Cloudflare Workers

in-progress

Deploy Next.js to your own Cloudflare account — Workers + R2 + Hyperdrive + Queues + Vectorize + AI + KV + Durable Objects, all wired up from a single nextdeploy.yml. Currently experimental — see status table below.

Source reference: cli/internal/serverless/cloudflare*.go (12 files), shared/nextcompile/, CLOUDFLARE_PARITY.md, NEXTCOMPILE_ROADMAP.md

Why Cloudflare

TODO global edge by default, generous free tier, integrated data products (R2 / Hyperdrive / Queues / Vectorize / AI), no cold start.

When to choose Cloudflare over AWS

TODO edge-first traffic, want native data primitives, comfortable with Workers compatibility surface, willing to live with current limitations.

What works today

TODO API routes, /public from R2, passthrough middleware, Server Actions (CSRF + JSON), pure-server RSC pages, image optimization, basic revalidate.

What does NOT work yet

TODO RSC pages with "use client" components, Suspense streaming, scheduled() exports, Hyperdrive binding injection, full middleware semantics — see CLOUDFLARE_PARITY.md.

Cost shape

TODO Workers pricing (requests + duration), R2 (storage + class A/B operations), Hyperdrive (per query), Queues (per operation). Compare to AWS for typical app.

Architecture at a glance

TODO diagram — local build → NextCompile → ESM worker → wrangler-style upload → R2 (assets) → Workers runtime + bindings → user.

How NextDeploy compares to wrangler

TODO wrangler is the underlying tool — NextDeploy adds metadata-driven detection, auto-provisioning, secret management, and unified config across VPS / AWS / Cloudflare.

Where to go next

TODO quickstart for first deploy, workers page for compile internals, bindings page to see what you can wire up.