AI coding assistants made a hand-built site cheaper to make, and the same question keeps coming up: is a no-code builder still worth it for a marketing site, or should you move to a framework like Astro? Each side argues about speed, cost and freedom. We checked those arguments against public numbers.
We build Framer sites for a living, so weigh what follows with that in mind. We tried to write the comparison we would want to read before choosing, with public numbers where they exist and a plain “we don’t know” where they don’t.
The short answer
Pick by who changes the site, not by speed. At the median, Framer and Astro sites pass Core Web Vitals on mobile at the same rate, 70.6% against 71.0%. Astro ships half the JavaScript and gives you every line of HTML. Framer lets you hand over a site your client’s team can change without you, with hosting, CMS, translation and staging on one bill. If you will touch the site every week anyway, Astro is hard to beat. If nobody on your client’s team writes code, the repo becomes the expensive part.
Speed is closer than either side says
HTTP Archive measures real Chrome visits to millions of sites every month and groups the sites by the technology they run on. A site passes Core Web Vitals when 75% of visits get a fast largest paint (LCP), a stable layout (CLS) and a quick response to input (INP). This is the share of sites that passed in August 2026, the latest month with data.
August was the first month in a year where Astro sites passed more often than Framer sites, by 0.4 points. On desktop Framer is ahead, 87.1% to 83.6%. The details split the way you would expect. Framer sites do better on loading: 82.5% pass LCP against 80.6%. Astro sites do better on interaction, 85.7% pass INP against 80.2%. Framer ships a React runtime that has to start before the page responds. Astro sends HTML and adds JavaScript only to the parts you mark as interactive.
Two caveats cut both ways. These numbers show how other people build with each tool, not what you could build with it. Astro sites include docs and blogs, while Framer sites are mostly marketing pages and portfolios. And among the million busiest sites, Framer drops to 39.7% against Astro’s 54.9%. That figure rests on only 358 Framer sites, so read it as a hint.
Weight is where Astro wins without argument. The median Framer home page loads 788 KB of JavaScript on a phone. The median Astro home page loads 396 KB.
If you build carefully, you can ship a page that loads in a fraction of a second with either tool. Nothing in the public data shows a large gap at the median. The claim that a coded site is faster to update is a different argument. It is about workflow, and it depends on who does the updating. With an AI assistant, you can change a component in a repo faster than you can click through a canvas. Your client’s marketer can’t change it at all.
What it costs to run
Say your client wants a 50-page marketing site with a blog, three people who edit it and two languages. On Framer that needs the Pro plan ($360 a year, billed yearly), one extra locale ($240) and two content editor seats ($240). That is $840 a year, or $1,080 if both editors need full design access. For more languages, see what a multilingual Framer site costs. Astro’s running cost depends on what you put around it.
Astro itself is free, and free hosting exists. Cloudflare serves static files “free and unlimited”, and Netlify’s free plan works until its 300 monthly credits run out. On a static site, each publish from the CMS usually triggers a deploy at 15 credits, and once the credits are gone every site on your account shows “Site not available” until the next month. We would not put a client’s site on that, so every Astro row in the chart starts on Netlify Pro at $240 a year. Vercel Pro costs the same, and Vercel’s free plan is for “personal, non-commercial use” anyway.
The CMS decides the rest. Sanity’s free plan gives 20 seats but only the Administrator and Viewer roles, so every one of your client’s editors gets admin rights. Its Growth plan adds an Editor role at $15 per seat a month. Prismic’s Starter plan fits three users and three locales exactly, at $10 a month billed yearly. Contentful’s free plan has room for ten users, but its usage terms say it “may not be used to support commercial use cases”, so a client site starts on Lite at $300 a month. The setup closest to Framer’s on-page editing, Storyblok’s visual editor, costs $1,089 a year on Growth.
None of these bills include people. An Astro site needs a developer for upgrades and new page layouts, and Astro shipped two major versions in 2026 and 116 releases in twelve months. A Framer site needs someone who knows Framer for new layouts too, though content editors can change copy and translations on their own. Put your own hours on both sides before you compare totals.
What comes in the box
| Framer | Astro | |
|---|---|---|
| Hosting and CDN | Included | Your host: Cloudflare, Netlify, Vercel |
| CMS | Included, 2,500 items on Pro | Add one: Keystatic, Sanity, Storyblok |
| Editing on the page | Yes, with content editor seats | Only with a visual CMS such as Storyblok |
| Languages | $20 per locale a month, hreflang set for you | Routing built in, hreflang by hand |
| Staging, branches, rollback | On Pro | Git, plus your host's previews |
| A/B tests and funnels | Paid add-on, $50 per 500k events | A separate tool |
| Redirects | On Pro | In your host's config |
| HTML | Generated, plus head and body snippets | Yours, every line |
| Leaving | No HTML export | Your repo, MIT licence |
| AI agents | Canvas agents on credits, outside agents through Framer's MCP | Any coding assistant on the repo |
- Framer
- Included
- Astro
- Your host: Cloudflare, Netlify, Vercel
- Framer
- Included, 2,500 items on Pro
- Astro
- Add one: Keystatic, Sanity, Storyblok
- Framer
- Yes, with content editor seats
- Astro
- Only with a visual CMS such as Storyblok
- Framer
- $20 per locale a month, hreflang set for you
- Astro
- Routing built in, hreflang by hand
- Framer
- On Pro
- Astro
- Git, plus your host's previews
- Framer
- Paid add-on, $50 per 500k events
- Astro
- A separate tool
- Framer
- On Pro
- Astro
- In your host's config
- Framer
- Generated, plus head and body snippets
- Astro
- Yours, every line
- Framer
- No HTML export
- Astro
- Your repo, MIT licence
- Framer
- Canvas agents on credits, outside agents through Framer's MCP
- Astro
- Any coding assistant on the repo
The “leaving” row deserves more than a cell. Framer’s help center says it plainly: “Framer does not offer HTML export for self-hosting.” You can take your CMS content out as CSV or JSON and keep your code components. Layouts, interactions and the editing setup stay in Framer. Third-party exporters exist, Framer does not support them, and their output varies. If being able to leave matters to you, this is the strongest argument for Astro.
Where each side overstates
The case for Framer gets a lot right. A marketing team can design and ship in one place, content editors at $10 a month can change copy and translations without a ticket, and branching, staging and rollback come with Pro. Two claims you will hear go too far. A/B tests and funnels are not built in. They come with the Convert add-on at $50 per 500,000 events, on Pro only. And “1,000+ pages” holds for CMS pages or an Enterprise plan. Pro stops at 700 designed pages and 40,000 CMS items, add-ons included.
The case for Astro is right about weight and control. Astro ships the least JavaScript of any platform HTTP Archive tracks, and code works with git, review and every AI coding tool. “No-code has had its day” leaves out the people who edit the site. It also leaves out the vendor question. Cloudflare bought the company behind Astro in January 2026. The MIT licence means anyone can fork it, and your host and your CMS are vendors too.
Pick the tool for the people who will edit the site, not for the person who builds it.
Weigh it yourself
We scored both tools from 1 to 5 on the criteria that come up most in this debate, using the numbers above. The weights are ours, set for a client whose marketing team edits its own site. Drag them to match yours.
At our weights Framer finishes three points ahead, because we weighted editing without a developer and developer time for a team that runs its own site. Raise control of the HTML, page weight and freedom to leave, and Astro pulls ahead.
What we would pick
Our pick depends on who runs the site and how often it changes.
Designers and writers run the site, it changes every week, it speaks more than one language: Framer.
A developer is already on the team, the site is mostly content, accessibility or page-weight targets are strict: Astro.
A site that works today on either one: stay. The median says a migration will not buy you speed.
So our advice comes down to what happens after launch. If you deliver the site and hand it over, Framer is the better fit: your client’s team edits it on the page, Framer hosts it, and nobody needs you to change a headline. If maintenance, updates and hosting are part of what you sell, Astro fits that model better: the repo, the host and the CMS stay in your hands, and every change comes back to you.
How we run this site
We should say how this site runs, because it shapes everything above. framerlabs.com is built in Framer, and we have barely opened the canvas. An AI coding agent connected through Framer’s MCP does the work. It builds the pages, writes these articles into the CMS, wrote the charts in this one as code components, and publishes a preview we check before anything goes live.
That changes the question this article started with. The best argument for Astro was that code works with every AI coding tool. The best argument for Framer was that a team can change the site without a developer. An external agent gives you both. It edits pages, components and the CMS the way it would edit a repo, and the site keeps Framer’s hosting, CMS, translations and on-page editing for everyone else.
It does not fix the “leaving” row, since Framer still has no HTML export. But if you already run a site on Framer, connect an agent before you plan a migration. You may find you no longer need one.
If you do want to move a site to Framer, we run migrations with this same setup. Tell us about your site, or see how it went on two Webflow sites.