A cloud platform marketed on aggressive per-second pricing and an AI-assisted autoscaler that claims to predict traffic before it arrives. Quillhaven offers compute, serverless functions, and a managed queue service across three regions, with a self-service model and support delivered primarily through a ticketing portal and community forum.
Share your experience to help others choose the right provider.
Credit where due, their functions product is quick and the cold starts are genuinely low, maybe 80ms for a small Node handler. I use it for webhooks and it's been reliable. I would not run a database or anything stateful on this platform based on what I've seen, but the functions are fine and cheap.
Signed up on the strength of the pricing page. First month was fine. Second month their autoscaler decided we were about to get a traffic spike, span up 14 extra instances at 3am, and left them running for six hours. We got no spike. There is no toggle to turn the prediction off, only a max instance cap that I hadn't set because nothing in onboarding suggested I should. Took nine days to get a response to the billing ticket and they credited half. Avoid.
Our whole API was down. Status page: all systems operational. Forum full of people saying the same thing. It was updated an hour and forty minutes in. That's the part I can't forgive — the outage itself is one thing, being told nothing is wrong while you're debugging your own code for an hour is another.
There is no phone number. There is no chat. There is a ticket portal where things go to sit for four days and a community forum where other customers try to help each other, which is honestly where I've got most of my answers. For a platform charging real money for production infrastructure this isn't good enough.
Modelled our workload against their calculator and got a number about 45% below what we were paying. Actual bills came in above what we'd been paying before, because the calculator assumes you provision exactly what you need and the autoscaler does not. Their docs are decent and the CLI is fine, so it's not all bad, but the core promise doesn't hold up.
Credit where due, their functions product is quick and the cold starts are genuinely low, maybe 80ms for a small Node handler. I use it for webhooks and it's been reliable. I would not run a database or anything stateful on this platform based on what I've seen, but the functions are fine and cheap.
Signed up on the strength of the pricing page. First month was fine. Second month their autoscaler decided we were about to get a traffic spike, span up 14 extra instances at 3am, and left them running for six hours. We got no spike. There is no toggle to turn the prediction off, only a max instance cap that I hadn't set because nothing in onboarding suggested I should. Took nine days to get a response to the billing ticket and they credited half. Avoid.
Our whole API was down. Status page: all systems operational. Forum full of people saying the same thing. It was updated an hour and forty minutes in. That's the part I can't forgive — the outage itself is one thing, being told nothing is wrong while you're debugging your own code for an hour is another.
There is no phone number. There is no chat. There is a ticket portal where things go to sit for four days and a community forum where other customers try to help each other, which is honestly where I've got most of my answers. For a platform charging real money for production infrastructure this isn't good enough.
Modelled our workload against their calculator and got a number about 45% below what we were paying. Actual bills came in above what we'd been paying before, because the calculator assumes you provision exactly what you need and the autoscaler does not. Their docs are decent and the CLI is fine, so it's not all bad, but the core promise doesn't hold up.
Getting data out of their managed queue service was painful — no bulk export, we ended up writing a consumer to drain it. Also discovered our account had to be manually closed by support, which took two weeks and one more billing cycle. Not accusing anyone of anything deliberate but it was frustrating. Two stars rather than one because the platform did technically work most of the time.