Framework Comparison

A detailed feature matrix for all 10 frameworks on Cloudflare Workers. Use this to make an informed decision, not just go with the most popular option.

Feature Matrix

Edge-Friendliness Ranking

Ranked by how well each framework fits the Cloudflare Workers runtime model (V8 isolates, no Node.js, binding-first architecture).

Tier S: Born for Workers

1. Hono - Native Worker. No adapters, no polyfills, no impedance mismatch. If Workers had a default framework, this would be it. Limited to APIs and server-rendered HTML.

2. React Router v7 - CF Vite Plugin, Worker entry point access, real bindings in dev. The closest thing to a full-stack Hono with React UI.

3. TanStack Start - Same integration depth as React Router v7, plus type-safe routing. Newer, smaller community.

Tier A: Strong Integration

4. Astro - CF Vite Plugin (adapter v13), real workerd in dev, zero-JS default. The best content framework on Cloudflare.

5. Vite + React SPA - CF Vite Plugin for asset config. No server code to worry about. Zero edge-runtime issues because there’s no edge runtime.

Tier B: Works Well via Adapters

6. SvelteKit - Solid adapter, good Cloudflare support, but dev runs in Node.js. You’ll catch some bugs only after deploying.

7. Nuxt - Nitro server engine handles Workers well. Same dev/prod gap as SvelteKit.

8. SolidStart - Works on Cloudflare Pages. Smaller community, fewer examples.

9. Qwik City - Resumability model is interesting but the Cloudflare adapter is less mature. Targets Pages.

Tier C: Compatibility Layer

10. Next.js (OpenNext) - Impressive feature coverage but fundamentally fighting the runtime. Node.js APIs shimmed, no workerd dev parity, community adapter with release lag.

Bundle Size Impact

Workers have a script size limit (3MB compressed free, 10MB paid). Framework server bundles vary significantly.

Gotcha: These are approximate ranges. Your actual bundle depends on dependencies. Check with wrangler deploy --dry-run.

Migration Paths from Next.js

Path 1: Keep Next.js (Least Effort)

Next.js app -> Add OpenNext -> Deploy to Workers

• Effort: Low
• Risk: Dev/prod parity gap, community adapter dependency
• Best when: Large existing codebase, can’t justify rewrite

Path 2: React Router v7 (Same Stack, Better Integration)

Next.js app -> Rewrite routes to React Router v7 -> CF Vite Plugin

• Effort: Medium-High
• Risk: Route migration, different data loading patterns
• Best when: Starting fresh or unhappy with Next.js lock-in

Path 3: Astro (Content Sites)

Next.js content site -> Rewrite pages to .astro -> Add React islands where needed

• Effort: Medium
• Risk: Different mental model (MPA vs SPA)
• Best when: Content-heavy site where most pages don’t need client JS

When NOT to Use Cloudflare

Be honest about the limits:

• Heavy compute: Image processing, video transcoding, ML inference exceeding Workers AI - use a traditional server
• Long-running processes: Workers have CPU time limits (10ms free, 30s paid per request)
• Large dependencies: If your node_modules pull in native bindings (sharp, bcrypt, canvas), Workers won’t run them
• Relational database needs beyond D1: D1 is SQLite-based. If you need Postgres features, you’ll need an external database and pay for latency
• WebSocket-heavy apps without Durable Objects: Standard Workers can’t hold persistent WebSocket connections. You need Durable Objects, which adds complexity