Leave Webflow or Framer without losing a URL
By Adrian Kuleszo, Co-founder, DesignMe. Updated 11 November 2026.
Download the PDF (21 pages)Chapter 01
Our own move, in numbers
We ran designme.agency on Framer for three years. In October 2026 we moved it to Astro: first commit on 7 October, live on 13 October.
Six days.
- 91 to 99
- mobile PageSpeed Performance on Astro, from 33 to 71 on Framer
- $0
- a month to host on Cloudflare, from about $150 on Framer
- 15 min
- at most for a typical page edit with Claude, from 3 to 5 hours
- 56
- pull requests merged in the first week after the switch
And the four weeks after launch, from our own scoreboard: the last 28 days to 14 November, against the 28 days before.
- +469%
- non-brand Google impressions: 39,632, from 6,969
- +355%
- non-brand Google clicks: 50, from 11
- +165%
- leads: 53, from 20
- +64%
- booked calls: 18, from 11
- +18%
- sessions: 12,838, from 10,836
- 263
- pages live, from 63 on Framer

Our homepage, before and after Astro

Here’s what changed on the homepage, measured with PageSpeed Insights on 17 October 2026 (desktop, the median of 3 runs on each site). One dot is one unit: grey is Framer, white is Astro.
Same homepage, before and after
- 0.52 MB
- Page weight, from 10.33MB on Framer
- 75
- Requests, from 512 on Framer
- 37 KB
- JavaScript, from 5,590KB on Framer
- 0 ms
- Blocking time, from 1,361ms on Framer
Not everything improved, and you should know that before you move: the homepage’s mobile LCP stayed at 2.9 seconds on both sites. A new framework doesn’t fix the largest image on the page for you.
And the part I care about most: most of those changes started as one sentence to Claude.
What made six days possible
- The redirect map first116 rows before any design, so no page lost its address.
- The source savedRaw HTML, CMS exports and 1,378 files, before a line of code.
- A design system before pagesTokens and shared sections first, then pages assembled from them.
- Claude on every pageOne request per change, reviewed on a preview link.
Chapter 02
Should you move at all?
Not always. Webflow and Framer are great tools, and we still build on both for clients. This is the test we use.
Is the site in your way?
Weekly canvas edits, a handful of pages, Ecommerce or memberships
Dozens of pages, pages built from data, edits through pull requests
If none of the signs below apply, stay
Stay where you are if
- Your marketing team builds new page layouts on the canvas every week, and nobody on the team works in code or with coding agents.
- The site is a handful of pages and the plan bill is the only thing bothering you. The move costs more than years of the plan.
- You rely on Webflow Ecommerce, memberships or localization.
- You need visual editing with no developer involved, ever.
Move if
- The site has grown into dozens of pages and the design has drifted. (Ours had 50+ colors and 30+ font sizes, added one section at a time.)
- You want pages built from data: comparisons, integrations, locations, docs. Same template, different content, and that’s exactly what code is good at.
- You want your team (and Claude, Cursor or Codex) to change the site through reviewed pull requests, without waiting on whoever knows the editor.
- You want to own the site: a repository you control, no per-site plan, no editor seats, hosting wherever you like.
- Page weight matters to you, and the builder’s runtime is something you can’t remove.
If none of these apply, stay.
Chapter 03
Astro or Next.js
Both are good. They’re built for different jobs.
The two frameworks


Most pages are the same for every visitor. Pages ship as HTML, JavaScript only where needed.
- Marketing pages, blog and docs
- Pages built from data
- Markdown or a headless CMS
- Static files on any host
The site is part of your product, and the engineers who build it will own it day to day.
- Shared code or sign-in with an app
- Logged-in or gated areas
- Pages built per visitor or request
- A team that ships Next.js daily
Astro and Next.js for a marketing site
| Topic | Astro | Next.js |
|---|---|---|
| How pages build | HTML and CSS, JavaScript only in islands | React end to end, static and dynamic parts mixed |
| Publishing from a CMS | A rebuild that takes a minute or two | Changes can appear without a rebuild |
| Hosting | Static files on any host | Fits Vercel best, runs elsewhere |
| Server code to patch | Very little on a static site | More on a server-rendered site |
Running both
Some teams run both: Astro for the marketing site, Next.js for the product. Everything else in this guide works for either.
The marketing site sits on your main domain and ships as static pages, so marketing can change it without touching the product. The product sits on an app subdomain, built and shipped by your engineers. Both read one shared set of design tokens, so a visitor going from the pricing page into the app sees one company.
- The marketing site, on AstroYour main domain: marketing pages, blog and docs, changed without touching the app.
- The product, on Next.jsAn app subdomain: sign-in, accounts and dashboards, shipped by your engineers.
- One set of design tokensColours, type and components in one package both use, so it feels like one company.
Chapter 04
What moves, and what breaks
Neither platform gives you a site you can build on. Framer has no HTML export at all. Webflow’s code export gives you HTML and CSS, but CMS lists come out empty, forms lose their backend and only the primary locale is included.
So the move is a rebuild, not a conversion. The content comes with you. The layout doesn’t.
Every part of the site has a home in code
- CMS collectionsMarkdown or typed data files, checked by a schema
- Pages and sectionsComponents on a design system with tokens
- Images and files on the builder’s CDNDownloaded and served from your own host
- FormsA form service, your CRM or a serverless function
- Interactions and effectsCSS or a small motion library, only where used
- Redirects in site settingsA redirects file, one 301 per row
- SEO fieldsFrontmatter, plus a generated sitemap
- The layoutRebuilt. The only part that doesn’t come across
What comes across from Webflow and Framer
| Part of the site | From Webflow | From Framer | On Astro or Next.js |
|---|---|---|---|
| CMS content | Each collection exports to CSV (or comes out through the Data API), rich text as HTML | Each collection exports to CSV with Framer’s CMS Export plugin | Markdown or typed data files, checked against a schema at build time |
| Page copy | Saved from the live pages | Saved from the live pages | Copied word for word from the saved HTML, nothing retyped from memory |
| Images and files | On Webflow’s hosting, linked from the CSV | On framerusercontent.com | Downloaded and served from your own host |
| Pages and components | Rebuilt | Rebuilt | Rebuilt on a design system with tokens |
| Interactions and effects | Rebuilt | Rebuilt | CSS or a small motion library, shipped only where used |
| Forms | Lose their backend | Need a new home | A form service, your CRM or a serverless function |
| Redirects | Exported as CSV | Copied out of site settings | Real 301s in the host’s redirect file |
What breaks without warning
- Images on the old CDN. They still load on launch day, then stop working when you cancel the plan. Search the built site for Webflow and framerusercontent.com URLs before launch. We moved 1,378 files off Framer’s CDN.
- Pages missing from the sitemap. Our own Framer sitemap was missing a page the navigation linked to. Crawl the site as well as reading the sitemap.
- Trailing slashes. Framer serves URLs without one. The new config has to match, or every URL changes.
- Redirects that aren’t 301s. On a static Astro build, redirects in the Astro config become meta refresh pages, and Cloudflare Pages uses 302 unless you write 301.
- Editing. The marketer who published on a Friday afternoon can’t any more, unless you decided who edits what before the build (chapter 6).
Chapter 05
The redirect map
The redirect map is the part most migrations get wrong, so start there, before anyone designs anything.
Our own site had 63 live pages. The map had 116 rows, because every source of old URLs turns up pages the others miss:
Where the rows come from
- Live site crawl
- XML sitemap
- Search Console
- Analytics
- Links you control
- Wayback Machine
- 116 rows, from 63 live URLsNormalise to paths, remove duplicates
Then every row gets one of three outcomes:
One decision per old URL
- Same URLKeep it. 65 of ours. No redirect
- Changed URLRedirect it. 301 to the new URL. 43 of ours
- RetiredRedirect it. 301 to the closest page. 8 of ours
Never mass-redirect to the homepage
How to build it
- Collect every old URLFrom a crawl, the sitemap, Search Console, analytics, campaign links and the Wayback Machine.
- Keep paths wherever you canAstro and Next.js let you choose every route, so there’s rarely a reason to change one.
- One hop to the closest pageNo chains, and never dump retired pages on the homepage.
- Write the host’s formatAnd check its default status code: Cloudflare Pages uses 302 unless you write 301.
- Test, then watchEvery row before launch, 404s daily after it. Keep the redirects for at least a year.
The first week of 404 reports turned up a few more old links, and our map is now 124 rows. Every page that earned links or rankings still has a working address.
What the file looks like
On Cloudflare Pages, where designme.agency runs, the map is a plain text file called _redirects in the build: old path, new path, status code, one row per line. Four rows from ours, the last one a retired page sent to the closest page:
public/_redirectsFour rows from ours
| Old path | goes to | New path | Status |
|---|---|---|---|
/call | /intro | 301 | |
/about | /about-us | 301 | |
/privacy-policy | /legal/privacy-policy | 301 | |
/projects/recruitu | /projects | 301 |
Write the 301. Leave it off and Cloudflare Pages sends a 302, which tells Google the move is temporary.
On Next.js the same rows go in next.config.js. permanent: true sends a 308, which Google treats like a 301:
next.config.js
module.exports = {
async redirects() {
return [
{ source: '/call', destination: '/intro', permanent: true },
{ source: '/about', destination: '/about-us', permanent: true },
];
},
};Keep the map itself in a spreadsheet (old URL, new URL, outcome) and generate the file from it. Ours lives in the repository as a CSV, and a script writes _redirects from it, so nobody edits the file by hand.
Test it before DNS moves
Check every row on staging before the switch. Each old URL should answer with one 301 and land on a page that loads: never a chain, never the homepage. A crawler in list mode, such as Screaming Frog, runs through a few hundred rows in minutes.
- One hopThe old URL goes straight to its final page.
- A 301, or a 308 on Next.jsPermanent, so Google moves the rankings across.
- A page that loadsThe new URL answers with a 200, not another redirect or a 404.
- Never the homepageRetired pages go to the closest real page.
After launch, keep going: new 404s from Search Console and the host’s logs go into the map the same day. Ours has grown from 116 rows at launch to 154.
An example check on staging
| Old URL | Status | Lands on | Result |
|---|---|---|---|
| /call | 301 | /intro, which loads | Yes |
| /about | 301 | /about-us, which loads | Yes |
| /privacy-policy | 301 | /legal/privacy-policy, which loads | Yes |
| /projects/recruitu | 302 | /projects, which loads | No |
The last row fails: a 302 tells Google the move is temporary. Change it to a 301 and run the check again.
Chapter 06
Choosing the CMS
Decide who edits what before the build, not after. It changes the build.
Editing the files
- Markdown and data filesFor developers and AI agents, through reviewed pull requests. A typo is a request.
- KeystaticA form on top of the same files, for titles, descriptions and simple fields.
For marketers who publish every week
Give them a headless CMS. These are the two we’d look at first. Both work with Astro and Next.js.
Structured content in Sanity Studio on your own domain. The site reads it through an API.
- Your own content types
- Drafts with a live preview
- Real-time editing for the team
- A free tier to start
A visual editor: marketers edit on a live preview and drag in approved components.
- Click-to-edit on the page
- Components from your system
- Translations and workflows
- A free tier to start
Here’s what we run: insights are Markdown files in the repository and the rest is typed data. Everything changes by asking Claude, and nothing goes live until a person approves the preview.
If your marketers want to publish a post on their own on a Friday afternoon, give them a CMS. (On a static Astro site, a publish triggers a rebuild that takes a minute or two.)
Chapter 07
The build, in order
This is the order we used on designme.agency, and the order we use for clients.
From inventory to staging
- Inventory every URLGive each one a route in the new project before a line of code.
- Save the sourceThe raw HTML of every page, every CMS collection as CSV, every asset mapped from its old URL to a local file.
- Design system firstThe builder’s values become tokens (type, color, spacing, light and dark), then shared sections. Pages are assembled from those.
- Parity, then improvementsThe first version matches the live site page for page: same copy, same SEO. Changes come after sign-off.
- SEO on every pageTitles, descriptions, canonicals, OG images and structured data, plus a sitemap with only indexable pages.
- Forms and trackingForms connected to your CRM, analytics and cookie consent set up and tested.
- Redirects in placeThe full map in the host’s format, every row tested on staging.
- Staging, side by sideThe full site on a private address, compared with the old one page by page.
Parity first is the rule people want to skip. Don’t. If traffic moves after launch, you’ll know it wasn’t the copy.
Chapter 08
Launch week, and after
Launch early in the week, so someone is around when Google starts re-crawling.
Around the switch
Before DNS moves
- Redirects testedEvery row of the map, on staging.
- No staging noindexChecked on every template.
- No old CDN linksIn images, videos, OG tags or CSS.
- Sitemap and robots.txtOnly indexable pages, and the staging block gone.
Launch day
- A way backOurs was one DNS change, with a five-minute path back to Framer.
- Sitemap submittedThe new one, in Search Console.
- Forms and analyticsTested on the live domain, with cookie consent set up.
The first weeks
- 404s and indexingWatched daily until the numbers settle.
- The old plan keptUntil every asset is off it.
- New 404s mappedEach one added to the redirect map the same day.
The launch checklist
Every step in this guide as one list. Print it, or copy it into your project tool.
Chapter 09
Timeline and price
Our own move took six days. That was our site, our team, no sign-off rounds and Claude doing a lot of the typing. For a client site, we plan 4 to 5 weeks for most sites, from the URL inventory to launch. Heavy integrations, lots of custom effects or a big CMS take longer.
Webflow or Framer to Astro, week by week
Coloured bars are our work, grey ones are yours.
- You: Access (Week 1 to Week 1), Preview reviews (Week 4 to Week 4), Kickoff, Model OK, Launch
- Scope and stack: Scope and stack (Week 1 to Week 1)
- Content model: Content model (Week 1 to Week 2)
- Design and build: Design system in code (Week 2 to Week 3), Pages and CMS (Week 3 to Week 4)
- Launch: SEO parity and redirects (Week 4 to Week 5), Launch and handover (Week 4 to Week 5)
- 4 to 5 weeks
- for most sites, from the URL inventory to launch
- $10,000
- the starting fixed price for a landing page sized site
- $15,000
- the starting fixed price for a full site or a headless CMS build
What the fixed price covers
- Design systemTokens, type, colour and spacing, and the shared components every page is built from.
- The rebuildEvery template and page, rebuilt on Astro or Next.js, matching the live site first.
- ContentEvery CMS collection exported, cleaned, moved into code or a CMS, and checked.
- RedirectsA 301 for every old URL that moves, in the host’s format, tested before launch.
- SEO and QATitles, descriptions, schema and the sitemap carried over. Every form and link tested.
- One price, agreed firstYou get the scope and the price on a call, before anything starts.
You get the scope and the price on a call, before anything starts, and the platform you pick doesn’t change it.
Running costs usually drop. A mostly static Astro site costs little or nothing to host for most marketing sites, and there are no editor seats for the site itself. A CMS like Sanity has its own plans, so price it for your team.
Want a number for your site? Book a 30-minute call and bring your URL. You’ll leave knowing which stack fits, roughly how long it takes and what it costs. And if the answer is “stay on Webflow” (or Framer), I’ll tell you that too.




