Svelteload
The stack behind every Nodebrush website. SvelteKit and Payload CMS, self-hosted Postgres and media, multilingual editing, and an AI connector that drafts changes for a human to publish.
What Svelteload is
SvelteKit renders the public site and Payload CMS handles the editing, with the database and every uploaded file on our own servers in Europe. Visitors get a fast site, editors get a familiar admin panel, and there are no per-document fees or vendor lock-in.
Editing day to day
An editor's week is mostly the same few moves. Build a page, get someone to sign off on it, write it in another language, and find it again six months later.
Block-based page editing
Pages are built from reusable section blocks such as heroes, banners, image grids, and contact forms. Editors drag, swap, and rearrange them without writing code or asking a developer.
Preview and publish from the page
Editors copy a preview link for any draft and send it to anyone, no CMS account needed on the other end. The links expire by themselves. Sign in on that preview site and the publish button sits right on the page you are reading, so the final check and the go-live happen in the same place.
Roles and access
Five roles, each with a clear ceiling. Readers look but change nothing. Contributors write drafts and cannot publish. Editors own the content visitors see, pages, posts and media, but not the settings behind it. Agents are for AI assistants and are blocked from publishing and deleting outright. Admins have the whole CMS, including configuration and inviting new people by email.
Multilingual content
Every text field is translated language by language without splitting the site into separate copies. You pick the languages when the site is built, English and Swedish or any others your customers read. Editors switch between them from one dropdown.
Nothing left half-translated
A review screen lists every document alongside the languages it exists in, so a page missing one of its translations stops being invisible. Open one and it shows word by word what changed since the last time someone signed it off.
Built-in search
Pages and blog posts are searchable on the public site, ranked by relevance. No external search service to host or pay for.
Self-hosted media
Uploads sit on our own object storage and are served from your own subdomain. Each file records where it is used across the site, so nobody deletes a picture that is still on a page, and stock photos carry their source link and licence document with them. No per-gigabyte egress fees and no third-party storage contract.
URLs that survive a rename
Rename a published page and the old address keeps working. The redirect is staged when the draft diverges and goes live the moment the page is published, so search results and old links do not break in the window between the two.
Fast with AI, maintainable without it
Content work has plenty of tedium worth handing off, and none of it should cost you a codebase you can no longer read. We gave the CMS its own Model Context Protocol server, an open standard, so it works with whatever assistant you already use rather than tying you to one product. Underneath, the code stays ordinary, with clean components, typed fields and normal version control. Drop the assistant tomorrow and the site is still readable, editable and extendable by hand. Everything is backed up and portable, so the whole project can move to your own hosting whenever you decide to take it in-house.
