Been building this for 3 weeks. Here's the honest technical breakdown.
What I Built
A fully automated marketing company run by AI agents:
- CalcFuel â 20+ free marketing calculators (top-of-funnel)
- A $19 AI prompt pack â sold via Stripe, auto-delivered via webhook
- Email opt-in funnel â MailerLite list building from calculator pages
- Bluesky/Mastodon auto-posts â cron jobs push tool updates to social
All orchestrated by Claude AI agents running on Paperclip with zero human code written.
The Stack (All Free Tiers)
| Layer | Tool | Cost |
|---|---|---|
| Frontend | Next.js + Vercel | $0 |
| MailerLite | $0 | |
| Payments | Stripe | 2.9% + 30¢ per sale |
| AI agents | Claude via Paperclip | Monthly |
| Analytics | Vercel Analytics + GA4 | $0 |
| Social | Bluesky API, Mastodon API | $0 |
The only real cost is the Claude API usage for the agents themselves.
The Automated Funnel Flow
Visitor lands on CalcFuel calculator
â Runs a free calculation
â Sees email opt-in CTA ("Get your full marketing checklist")
â Joins MailerLite list (opted-in)
â Receives welcome sequence (AI-authored)
â Welcome email links to $19 prompt pack
â Stripe checkout â webhook fires
â PDF delivered automatically within 30 seconds
Zero manual steps. I could be asleep and it all works.
The Part That Surprised Me
The Stripe webhook auto-delivery was surprisingly simple:
// app/api/stripe-webhook/route.ts
export async function POST(req: Request) {
const event = stripe.webhooks.constructEvent(body, sig, secret);
if (event.type === "checkout.session.completed") {
const session = event.data.object;
const email = session.customer_details.email;
// Generate personalized marketing plan with Claude
const plan = await generateMarketingPlan(session.metadata);
// Send delivery email with PDF + plan
await sendDeliveryEmail(email, plan);
}
}
The customer gets their download before they've even left the "thank you" page.
What Actually Drives Traffic
The free calculators on CalcFuel are the real traffic engine. Tools like:
People searching for "calculate marketing ROI" land on the free tool, get value immediately, and see the email opt-in. No ad spend. Pure SEO + organic.
The AI Agent Architecture
The agents run on a 15-minute heartbeat:
- Check inbox â any new Paperclip tasks assigned?
- Execute â code, deploy, write content, post to social
- Update status â comment on the issue with results
- Exit â no persistent state between runs
Each agent is a fresh Claude context. Shared memory lives in markdown files that every agent reads on startup. It's surprisingly robust.
The Hard Numbers
- Calculators built: 35+
- Dev.to articles published: 13
- Bluesky posts: 5
- Mastodon posts: 5
- Email opt-ins: still growing
- Revenue: working on it ð
The goal is $200 AUD by May 18. We're running the experiment in public.
Lessons Learned
What worked:
- Free tools as top-of-funnel convert better than cold content
- Automated social posting keeps the presence alive even during "off" hours
- Stripe webhooks + Claude = surprisingly good auto-fulfillment
What didn't:
- Twitter/X API is dead for free tier (402 CreditsDepleted on every call)
- Reddit new app registration is blocked for everyone now (Nov 2025 change)
- Sending too many similar articles to Dev.to tanks your account reach
The surprising insight: AI agents are great at building things but need very specific instructions to do the right thing. The challenge isn't the technical capability â it's the governance (making sure agents don't violate spam laws, run up costs, or build the wrong thing).
The Code
CalcFuel is on GitHub: getmarketingai-byte/calcfuel
Fair warning: it was written by AI agents, so the code quality is... AI-agent quality.
Try the Tools
- CalcFuel.com â All the free calculators
- 50 AI Marketing Prompts â $19 AUD â The paid product this funnel sells
- Support CalcFuel â ($5) â If this was useful
Happy to answer questions about the agent setup in the comments.
Top comments (0)