Heftli

Invoicing SaaS for freelancers and Swiss SMEs

Role: Sole designer, engineer, operator
Launched: January 2026
Stack: TypeScript, React, Supabase, Stripe, PostgreSQL

Swiss freelancers and small businesses have to issue QR-Bill invoices, a national standard with strict formatting rules that most international invoicing tools either do not support or bolt on badly. The alternatives were enterprise accounting suites with a setup cost measured in days, or spreadsheets.

I wanted something that handled QR-Bill natively, tracked time against clients, and did not require a bookkeeping background to operate.

Retrofitting tenancy is expensive, so row-level security and tenant scoping went in before the first feature. Supabase provides the Postgres layer and auth; every query runs under an RLS policy rather than an application-level filter, so a tenancy bug cannot leak data across accounts.

Stripe handles subscriptions with webhook reconciliation. Billing state is derived from Stripe rather than mirrored, which removes an entire category of drift bug.

The QR-Bill generator is a self-contained module producing Swiss-standard payment parts, validated against the specification rather than against sample output.

Row-level security enforced at the database, not the application
Stripe as the source of truth for billing state
Background jobs for reconciliation and scheduled invoicing
Migrations run in CI against a production-shaped dataset

There is no team. I write the migrations, handle the failed-payment edge cases, monitor the background workers, answer the support email, and ship the frontend.

That constraint shaped the architecture more than any preference did. Anything requiring routine manual intervention got automated or removed, because there is nobody to hand it to. The system has to be boring enough to leave alone for a week.

It is also why the advice I give on architecture is worth taking. Most architects have never operated what they designed.