Shipping a production SaaS as a team of one with Claude
Her Aviation Era runs in production with paying users, three payment gateways, governed database migrations, and daily encrypted backups. It was built and is operated by exactly one engineer. This is how.
Treat the agent like a teammate, not autocomplete
The biggest mindset shift is moving from "AI writes snippets" to "AI owns workflows." My agent doesn't just write components — it creates feature branches, applies migrations to staging, watches CI, opens pull requests, and merges after my in-chat approval. I review outcomes, not keystrokes.
Guardrails make speed safe
Solo doesn't mean reckless. Every change flows through the same pipeline a good team would use:
- Feature branch → staging environment → production, with verification at each stage
- CI gates (lint, type-check, build) on every pull request
- Database migrations as append-only files, applied to staging first, gated behind a manual confirmation for production
- Row-Level Security audited on every table, and a daily encrypted 3-2-1 backup pipeline
The agent executes these steps; the process ensures a bad day can't become a bad month.
Iterating on monetization: the unified payment page
The same workflow let me rebuild monetization end-to-end. The scattered buy-buttons became a single checkout page backed by an orders-and-entitlements ledger, with a separate payment package for each product — membership, CV templates, coaching sessions — so users pay only for what they need instead of facing one all-or-nothing price. Every purchase also unlocks a free toolkit bonus, a small freebie that gives buyers an immediate win and pulls fence-sitters over the line. Designing, shipping, and verifying that whole surface — UI, ledger schema, gateway wiring — followed the exact same branch → staging → approval pipeline as any other feature.
Where the human stays in the loop
Product judgment, approval gates, and anything irreversible. I decide what to build and when to ship; the agent compresses everything between those two decisions. That division of labor is what makes a team of one feel like a team of five.
More detail on the architecture is in the Her Aviation Era case study.