Building a CRM With Claude Code for Local

There's a certain category of YouTube video that exists somewhere between tutorial and infomercial—technically instructive, but also deeply invested in selling you on a particular vision of how the world works. Hamish from Income Stream Surfers has made one of the better examples of the genre: an 11-minute walkthrough of a CRM he built using Claude Code, Astro, Convex, Clerk, Stripe, and a few supporting integrations, all wired together to manage leads for local businesses. The €10k/month figure is right there in the title. The system, to be fair, is real.

What makes the video worth unpacking isn't the income claim—those are ambient noise on YouTube at this point—but the underlying technical argument, which is actually coherent and worth examining on its own terms.

The Problem Being Solved Is Genuine

Before getting to the stack, it's worth taking the business problem seriously, because Hamish does. Most small businesses—a venue, a tradesperson, a hospitality company—get leads through a chaotic mix of web forms, DMs, emails, and the occasional Post-it note. They don't have a Salesforce subscription. They don't have a sales ops team. What they have is someone checking their inbox inconsistently and forgetting to follow up on the inquiry that came in on Thursday.

"Most small businesses lose money on not getting leads, but on forgetting to follow up on leads, on not following up properly," Hamish says in the video. That's not a revolutionary insight, but it's an accurate one. The CRM market has known this for thirty years. The question has always been: how do you get a business owner who's running a venue and organizing hen parties to actually use a CRM? The answer, historically, has been: you don't. They pay for one, ignore it for three months, and go back to the inbox.

The bet being made here is that AI-assisted development has lowered the build cost enough that you can create a custom, stripped-down CRM specifically for one business type—with exactly the fields, automations, and integrations that business needs, and none of the bloat that makes off-the-shelf tools feel like piloting a commercial aircraft to drive to the shops.

What the Stack Actually Does

Hamish's recommended configuration: Astro for the marketing site (chosen over Next.js specifically for its SEO behavior out of the box), Claude Code to scaffold and build the dashboard logic, Convex as the database and backend-as-a-service layer (with Supabase as an alternative), Clerk for authentication and role separation, Stripe for payments with webhook-driven status updates, PostHog for analytics, and Twilio plus Resend for SMS and email automation respectively. Vercel hosts the whole thing.

The lead flow he demonstrates: a prospect fills in an inquiry form on the marketing site → the CRM captures the lead and assigns it "New" status → an acknowledgment email goes out automatically → if no response within 24 hours, a follow-up nudge fires via both email and SMS → once a quote is sent, the CRM status updates to "Quoted" → Stripe generates a payment link embedded in the quote → when payment hits, a webhook marks the lead as "Won."

It's a classic pipeline: New → Contacted → Quoted → Won/Lost. Nothing architecturally novel. What's novel is the claim that Claude Code can scaffold most of this with minimal manual coding—and that this is now accessible to people who aren't career software engineers.

"Claude Code can control for you," Hamish says, describing how the AI handles database schema design, backend logic, auth setup, and API integrations. He notes that you no longer need to hand it documentation to work with third-party APIs—it already knows them.

That last claim is worth holding lightly. Claude Code is genuinely capable of scaffolding complex applications, and the tool has improved substantially in its ability to reason about integration patterns. But "production ready" applications still require debugging, edge case handling, security review, and maintenance. The video shows a working demo; it doesn't show the two weeks of iteration that likely preceded it.

The Interesting Tension Here

There are two distinct things being sold in this video, and they're worth separating.

The first is a technical methodology: use Claude Code to build bespoke, lightweight CRMs for local businesses faster than traditional development would allow. This is a defensible and interesting idea. The tooling genuinely has reached a point where a developer with strong prompting skills can compress weeks of CRUD app development into days. The stack Hamish describes is reasonable—Astro is genuinely good for content-heavy sites, Convex is a solid BaaS choice, Clerk is well-regarded for auth. None of these are unusual or questionable picks.

The second thing being sold is a business model: build these CRMs for local businesses and sell them. "This is a really strong thing to sell to people. I'm telling you people are going mad for CRM stuff right now," Hamish says. And here's where the video becomes less instructive and more aspirational. The gap between "I can build this" and "I can sell this and make €10k/month" contains a lot of territory the 11-minute runtime doesn't cover: client acquisition, scoping conversations, ongoing support, the difference between building a system for your own business and building one that a non-technical business owner can operate without calling you every Tuesday.

The €10k/month figure presumably refers to Hamish's own operation, which includes a venue business (Boundary Park appears to be the live example), an SEO tool (Harbor SEO, which he plugs mid-video), and a broader content and consulting business. That's not a CRM product generating €10k. That's an ecosystem of which the CRM is one functioning component. The conflation is common in this genre of video and worth noticing.

What Claude Code Actually Contributes

The framing of "no-code" is technically misleading but practically interesting. Hamish is writing prompts to Claude Code, reviewing output, debugging, and iterating—that's a form of coding, just not in the traditional sense. What Claude Code compresses is the boilerplate: schema setup, auth configuration, webhook handling, basic CRUD operations. It doesn't replace the architectural thinking. You still need to know that Clerk and Convex need to talk to each other, that Stripe webhooks require idempotency handling, that PostHog events need deliberate instrumentation.

"It's really good for database schema. It's really good for lead tracking and state. It's amazing for live API integration," Hamish says of Claude Code. This broadly tracks with how practitioners are using these tools—AI-assisted development accelerates the well-trodden paths (authentication, standard integrations, data modeling) and still struggles with genuinely novel architecture or security-sensitive logic.

The honest picture is: an experienced developer using Claude Code could build something like this significantly faster than they could without it. A complete beginner using Claude Code could probably get further than they'd expect—but "production ready" would require more careful review than the video suggests is necessary.

The Broader Pattern

What Hamish is describing—stripped-down vertical CRMs built fast with AI assistance, sold as a service to local businesses—is a real market opportunity, and one that several developer-entrepreneurs are quietly pursuing. The traditional CRM vendors (Salesforce, HubSpot, Zoho) have spent years building horizontally, trying to serve every industry. That creates genuine gaps for vertical, purpose-built tools that do exactly what a landscaping company or a wedding venue needs and nothing else.

The tools to build those vertical products have gotten dramatically cheaper and faster to use. That part of the story is accurate. The missing ingredient isn't usually the tech; it's distribution, trust-building with non-technical clients, and the operational weight of supporting software you've sold to businesses that depend on it.

Building a CRM with Claude Code for your own business is a different project than building CRMs as a business. Both might be worth doing. They're just not the same project.

— Dev Kapoor, Open Source & Developer Communities Correspondent, Buzzrag