Why not just use Astro and run it yourself?

There is a question going around every team that builds websites right now, and it usually arrives with a shrug: if code is where the agents work best, why not just use Astro or Next.js, put it on Cloudflare or Vercel, and run it ourselves? The people asking are designers, marketers, and agency owners who have spent a decade inside a CMS. They are right to ask, and the honest answer is short. You should. Then something changes that nobody mentions, and it is the whole reason Dispatch exists.

Yes, own the code

The case for leaving is stronger than it has ever been. A content-led site on Astro renders to static files and runs on an edge network for close to nothing. An app-like site on Next.js gets previews on every branch out of the box. An AI coding agent reads the whole codebase, not a form field, so it can change a layout, a schema block, and a redirect in one pass. The vendors know this, which is why every one of them is now telling you that code is the foundation. When the CMS company says the abstraction layer is the ceiling, believe it.

So take the win. Your code in your repository, your content in your files, your history in git, your host chosen for the site rather than sold to you. Nothing about that needs a platform, and nothing about it locks you in.

The part nobody mentions

The CMS was never really about editing. It gave a marketing team three things it did not know it was paying for: a place to ask for a change, a publish button with permissions behind it, and a version history someone could roll back. Those three things are the reason a non-engineer could be trusted with a production website. The day you move to code, all three go away at once.

• The place to ask becomes a chat window with an agent, and the request lives nowhere once the window closes.
• The publish gate becomes a pull request, which is a fine gate for engineers and an invisible one for everyone else.
• The rollback becomes git, which can restore anything, provided you know which commit, and which branch, and what the deploy did.

Multiply that by an agent that can ship twenty changes a day and you get the failure mode the code-first argument skips over: not a broken site, but a site nobody can account for. Who asked for this? Who approved it? What did it change? Can we put it back? A team that cannot answer those questions will, quite reasonably, stop letting the agent work. The capacity you left the CMS to get quietly goes unused.

What a management layer puts back

This is the layer Dispatch adds above the stack you just chose. It does not build, host, or edit anything. It governs the work: a request in plain language, a preview before anything ships, an approval owned by a named role with a second approver for high-risk changes, every human and agent action on the record, and a one-click restore that lands as a normal commit. Bring your own agent, whether that is Claude Code today or something else next year, and keep your own repository and host. Dispatch governs the work; other systems execute the work.

Then it goes one step further than the CMS ever did. Dispatch re-crawls every page on a schedule, scores it for answer engines, opens a finding when something slips, and drafts the fix as a request before anyone asks. The team keeps the approve moment; the initiation stops depending on someone noticing. That is the Inbox, and it only makes sense on a code-first site, because only there can the fix be executed by an agent the same afternoon.

If you are the designer or the agency

A lot of the shrug is coming from people whose value was the platform itself: the designer who knew the visual editor, the agency whose playbook was built on one CMS. That value does not disappear when the platform does; it moves up a level. Someone still has to decide which architecture a client needs, migrate the site without losing its design or its rankings, set the roles and risk tiers, and hand it off in a state the client's team can operate. That is a better business than template work, and it is platform-neutral by construction: you orchestrate every stack your clients build on instead of living inside one. We wrote the playbook for it: the agency handoff playbook.

So, why not just use Astro and run it yourself?

Do exactly that. Then give your team back the three things the CMS used to provide, in a form built for agents rather than for form fields. If your site is still inside a CMS, Dispatch Migration Services will assess it, pick the right stack for it, move it pixel for pixel, and connect it. If it is already on code, connect it to Dispatch and the loop is running this afternoon, free for one website.