Done for you · From $2,500

Get your own
AI-ready website.

I build the site, wire in the agents that maintain it, and hand you the keys. They keep working after I'm gone.

Price
From $2,500
Timeline
About 2 weeks
You keep
Repo, accounts, code

The problem with a normal website

Someone builds it, it looks good for a year, and then it sits there. The blog stops in March. Internal links rot as pages get renamed. Nobody notices the post that used to bring 4,000 visits a month has quietly dropped to 200, because noticing is somebody's job and that person is busy.

This is the same site plus a set of agents whose entire job is the noticing. They read your Search Console and analytics on a schedule, find the pages that are slipping, and open a pull request with the fix. You approve or you don't.

I run 100+ agents across my own nine sites on exactly this setup. That's what I'm installing for you.

Running it now

Four sites on this stack

Different shapes — a local business, a publication, a personal site, a SaaS product. Same harness underneath. Every number below is the site's own public claim, not a projection.

Pickleland homepage — indoor pickleball courts in Austin
Local business · Austin, TX

Pickleland

Nine indoor courts in Pflugerville. Bookings, memberships, events, and a Spanish locale. The agents keep the class and event pages current, which is the content that actually converts for a venue.

pickleland.com →
Future Sharks homepage — founder interview publication
Publication · since 2017

Future Sharks

1,042+ founder stories and 100K+ monthly readers. The highest-volume site on the stack, and the one that proves the content agent holds up past a few hundred posts.

futuresharks.com →
alejandrorioja.com homepage
Personal brand · 13 locales

This site

Roughly 5,000 posts across thirteen languages. The SEO agent runs weekly here and the metadata budgets are enforced at build time, so a bad title fails CI instead of shipping.

alejandrorioja.com →
Mentioned homepage — AI answer visibility tracking
SaaS product

Mentioned

Tracks whether AI assistants name your brand. A product site with an app behind it — the case for the stack handling marketing pages and a real dashboard in one repo.

mentioned.at →

What actually runs after I leave

Five jobs on a cron, each one opening pull requests against your repo:

  1. 01
    SEO agent — weekly. Reads Search Console and GA4, ranks your pages by what's decaying fastest, and drafts the rewrite. Catches the slow bleed that nobody has a calendar reminder for.
  2. 02
    Content agent — three times a week. Drafts against a brief and a voice file you write once during setup. It produces a first draft that needs an editor, which is the honest description of what these models do well.
  3. 03
    Link and 404 healer. Crawls the built site, finds internal links pointing at pages that moved, and repairs them. Every dead inbound URL gets a redirect to the closest live page instead of a 404.
  4. 04
    Freshness watch. Flags posts whose facts, prices, or screenshots have aged past a budget you set per section. Fixing a pricing page that's quietly two years out of date beats publishing three new posts.
  5. 05
    Performance and security budgets. The build fails if a page gets heavier than the budget, if the security headers regress, or if a required piece of metadata goes missing. These run on every deploy, not on a schedule.
The part I care most about. Every check distinguishes "I reached your site and it served the wrong thing" from "I never got an answer." That sounds pedantic until an agent with a blocked network connection confidently reports your newsletter is dead and you spend a day on it. That happened to me. The fix is in the code you're buying.
Optional bolt-ons

Two things that make it sharper

Neither is required and neither is bundled into the price. Both are mine, both plug into the same agents, and I'll wire either one up during setup if you want it.

SiteWide MCP · $9/mo

Give the agents real numbers

Out of the box the agents read your Search Console and analytics through whatever access you've set up. SiteWide is one connector that hands them GA4, Search Console, Google Ads, Bing, and Ahrefs together — so "which pages are decaying" is answered from live data instead of a guess. It's the difference between an agent that rewrites what it thinks is slipping and one that rewrites what actually is.

See SiteWide MCP →
Mentioned · free tier

Track whether AI names you

Ranking on Google and getting named by an AI assistant are now two different jobs, and only one of them shows up in Search Console. Mentioned watches whether assistants name your brand when a buyer asks about your category, and which competitors get named instead. Feed that back into the SEO agent and it has a target that Google's numbers can't give it.

Check your brand on Mentioned →
The engagement

About two weeks, four phases

I work in your repo from day one, so you watch it happen rather than getting a reveal at the end.

Days 1–2

Scope and access

We agree what the site has to do, I get collaborator access to your GitHub and Cloudflare, and if there's an existing site I audit what's worth carrying over.

Days 3–7

Build and migrate

The site goes up on your Cloudflare account. Existing content converts to Markdown, media moves, and the old URLs get a redirect table so your rankings survive the switch.

Days 8–11

Wire the agents

Charters, schedules, budget caps, and the build gates. We write your voice file together — it's the input that decides whether the drafts are usable.

Days 12–14

Handoff

A live walkthrough, written docs in the repo, and I watch the first real agent run with you. Then I remove my access.

What it costs you every month after that

I'd rather you see this before you buy than discover it in month two. For a site publishing a few times a week:

$0–5
Cloudflare hosting
$20–200
Your coding-agent plan — $20 covers most
$30–90
Agent token spend
$0
Owed to me after handoff

Every agent run carries a hard dollar cap I set during setup, so a job that goes sideways stops itself rather than running up a bill. You can lower those caps any time — they're two lines in a config file, and I show you where.

When you shouldn't buy this

Three cases where I'll tell you to go elsewhere, and I'd rather say so here:

  • You want a five-page brochure site. A designer on Webflow will do it better and for less. The agents only earn their keep if you're publishing regularly.
  • Nobody on your side can approve a pull request. The whole safety model rests on a human saying yes. Without that you either merge everything blind or nothing ships, and both are worse than not having the agents.
  • You want the writing handled permanently. This gets you a machine that produces drafts. Somebody still edits them. If what you actually want is for content to be off your plate entirely, hire a writer — that's a retainer, and a different conversation.

The setup pays for itself when you're publishing at least a few times a month and someone currently spends their Friday afternoons on it.

Setup — common questions.

What do I need to have before we start?
Three accounts, all in your name: GitHub, Cloudflare, and a subscription to whichever coding agent you want driving the work. If you already have a domain, keep it where it is — I point DNS at Cloudflare without moving the registrar. I never take ownership of your accounts; I work as a collaborator you can remove the day we finish.
Does this lock me into one AI provider?
No, and it's worth being precise about what that means. The runner is a shell script that hands a prompt to a coding-agent CLI and commits what comes back — the model is a swappable part, not the architecture. It ships wired for Claude because that's what I run on my own sites and what I've tested at volume. Pointing it at a different CLI is a contained change to one file, and if you tell me during scoping which one you want, doing it is part of setup rather than a change order. What I won't tell you is that every model is equally good at this: the drafting quality gap between the frontier models and the cheap ones is real, and you'll feel it in how much editing each draft needs.
Can I run it on a free model?
You can point it at a free tier or a local model, and for the mechanical jobs — broken links, redirects, stale-date flagging, metadata — that works fine, because those are rule-following tasks rather than writing tasks. Where free tiers fall down is the two that matter most: drafting and SEO analysis. The free tiers also rate-limit, which a scheduled fleet hits quickly. My honest advice is to run the cheap paid tier, around $20 a month, and put the free models on the mechanical jobs if you want to trim the bill. I'll set it up either way and tell you which jobs I'd move.
What does it cost to run after handoff?
Cloudflare hosting is $0 to $5 a month for most sites — static pages on the free tier, and the form handlers stay well inside the free request ceiling. The coding agent is $20 to $200 a month depending on the plan; the $20 tier is enough for a site publishing a few times a week. The agents spend tokens on top of that, and I set a hard per-run dollar cap during setup so a runaway job stops instead of billing you. Budget $50 to $150 a month all-in.
Do I own the code?
Yes. It lands in your GitHub organization under a permissive license, and it keeps working whether or not you ever talk to me again. There is no license server, no phone-home, and nothing that expires. If you want to hand the repo to a different developer in six months, nothing stops you.
What happens when an agent writes something wrong?
Agents open pull requests. They do not push to your live site. You or someone on your team approves every change before it goes out, and the build refuses to merge anything that breaks an internal link, drops required metadata, or blows the page-weight budget. The realistic failure mode is a draft that reads flat and gets rejected, which costs you a few cents in tokens.
Can you migrate my existing WordPress site?
Usually. I have scripts that pull the post corpus, download the media, convert to Markdown, and build the redirect table from the old URLs so you don't lose rankings. Migration is the single biggest driver of scope: a 40-post blog is straightforward, a 4,000-post site with custom fields and a decade of plugin cruft is a bigger conversation. Send me the URL and I'll tell you which one you are before you pay anything.
How is this different from hiring someone on Upwork?
Most of what I'm handing over is the harness that keeps agents from doing damage — the lockfiles that stop two agents pushing at once, the budget caps, the build gates, the rule that a check which cannot reach your server reports 'inconclusive' instead of announcing your site is down. I wrote that because it broke on my own sites first. You're paying for the failure modes I already found.
Do you keep working on it after handoff?
Only if you want that, and it is a separate monthly arrangement rather than part of this price. Most people take the handoff, run it themselves, and email me a few times in the first month. That is the outcome I am aiming for.

Send me your site.

Tell me the URL and what you publish. I'll come back with a scope and a number, or tell you this isn't worth it for you. Both are free.

↵ to see all results esc esc to close