Astro comes up more and more often as an alternative to Webflow. As a serious Webflow fan, I have to admit there are now projects where the platform has stopped being the best option, and the Astro framework is one of the reasons why. It is also the setup for teams that want to update their website with AI instead of learning Webflow.
I have worked with Webflow for years, with teams of every size and technical level, and it is a genuinely great platform. What I keep seeing, though, is that nobody on those teams uses the Designer themselves. There is a learning curve, and people would rather not spend their time on a tool, which I completely understand.
Webflow's answer to this is components, and to a point it helps. In practice, though, no component system, however elaborate, ever removes the need to come back to a Webflow developer on a large project. Components cover the simple cases, and every new page tends to differ from the last one.
Meanwhile the same people have no trouble at all putting a page together through Claude or ChatGPT. The process of working on pages in that case is exactly what it should be, and it is hard to imagine anything better.
It does, and with a certain amount of persistence you can get some results out of it. It is a long way from the experience people get working with Claude or ChatGPT directly, and the MCP is essentially a layer built on top of the existing interface. Anyone claiming it delivers a revolutionary gain in speed, or changes the process for the better, has most likely not tried to build anything real with it.
Any website ends up being a set of HTML files on a server, and that includes Webflow. The Astro framework is simply one way of working with those files. Used properly, it lets you reuse what has already been built, keep your content structured, and hold the whole project in good order.
In general you will forget it is there at all, because it never gets in your way. You genuinely stop thinking about it.
| Webflow | Astro | |
|---|---|---|
| Who changes pages | Someone who knows the Designer | Anyone on the team, by asking Claude or ChatGPT |
| Who edits posts and collections | Editors in the Webflow CMS | Editors in a CMS like Sanity, or anyone by asking Claude or ChatGPT |
| Hosting | Webflow, as part of the site plan | Cloudflare or any host you choose |
| Who holds the files | Webflow. The code export leaves CMS content out | You, in your own GitHub account |
Not moved: online stores and member areas.
I list every page, URL, form and redirect on your current site.
Each page is rebuilt one to one in clean HTML, with the same URLs, copy and SEO settings, so the switch is as smooth as it can be.
Pages go to GitHub, which keeps every version. A blog or collections go to a CMS like Sanity, which your team edits directly.
The site goes live on Cloudflare, a hosting service, or on your own server if you need that.
You and your colleagues ask Claude or ChatGPT for most changes, in their app or right from Slack. Your content team can also edit posts and collections directly in the CMS. Your site's rules check every change, and it's live in about a minute.
Webflow handles hosting for you: the HTML gets built from what you made in the Designer and served from Webflow's infrastructure. With Astro, that part is yours to arrange.
A good setup right now is to create the project through the AI assistant of your choice, keep the source in GitHub, and host it wherever you like. Cloudflare in particular gives you an excellent experience for pushing updates and working on the site day to day.
Because the files are no longer locked inside a proprietary platform, you can update the website directly from Claude or ChatGPT and publish from there. You do still need to keep an eye on whether what comes out fits the rest of the site, and the skills set up inside the project catch a good part of that.
glevsky.com is an Astro project in GitHub, hosted on Cloudflare. I change it by asking Claude Code, and a change is live in about a minute.
No. I still believe in not touching what works, and in making changes only when the need for them is clear.
If you are planning a site of up to thirty or fifty pages with a blog updated every couple of weeks, then, however much it pains me to say it, you do not need Webflow. Astro is a full alternative for that.
With dozens of collections and hundreds of pages, a migration needs to be approached very carefully. The process is not quick, it demands a lot of attention, and things can get lost along the way. Moving for the sake of moving is unlikely to justify itself here.
Look instead at what specifically is not working for you in Webflow and deal with that directly, inside your current setup, rather than replacing the whole platform.
Not quite. Whether you are migrating an existing site or starting a new one from a Figma design, it takes a considerable amount of time up front, building the first pages and establishing the design system that every page after them is built from.
For all the intelligence of current models, they are still some way from working independently, without a human in the loop. More importantly, the architectural decisions at the start of a project are not theirs to make, and those decisions are what determine how comfortable and how fast the project is to work on later.
Getting that part right is exactly where I come in. Once the foundation is sound, building on the site is a breath of fresh air and new pages really do come together quickly.
About 1 to 2 weeks for a small site. The old site stays up the whole time.
I cannot guarantee that, and nor can anyone who tells you they can. The move leaves Google nothing new to judge: the same URLs, the same titles, the same copy, served faster. That is as close to zero impact as it gets.
Pages go to GitHub, which keeps every version. A blog or collections go to a CMS like Sanity, which your team edits directly.
Nothing. Your email and anything else on your domain stay untouched by the move.
Ask for the previous version back. Every change is saved, so any earlier version can come back.
Yes. On the call I open my demo project and make a change on it live, so you see the whole thing before you decide.
Not yet. I haven't built stores or sites with logins this way yet, so I don't take them on.
The same as any other move: $1,999 for a small site, from $5,000 for a normal marketing site.
$1,999
for a site up to 5 pages
Book a 30-minute call→A bigger site with a blog starts at $5,000.
Included
After the move
Claude or ChatGPT, about $20 a month. Hosting is usually free. After handover you owe me nothing. If you want support, I stay on for fixes and bigger changes.
