The Tech Stack Guide · A free resource from Matt Mayo's AI talk 10 pages · US Letter · Print or save as PDF
m·mMortgage Mayo
A Free Resource
From Matt Mayo's AI Talk

The Exact Tech Stack I Used to Build mortgagemayo.com with AI.

A practical guide to the tools, what they do, why I chose them — and what to use instead if you're not technical.

This is the actual tech stack behind mortgagemayo.com — the site that produced ~$5M in pipeline in 75 days. I built the initial version the same day I took Luke Shankula's Build Your Website with AI training, then spent the next several weeks of evenings and weekends iterating — adding hyperlocal pages, FAQ schema, conversational content, and the AI-readability optimizations I talked about in the conference session.
This guide is for two audiences: (1) Technical LOs who want to copy or adapt the exact stack, and (2) non-technical LOs who want to know what to look for if they hire someone — or what no-code platform to choose instead.
Matt Mayo
mortgagemayo.com
~ $40/mo
All-in stack cost
SummarySix tools · Two audiences · One stack

The stack at a glance.

ToolWhat it doesCostNon-technical alt.
AstroFront-end framework.FreeWebflow, Framer, Squarespace
SanityHeadless CMS.Free tierWordPress, Webflow CMS
VercelHosting + deployment.Free tierWebflow Hosting, Squarespace
Tailwind CSSStyling system.FreeBuilt into Webflow / SS
Cursor + Claude OpusAI dev partner for the initial build — visual code editor with Claude's most capable model.$20/mo
Cursor Pro
Hire dev ($5K–$10K)
Claude CodeAI coding tool for ongoing maintenance — use in browser, desktop app, or terminal (pick what's comfortable).$20/mo
Claude Pro
Same as above
Monthly cost
~$40/mo
One-time cost
$0
Time
4–6 weeks evenings + weekends

NoteYou don't strictly need both AI tools — each fits a different workflow.

Non-technicalThe path I'd take · And did

For the non-technical LO.

If you don't write code and don't want to learn, you have a few real paths. I'll be honest about which one I'd recommend first — because it's literally what I did.

Why it winsApples to apples · Three paths compared

Why this path wins.

Luke's workshop Hire a developer No-code (Webflow etc.)
Upfront cost $97 $5,000–$10,000 $14–$39/mo forever
Time to live site One workshop session 4–6 weeks Few hours
Ongoing cost $20–40/mo (AI tools) + domain ~$15/yr $20–40/mo if updating with AI; or pay dev per update $14–$39/mo to platform forever
You own the site Yes — completely. Yes, but hard to update without the dev Sort of — you rent the platform
Update yourself Yes, using AI No (unless you also subscribe to AI tools) Yes, in platform
Same architecture as Matt's Yes (same stack) Depends on dev No
AlternativesIf Luke's workshop isn't the fit

If that's not the right fit…

Two other paths. Both work — they just cost more, take longer, or trade off ownership.

Alt 01 · Hire a developer

Hire a developer to build the same stack.

Cost
$5,000–$10,000 one-time, plus $20/mo to update yourself later.
Timeline
4–6 weeks if you have content ready.
Pros
Same flexibility, performance, SEO results as my site.
Cons
You don't know how to update it — back to the dev (or learn anyway).
Where
Upwork · Toptal · Fiverr Pro · Twitter/X (search "Astro developer" or "Sanity developer").

The spec to hand them:

Alt 02 · No-code platform

Use a no-code platform.

Webflow $14–$39/mo
Closest to what I built. Has CMS, supports custom code for schema.
Framer $5–$25/mo
Newer alternative. More design flexibility.
Squarespace $16–$49/mo
Easiest setup. Less flexible for custom schema.
Trade-off to know: with no-code platforms, you don't truly own the underlying site — you rent the platform. If pricing changes, features sunset, or the platform shuts down, you have to migrate. With Luke's workshop path, you own the code outright.
Stack detailsTool by tool · Why I chose each

For the technical LO.

Six tools. Each gets the same treatment: what it does, why I chose it, what I considered instead, and where to read the docs.

01 · Front-end framework

Astro— the front-end framework

What it does
Generates static HTML by default, with optional dynamic routes. Ships almost no JavaScript by default — insanely fast page loads, which Google rewards.
Why I chose it
Best performance-per-effort of any modern framework. Excellent SEO out of the box. Plays nicely with Sanity. AI tools write Astro really well.
Considered
Next.js (more powerful but heavier), Eleventy (lighter but less batteries-included), Gatsby (older, slower).
Docs
astro.build/docs
02 · Headless CMS

Sanity— the content management system

What it does
Stores your content as structured data, separate from the front-end. You edit content in Sanity Studio and the front-end pulls it in.
Why I chose it
Real-time collaborative editing. Strong content modeling. Free tier covers small-to-medium sites. Excellent developer experience.
Considered
Contentful (more expensive), Strapi (open source, self-hosted = more work), Storyblok, DatoCMS.
Docs
sanity.io/docs
03 · Hosting & deployment

Vercel— hosting and deployment

What it does
Hosts your Astro site, auto-deploys when you push to GitHub, edge caching, analytics.
Why I chose it
Best deploy-from-git experience. Free tier is generous. Edge functions let me do SSR for dynamic location routes.
Considered
Netlify (close second), Cloudflare Pages (cheaper at scale), GitHub Pages (limited).
Docs
vercel.com/docs
04 · Styling

Tailwind CSS— the styling system

What it does
Utility-first CSS. Compose styles directly in HTML using class names.
Why I chose it
AI tools write Tailwind extraordinarily well — effectively the lingua franca of AI-assisted front-end development. Performance excellent because unused styles are stripped at build time.
Considered
Vanilla CSS, CSS Modules, styled-components.
Docs
tailwindcss.com/docs
05 · The initial build

Cursor + Claude Opus— AI dev partner for the initial build

What it does
Cursor is an AI-first code editor (a fork of VS Code with AI deeply integrated). Inside Cursor you pick which AI model writes the code — I used Claude Opus, Anthropic's most capable model for complex coding work. Describe what you want in plain English, Claude writes the code, changes show up as a diff, you review and accept or iterate.
Why I chose it
I'm not a professional developer. Cursor with Opus let me describe features in plain English and get production-quality code back. Opus specifically punches above its weight on complex multi-file changes — building hyperlocal page templates with Sanity integration would have been brutal without it.
Cost
$20/mo for Cursor Pro · includes generous AI usage.
Docs
cursor.com
06 · Ongoing work

Claude Code— what I use now for ongoing maintenance

What it does
After the initial build, I shifted day-to-day site work to Claude Code — Anthropic's AI coding tool. Same Claude model under the hood, different interface from Cursor.
Where to use
  • In your browser at claude.aithis is what I use most
  • In the Claude desktop app
  • In a terminal/CLI — more advanced, optional
  • As an extension inside VS Code or other IDEs

The browser and desktop versions are completely approachable — you're not in a terminal. It looks like a regular Claude chat, except Claude has direct access to your project files.

Why I switched
Once you know your codebase, Claude Code is faster for targeted changes. You describe what you want, Claude reads the relevant files, makes changes across multiple files at once, and shows what it did. Significantly faster than the diff-by-diff loop in Cursor for maintenance work.
Cost
Included in Claude Pro — $20/mo.
Docs
docs.claude.com
Use Cursor + Opus when

Initial build. Complex feature work. Anything where you want visual diff review.

Use Claude Code when

Ongoing maintenance. Multi-file edits. Anything where you've described what you want clearly.

Other alternatives I considered
LessonsFive things · Looking back

What I'd do differently today.

  1. 01

    Start with the content, not the code.

    I spent the first week obsessing over framework choice. The content — what each page actually says — turned out to matter ten times more. Write the headlines, FAQs, and city pages in a Google Doc first.

  2. 02

    Use existing templates more aggressively.

    Astro has free, production-quality starter templates. So does Tailwind UI. Starting from a polished template and editing down is faster than building from scratch — and easier on AI.

  3. 03

    Don't over-engineer for traffic you don't have yet.

    Edge functions, ISR caching, custom analytics pipelines — I built things for 100K monthly visits while sitting at 800. Ship the simple version. Re-architect when the traffic exists.

  4. 04

    Get the FAQ schema right from day one.

    Retrofitting structured data after the fact is annoying. Bake JSON-LD into your page templates before you publish anything.

  5. 05

    Hire a designer for one day.

    I'm decent at design but not great. A four-hour session with a real designer on type, spacing, and color would have saved me weeks of "this doesn't look right" iteration. Worth $400–$800 even at the very start.

NumbersWhat this actually costs · All in

What this stack actually costs me.

ItemMonthlyAnnual
Astro$0$0
Sanity (free tier)$0$0
Vercel (free tier)$0$0
Tailwind$0$0
Cursor Pro$20$240
Claude Pro (for Claude Code)$20$240
Domain~$1$12
Total~$41~$492
You don't strictly need both AI tools. Cursor alone works if you stay in the visual editor. Claude Code alone works if you're comfortable in a terminal from day one. I happen to use both because they fit different parts of my workflow.
Questions about the stack — or which path to take?

DM @mortgagemayo on Instagram.

Happy to point you in the right direction — whether that's hiring someone, going no-code, or picking up Claude Code yourself.
P.S. If you saw LoanDeck at the end of the talk and want a closer look at the loan comparison tool I'm building, here's the waitlist: loandeck.app. Beta users lock in 50% off launch pricing.