Running a Next.js Site on Cloudflare Pages

14 Jul 2021 | 10 min read

next.js on cloudflare pages

JAMstack Overflow

These days there are so many providers for JAMstack hosting, it's hard to choose. Netlify, GitHub Pages, Vercel, Heroku, ... the New Dynamic has a list of over 30 different hosting/deployment tools, with new ones being added regularly. However, the new kid in town caught our attention: Cloudflare Pages. They have only been around for a few months now, but since we already use Cloudflare for DNS and CDN, consolidating tools could be a nice win.

Cloudflare Pages radically simplifies the process of developing and deploying sites by taking care of all the tedious parts of web development. Now, developers can focus on the fun and creative parts instead. –

Next.js on Cloudflare

Given they're just starting up, the features are still a bit limited. One of the biggest drawbacks as of this moment is the lack of a server to generate dynamic content. Currently, Cloudflare uses the Next.js static HTML export to prerender all pages to plain html. Given we don't use any server-side rendering capabilities at the moment, this was good enough to move forward now, but we are also excited to see the further development of Cloudflare Pages.

Getting started

Let's create a project:

cloudflare create a project first screen

Assuming you have a Next.js website and both next build and next export run through without any errors (you'll probably see an error when using next/image, we'll get to that in a moment), make sure everything is committed and pushed to a repository. In addition, make sure you create alias scripts in package.json for build and export:

"build": "next build",
"export": "next export",

This is necessary to run post-build scripts to generate additional content such as a sitemap.xml, robots.txt, RSS/Atom feeds etc.

Log in to your Cloudflare Dashboard and head to "Pages" on the right. "Create a project" and connect to your repository.

When you're in the build configuration section, they offer "Next.js (Static Export)" as a framework preset, but since this preset uses next commands by default, pre/post build hooks from our package.json are ignored. Instead, do not select a preset and configure the build options manually. The Cloudflare "build command" should be:

npm run build && npm run export
# or
# yarn build && yarn export

Similarly, the static files will be exported to a directory called out, so set the "build output directory" field to this. Your input should look like this:

cloudflare build settings screen

Once saved, Pages will automatically initiate the first build. Once the build is successful, you can then head to the * URL that will be shown at the top of the page. In a Pull Request, Cloudflare comments the build status and preview url.

If you are already using Cloudflare DNS, you can connect your custom domain in the next step—Cloudflare will automatically generate the CNAME for you with a click on the "activate domain" button:

cloudflare connect custom domain

Congratulations, you're now running your Next.js site on Cloudflare Pages! Now, the fine print.


The Problem

If you switched to the new next/image component in Next.js 11, you'll see the following warning during export:

Error: Image Optimization using Next.js' default loader is not compatible with `next export`.
Possible solutions:
- Use `next start` to run a server, which includes the Image Optimization API.
- Use any provider which supports Image Optimization (like Vercel).
- Configure a third-party loader in `next.config.js`.
- Use the `loader` prop for `next/image`.
Read more:

Since we want to do image optimization and we don't want any new tools, we've decided to look at another Clouflare option—a Cloudflare Worker.

Create a Worker

To create one in your Cloudflare Dashboard, go to "Workers" > "Create a Worker". Cloudflare Docs has an entire article about Resizing Images with Cloudflare Workers, and prepared the script for you to copy into the {} Script box:

addEventListener('fetch', (event) => {
async function handleRequest(request) {
let url = new URL(request.url)
let options = { cf: { image: {} } }
if (url.searchParams.has('fit')) = url.searchParams.get('fit')
if (url.searchParams.has('width')) = url.searchParams.get('width')
if (url.searchParams.has('height')) = url.searchParams.get('height')
if (url.searchParams.has('quality')) = url.searchParams.get('quality')
const imageURL = url.searchParams.get('image')
const imageRequest = new Request(imageURL, {
headers: request.headers
return fetch(imageRequest, options)

In the top right, you can change the name of the worker. Once saved, note down the URL, we'll need that.

cloudflare image worker preview

Configure the Loader

As the error gives us possible solutions, unfortunately we can't "Configure a third-party loader in next.config.js." - there is a small list of pre-existing loaders, but Cloudflare isn't one of them and they're no longer adding new default loaders. So we're using the loader prop for 'next/image':

// replace [yourprojectname] and [] with your actual project name and (custom) domain
const cloudflareImageLoader = ({ src, width, quality }) => {
if (!quality) {
quality = 75
return `https://images.[yourprojectname]${width}&quality=${quality}&image=https://[]${src}`

And the <Image>:

const MyImage = (props) => {
return (
alt="Picture of the author"

This would be too cumbersome to add this loader to every single <Image> tag you have in your project. So we created a custom Image component (/components/Image.jsx):

import Image from 'next/image'
// replace [yourprojectname] and [] with your actual project name and (custom) domain
const cloudflareImageLoader = ({ src, width, quality }) => {
if (!quality) {
quality = 75
return `https://images.[yourprojectname]${width}&quality=${quality}&image=https://[]${src}`
export default function Img(props) {
if (process.env.NODE_ENV === 'development') {
return <Image unoptimized={true} {...props} />
} else {
return <Image {...props} loader={cloudflareImageLoader} />

Node 16 Hiccup

If you're running on Node 16, you may already have had issues with Images on Next.js, and because we don't need image optimziation during development, we've removed the loader and added unoptimized={true} for the development environment.

Now you can search/replace all import Image from next/image to import Image from components/Image. To resolve the import path components/Image, you need to add a jsconfig.json to your project so that components/Image will resolve to ./src/components/Image:

"compilerOptions": {
"baseUrl": "."

Once you've done all that, you'll probably still see the same error during next export, so let's add a fake third-party loader into next.config.js (this fake loader will not be used—it's just a hack to avoid the build error):

images: {
loader: 'imgix',
path: ''

npm run export should now run successfully without any errors!


Cloudflare Pages also offer a GitHub integration that provides Pull Request previews, posting a comment to each Pull Request with the deployment status, which is extremely useful. (Other services provide this as well, for example, Vercel.)

However, one thing that was missing: a direct link out to the preview. In order to retrieve the preview URL, we built a GitHub Action Cloudflare Preview URL that waits for the deployment to be ready and then returns the URL. This is useful to run E2E tests, URL link checks etc. on a real website before going live.

There's room for improvement though:

  • build times are still very long (2+ minutes to initialize the build environment).
  • more flexibility with environment and build variables to expose system/build related information in user-defined variables.
  • use of the GitHub Deployments API, which fires an event when a deployment is ready, instead of polling the Cloudflare API.


You can use next-sitemap to generate a sitemap for all your pages automatically after build. Follow the README and add the next-sitemap.js as well as the post-build hook. This works out of the box with static site generation and the sitemap.xml and robots.txt will be copied into the export (/out) folder during npm run export. You should add both files to .gitignore to prevent them being added to the repository.

RSS Feeds

If you publish regularly and you want to give your audience a way to subscribe to your blog, RSS/Atom is the standard format for that. Most tutorials you'll find about RSS generation require server-side rendering from your dynamic content, though. But we can also solve this with a post build hook.

First, we need a script to generate the feed from our articles. We put this in utils/generate-rss.js. We use feed to help generate the XML and JSON files for RSS and Atom.

import fs from 'fs'
import { Feed } from 'feed'
import getPosts from 'utils/getPosts'
// site.js exports the default site variables, such as the link, default image, favicon, etc
import meta from 'content/site.js'
async function generate() {
const feed = new Feed({
title: meta.title,
description: meta.description,
image: meta.image,
favicon: meta.favicon,
copyright: meta.copyright,
language: meta.language,
feedLinks: {
json: `${}feed.json`,
rss2: `${}feed.xml`,
atom: `${}atom.xml`
// we store blog articles in content/articles/article.mdx
// you can change the path and regex here for your project.
const posts = ((context) => {
return getPosts(context)
})(require.context('content/articles', true, /\.\/.*\.mdx$/))
posts.forEach((post) => {
title: post.title,
id: `${}${post.slug}`,
link: `${}${post.slug}`,
date: new Date(,
description: post.description,
image: `${}${post.coverImage.src}`
fs.writeFileSync('./public/feed.xml', feed.rss2())
fs.writeFileSync('./public/feed.json', feed.json1())
fs.writeFileSync('./public/atom.xml', feed.atom1())

This is a little tricky. As you can see we're using require.context which isn't available outside the Next/Webpack environment. We're using a modified webpack config in next.config.js to compile the script and put it into the build directory:

webpack: function (config, { dev, isServer }) {
if (!dev && isServer) {
const originalEntry = config.entry
config.entry = async () => {
const entries = { ...(await originalEntry()) }
entries['utils/generate-rss.js'] = 'utils/generate-rss.js'
return entries
return config

Finally, we'll need to run the script after build (another post-build hook). In order to run this in parallel with the sitemap generation and keep everything neat and tidy, we're using npm-run-all:

"export": "next export",
"build": "next build",
"postbuild": "run-p generate:sitemap generate:rss",
"generate:rss": "node ./.next/server/utils/generate-rss.js.js",
"generate:sitemap": "next-sitemap",

You can now add the feeds into your _document.js <Head>:

<Html lang="en">
title="Opstrace RSS2 Feed"
title="Opstrace Atom Feed"
title="Opstrace JSON Feed"
<Main />
<NextScript />


Getting a first build on Cloudflare is incredibly easy, fast and convenient. Give Cloudflare permission for your repo, build, deploy—done. And it's free. If you use them for DNS, no manual DNS changes need to be made because Cloudflare does that for you. As they're feeding your site directly into their incredibly powerful content delivery network (CDN), you can expect the best load performance available.

With Cloudflare you also get some decent Account Analytics out of the box without any tracking snippets. (Also, script blockers cannot side-step this tracking.) They've also recently added Web Analytics. Overall it's a great offer for JAMstack sites and reduces the need for yet another tool to log in to and maintain.

Preparing Next.js was a little bit of work, but there were no serious blockers that prevented us from deploying with and hosting on Cloudflare Pages. We're now experimenting with Cloudflare Workers and Google Cloud Functions to add some server-side capabilities to our site, for example to collect feedback or handle our Stripe subscriptions. There are a few known convenience features missing from the build environment, but Cloudflare will probably add those soon. And maybe they'll even support Server-Side Rendering (SSR) capabilities for Next.js sites, too.

If you have any questions or comments, please reach out on Twitter or start a discussion on GitHub.

Acknowledgements & Special Thanks

Patrick Heneise

Chief Problem Solver and Founder at Zentered

I'm a Software Enginer focusing on web applications and cloud native solutions. In 1990 I started playing on x386 PCs and crafted my first “website” around 1996. Since then I've worked with dozens of programming languages, frameworks, databases and things long forgotten. I have over 20 years professional experience in software solutions and have helped many people accomplish their projects.

pattern/design file