Decap CMS

What Decap CMS actually is

Decap CMS (github.com/decaporg/decap-cms) solves one narrow problem well: your site is a folder of Markdown files in a repository, and the person who wants to change a sentence does not use Git. It installs under the admin path of the site you already have. The quick version, in the project's own description, is a single HTML file plus a configuration file, with the CMS itself loaded from a CDN; the fuller install goes through npm and gives you more room to customise. A YAML config declares your content model, the editor builds forms from it, and each save becomes a commit.

The editing surface is assembled from widgets, and the set is wide enough for most sites: Markdown and rich text, images and files, dates, numbers, booleans, selects, lists and nested objects, plus a relation widget for pointing one entry at another. Sign-in is pluggable rather than assumed — the repository ships backends for GitHub, GitLab, Bitbucket, Azure, Gitea and Forgejo, a hosted Git gateway, and a proxy for local editing. Media can live in the repository or go out to Cloudinary or Uploadcare. The licence is MIT, and a paid tier called Decap Turbo exists for teams that want centralised user management, finer roles, a database proxy or supported onboarding.

One naming note that will save you an hour of confusion: this was Netlify CMS until February 2023, and the project says so in its own README. A large share of the tutorials you will find still use the old name and the old package, and the configuration has moved on since they were written.

The trade you are making

The commit-per-save model is the whole design, and it cuts both ways. Your content is versioned by default and can be reviewed as a pull request if you want it to be, so a copy change can go through the same process as a code change — genuinely unusual, and valuable on a team where nobody is sure who edited the homepage. The cost is latency. A typo fix is a build, so if the build takes four minutes then correcting a typo takes four minutes. For a site that publishes weekly this is irrelevant. For one publishing hourly it is the wrong tool, and something with a database behind it is the right one.

It helps to place it against its neighbours rather than judging it alone. Payload sits at the other end of the same spectrum — a full framework with its own database, which is what you want when the content is really an application. Ghost is a finished publishing product rather than a layer over your repository. Decap is the thinnest of the three, and thin is the point: it adds an editing screen to a site that already works, and it takes nothing else over.

Who edits the page after it ships

We do not run Decap ourselves, so this page claims nothing about operating it. What we can speak to first-hand is the question it exists to answer, because we came at it from the other side. We publish landing pages for our own customers, and the editing interface is a conversation: the customer asks for a change and the agent makes it. One customer took a page through three rounds of revisions in a single sitting, and the striking part was where the time went. None of it went into the editor. All of it went into deciding what the page should say.

That is the argument for keeping this layer boring. Whichever option you choose, the bottleneck moves to the copy within about a week, so pick the one that costs you least to maintain and spend the difference on the words. The reason we would still put a Git-based editor ahead of something heavier for a small site is a second lesson, learned the expensive way: what a static site generator emits is finished HTML, which quietly avoids the failure that cost us months of indexing on our own dynamically-rendered pages — a crawler being handed a shell instead of a page. A setup that commits Markdown and rebuilds a static site cannot make that mistake.

Publishing is also where the tooling stops helping. A site somebody can edit is not a site anybody has read, and the work after that — being found, being chosen, being followed up — is a different discipline from the one that got the page live. That is the gap AgentCeres — the AI Growth Officer at agentceres.com — is built for: a managed AI marketing team that drafts the work and waits for a human to approve anything that goes out. If you are still choosing the stack underneath it, AstroWind is a starting template rather than an editor, and Webstudio is the visual-builder answer to the same question.