Here's something we haven't said out loud until now: your WebForCheap website has a full API behind it, and any AI coding agent can drive it.
That means the assistant you already talk to — Claude Code, pi, Hermes, OpenClaw, or whatever harness you run — can update your homepage, publish a blog post, swap out a logo, push the change live, and roll it back if you change your mind. No dashboard. No copy-pasting into a form. You describe what you want in plain language, and your agent does the rest.
We built WebForCheap on AI from day one. This is the natural next step: handing you the same controls we use, in a form your own tools can pick up and run with.
What "agent-managed" actually means
Every WebForCheap site is editable through a clean HTTP API. We've packaged that API as a drop-in skill — a small set of instructions and credentials your agent loads once and then just knows how to use.
Once it's installed, you stop thinking about endpoints and start thinking about outcomes. You say "add a blog post about our new Saturday hours" or "make the hero headline say X," and your agent translates that into the right calls: uploading your content, kicking off a build, checking the result, and showing you a preview link before anything goes public.
The skill works with any modern agent harness. We've tested it inside Claude Code, but the same instructions run just as well in pi, Hermes, OpenClaw, and other tool-using assistants. If your agent can read a file and make a web request, it can manage your site.
Set it up once, locally
Getting started takes about a minute. You add three things to your agent's environment:
- Your site ID — which site the agent is allowed to touch
- A scoped API token — a wfc_... bearer token that grants exactly the permissions you choose
- The API base URL — the WebForCheap endpoint your agent talks to
Drop those into your secrets store (or your agent's config), point it at the WebForCheap skill, and you're done. From then on, every request your agent makes is authenticated automatically. You never paste the token into a chat, and you never hand-edit a config again.
The materials-first workflow
This is the part people love once they see it. Before your agent makes a change, it can upload materials — the raw inputs a build should use.
Materials are exactly what you'd hand a designer: your logo, product photos, a PDF brand guide, a CSV of services, or a plain-text copy deck for a new page. Your agent uploads them straight to your site's build storage, then references them by name when it asks for the change.
So instead of trying to cram an entire page of copy into a single instruction, the flow looks like this:
- Your agent writes your content to a file and uploads it as a material
- It starts a build that says, in effect, "create this page from the file I just uploaded"
- The build pulls the material in automatically and renders it into your site's design
It's the same way we work internally, and it's why the results look like a designed website instead of a wall of pasted text. Images, PDFs, and text files are all supported, with sensible size limits so a build never chokes on a giant upload.
Builds in minutes, not days
When your agent starts a build, the WebForCheap engine figures out whether you want a small revision or a full rebuild.
Most changes are revisions — surgical edits like updating a headline, adding a section, or dropping in a new blog post. Those finish in roughly two to four minutes. A full rebuild — starting a site over or doing a major redesign — takes a little longer. Either way, your agent polls the build until it's done and tells you the moment it's ready.
You get a live progress signal the whole time: what step the build is on, how far along it is, and a clear "complete" or "failed" at the end. No refreshing a dashboard wondering if anything is happening.
You always see it before the world does
Nothing your agent builds goes live automatically. Every change lands on a private preview URL first.
Your agent hands you that link, you look at the real page exactly as visitors would see it, and only then do you decide. When you say "ship it," the agent promotes the preview to production. Until you say so, your live site stays untouched.
We made this deliberate. An AI agent is fast and capable, but publishing to real customers is your call, every single time.
Made a mistake? Roll it back
Every build is versioned. Your agent can list the full history of changes — each with its timestamp and the instruction that created it — and roll your live site back to any previous version in seconds.
So if a change doesn't land the way you hoped, there's no panic and no support ticket. You just tell your agent "go back to yesterday's version," and it does.
Scoped, safe, and yours
The token your agent uses is locked down on purpose. It can only touch the site you've assigned it to — never your billing, your account settings, or anyone else's site. Each token carries only the permissions you grant: read-only, build-and-edit, or full publish rights.
That means you can hand a junior teammate's agent a token that can draft and preview changes but can't publish, while your own agent keeps the publish button. The API enforces it on every request, so there's no way to color outside the lines.
Why we're doing this
We started WebForCheap to make a real, well-built website affordable for small businesses. Putting the controls directly into your AI agent is the same idea, taken one step further.
You probably already have an assistant you trust for everything else. Now it can run your website too — write the copy, design the page, preview the result, and publish on your word. That's web management that fits the way you already work.
If you're a WebForCheap customer and want your agent wired up, reach out and we'll get you a token and the skill. If you're not a customer yet, this is one more reason to be.
