-route-property-order rule) and npm run check passes.
Devtools verified locally; router.matches state looks correct.
Cloudflare bindings typed (cf-typegen) and streaming tested via curl -N.
