Astro 6: What Actually Changed

I spent a weekend migrating a project from Astro 5 to 6. The good news is that most of the work was updating Node.js. The bad news is that the migration wasn’t as smooth as the docs suggest.

Node.js 22 is now required. That’s it. That’s the headline. If you’re on Node 20 or older, you need to upgrade first. I use nvm, so it was a couple of commands. If you’re on a shared hosting environment or a CI pipeline with pinned versions, you’ll need to update those too.

The Node.js requirement isn’t arbitrary. Astro 6 uses ES2026 features under the hood. Things like the Temporal API for dates and Iterator Helpers. These aren’t just nice-to-haves anymore. They’re part of the framework’s core. Which means if your project depends on older Node versions for some reason, Astro 6 is not going to work for you. Period.

Build performance improved noticeably. My project went from about 45 seconds to 32 seconds. That’s not a typo. The pages are the same, the integrations are the same. It just gets done faster. Dev server startup also feels snappier, though that’s harder to quantify.

Content Collections got a transform option. This is actually useful. You can now modify entries before they reach the renderer. I used it to calculate reading time for each post. Before Astro 6, I was doing that in the template layer, which felt wrong. Now it happens during the build, and the value is available everywhere.

Server Islands arrived. This is Astro’s answer to the hydration problem. Instead of hydrating entire component trees, you mark individual components as islands. They render server-side and hydrate on the client independently. It’s a subtle change but it matters for performance on pages with lots of interactive elements.

View Transitions are now built in. You just add the component to your head and they work. No extra configuration. The integration feels complete rather than experimental.

The breaking changes are manageable but real. fetchContent is gone. You use getCollection now. The config format changed slightly. SSR options moved from vite to the top level. Nothing catastrophic, but you will hit these if you’re upgrading from 5.

Image optimization got simpler. The dedicated Image component is no longer necessary for most cases. Astro handles optimization automatically now. If you were using the Image component extensively, you’ll want to audit your templates. Some things might stop working if you just remove imports without checking what the component was doing.

All the official integrations got version bumps. React 6, Vue 6, Svelte 6, Tailwind 6, MDX 6. If you’re using any of these, expect to update them too. The integration versions are tied to Astro’s version, so they move together.

Astro 5 will be in maintenance until December 2026. Astro 4 gets security fixes only. If you’re still on Astro 3 or older, you’re behind. A lot behind.

The migration path is straightforward. Update Node.js. Run npm install astro@6. Update your integrations. Run astro check. Fix the things it finds. It’s not zero effort, but it’s not a rewrite either.

I’d say the biggest win is build performance. The 29% improvement is real and noticeable on larger projects. The Content Collections transform is the feature I use most. Everything else is incremental improvements.

Astro keeps improving quietly. Version 6 doesn’t revolutionize anything, but it makes the things you do every day a bit faster and a bit easier.