← Case Studies
Product · SaaS·2026·Founder · Migration & infra

ChatDrive

ChatDrive is a live-chat widget businesses embed on their sites with a single script tag. It began life entirely on a no-code platform — fine to start, a liability to scale. The job was to move it onto infrastructure I own and can secure, without asking a single customer to touch their code.

Visit the live site
ChatDrive
1 script tagcustomers never re-install
Zero-downtimereversible cutover
Owned stackVercel · Supabase · Resend
Billing-criticalrevenue rides on uptime
01

The problem

The whole product — widget, backend, and billing — lived on a no-code host. Customers had already pasted an embed snippet onto their own sites, so any change that forced a re-install would mean lost installs and lost revenue. And because billing depends on the widget staying live, the migration had to be effectively invisible: the kind of move where a single dropped event costs money.

02

The build-vs-buy call

Own the stack. Move hosting to Vercel and the backend to Supabase and Resend — infrastructure I control, can secure, and can scale — while preserving the exact embed contract so no customer ever knows it happened.

03

The approach

A migration designed to be reversible and zero-downtime from the start: preserve the embed script and its API contract, stand up the new backend in parallel, bridge legacy traffic so old and new run side by side, and cut over by DNS with a rollback path the entire way.

The security gaps that come free with a no-code backend get closed in the move — real ownership checks on every route, embed-domain allowlisting, and rate limiting where none existed. Migrating is the moment to fix them, not paper over them.

04

Why it matters

This is the work I mean when I say strategy and execution come from the same hands: the person making the call to leave the platform is the same person writing the migration — and carrying the risk if a single event drops.

What made it work

  • Preserved the one-line embed — zero customer re-installs
  • Reversible, zero-downtime cutover with a legacy-traffic bridge and rollback path
  • Moved to owned infrastructure: Vercel + Supabase + Resend
  • Security tightened in the move — ownership checks, domain allowlisting, rate limits
Next.jsSupabaseVercelResend

Want something built like this — where the person who makes the call is the one who ships it?