TL;DR: Cancelled entire Webflow stack, now running on Cloudflare + Sanity for $0/month. Cleaner animations, top PageSpeed scores, faster builds. Every layer between AI and code makes Claude less capable. Clients stopped paying Webflow and stopped paying retainers. Webflow still wins for big team handoffs and first-time site owners.
I loved Webflow.
It was the best platform. It gave the most freedom for impressive marketing websites with awesome animations. I had custom components, a Relume workflow, client sites, the whole thing.
Here is what I was paying:
- Webflow CMS plan: ~$29/month
- Localize add-on: ~$12/month
- Analyze: ~$12/month
- Freelancer plan: ~$24/month
Total: roughly $80/month just to keep my own site running. Every client paid some combination of the same stack.
The cost was not the problem. Webflow earned its keep. But I saw an opportunity to integrate AI into the workflow to multiply quality and speed.
Why I tried to leave
I had been seeing what Claude Code is capable of and wanted that energy inside Webflow. From research to full builds to ongoing fixes.
I tried the Webflow MCP. The connection between Claude Code and Webflow.
It was rough. Claude had a hard time interacting with the MCP.
The freedom I expected was not there. Claude could not build what I wanted. It kept getting stuck, consuming time and tokens excessively.
The MCP is improving. But I realized the things I actually wanted AI to handle were:
- Build pages from scratch
- Install analytics stacks
- Connect domains
- Manage backups
- Change SEO/AEO settings across entire sites
- Localize entire sites
- Monitor for broken links and performance issues
- Update CMS items on schedule
- Push PageSpeed scores above 95%
All of those would be much easier if Claude could talk directly to code.
So I looked for a different stack.
What I pay now
Cloudflare + Sanity. Zero.
- Free hosting on Cloudflare
- Free CMS through Sanity
- Free analytics through Cloudflare, Microsoft Clarity, and GA
- Two languages, no add-ons
I migrated my own site first as an experiment. Once the workflow was established and saved as Claude Code skills, the next migrations became easier.
Smaller clients moved next.
When their sites moved, they dropped their retainers. The clients did not need ongoing content updates anymore. Sanity handles that.
They also stopped paying Webflow. They got cheaper sites and I got my time back.
What I did not predict
The earlier wishlist items? Claude accomplishes all of them. Capability was never in question. That is why I left Webflow.
What I did not predict was how the work would feel.
Before, every animation idea came with trade-offs. Will it slow the site? Will it conflict with existing elements? Will Webflow break it in updates? Half my energy went to those questions before writing any code.
Now, when something does not flow, I just fix it.
The only question is whether the thing should exist.
Every layer between AI and your code makes Claude dumber
Easier tools are now the slow tools.
Webflow MCP, Framer plugins, WordPress page builders. Each abstraction consumes tokens and reduces Claude’s freedom. Strip the layers and Claude moves at full speed.
It’s the best decision to make and people still won’t make it
People will skip this jump even when it is cheaper and better long-term.
Why? Because of the learning curve. Because of the time it takes at the start to actually solve problems.
It requires working with the CLI and Claude Code, which many people find intimidating. You need to:
- Learn to prompt effectively
- Learn to give enough context
- Work with Claude’s token costs
- Solve problems and learn a different work approach
It takes time for people to change their beliefs. Maybe their entire system is built on Framer. Maybe all their clients use Webflow. That is why WordPress still dominates the website builder market, despite not being built for current AI tooling.
To be clear: I am not vibe coding. I understand everything shipping. I decide the framework. I decide between Astro or Next. I walk through code with Claude to ensure I understand each part and how they integrate. Claude executes. I do not care about syntax.
The compounding employee
Every site I build teaches Claude something new about my work. I have a starter repo on GitHub. I have skills capturing my workflow.
Each site updates the skill with gotchas, patterns, and things that almost broke. The next site starts smarter.
It is like an intern who gets better every week. Webflow, Framer, WordPress. None of them give you that.
With Claude Code, you compound.
Where Webflow still wins
Webflow is not dead. Far from it.
Big teams with marketers touching the site. The visual editor is friendlier than Sanity for someone who has never used a CMS. The handoff to a client who wants to make changes is smoother.
First-time site owners. If you have never built a website before, the Webflow learning curve is much lower than learning Claude Code in a terminal. That is a fair objection. For second or third sites, the math flips. For the first one, Webflow probably still wins.
But the gap is closing. Sanity already lets clients edit copy and swap images independently. The tooling around AI plus content is moving fast. The “Webflow is friendlier for clients” argument gets weaker every quarter.
Open your billing page
Look at what you are paying for your no-code stack. Webflow, Framer, WordPress, Squarespace, whatever it is. Add it up. Add what your clients are paying on top.
Then ask: is this still earning its keep? Or is it muscle memory from a different era?
If you cannot justify the bill, you have your answer.