Shipped a fix for the webhook failures nobody notices until a customer complains
A silent webhook failure cost a client their invoice — and they only found out when a customer complained. Here's what I built to stop it happening again.
Last month a client told me a payment webhook from Stripe just... never arrived on their end. No error, no alert, nothing. Their server had a cold start, timed out, and the webhook was gone forever. They only found out because a customer emailed asking where their invoice was.
That's the part that got me — it wasn't a bug in their code. It was silence. No log, no retry, no way to even know it happened until a real person complained.
So I built Webhook Proxy. It sits between Stripe/Make/Zapier and your actual endpoint, logs every payload instantly, and retries failed deliveries automatically (or with one click if you want to trigger it yourself). Zero SDK — you just swap the destination URL.
Free tier is 5,000 webhooks/month, no card needed, mainly because I wanted real feedback before locking in pricing.
Still early days — no paid users yet, just trying to get this in front of people who've actually been burned by a silent webhook failure. If that's you, curious what you did about it (manual Stripe dashboard replay? custom logging? nothing, and found out the hard way?).
-
1The missing invoice is a good example of why I’d check the resulting application state as well as webhook delivery. A proxy can show that an endpoint returned 2xx, but that alone may not prove the invoice was created or access changed. Do you have a way to detect and reconcile events that were acknowledged but never produced the expected downstream state? That seems like a useful complement to retries.5 minutes ago Reply
-
1That silent failure → customer complaining about a missing invoice is a brutal discovery path. On the freelancer side I see a parallel: the invoice exists, but after a bad week nobody owns the next chase, so the “failure” is just as invisible until cash gets weird. I’m recruiting a few ADHD/ADHD-ish freelancers for a free ~20-min observed silent watch. I’ll provide a fake sample overdue (e.g. Northstar Studio, $1,240, due Sep 16, 14 days late) so you don’t use real client data. You enter it, and we watch what next action + chase message the system surfaces. Not a sales call. Reply YES + your timezone (optional window this week) — I’ll confirm a slot.7 hours ago Reply
-
1Stripe already retries failed deliveries, so the key distinction seems to be what happens after those retries. Have you identified that specific gap in the client incident?a day ago Reply
