Ask most CMS vendors what happens if their service goes down, and the honest answer is: your content stops loading. Carapace can't have that problem, because there's no service to go down.

## Content as files, not rows

A Carapace post is an HTML file sitting next to a small JSON sidecar of metadata — title, date, tags, draft status. That's the entire schema. There's no database migration to run when you add a field, because "adding a field" just means writing a new key.

```json
{
  "title": "No Database, No Headless CMS",
  "date": "2026-07-13",
  "tags": ["cms", "architecture", "git"]
}
```

## Drafts are branches

A draft post is a branch that hasn't merged to `main` yet. Reviewing a post before it ships looks exactly like reviewing a pull request, because it is one.

> Publishing is a merge. Unpublishing is a revert. Your entire editorial history is `git log`.

No export tool required to leave, either — the content was always just files.
