Infrastructure control plane for your coding agents. Agents deploy, you stay in control.
New York 🇺🇸 / Paris 🇫🇷
Joined October 2019
- Tweets2.7K
- Following956
- Followers3.1K
- Likes2.4K
Pinned Tweet
You missed our “Heroku to AWS in One Command” live session yesterday - no problem. Here’s the summary and recording: qovery.com/blog/heroku-to-aw…
1/ We planned to move every Qovery-managed cluster from ingress-nginx to Envoy Gateway by the end of March. It took until October. 🧭
3/ What slowed us down: Envoy is stricter than NGINX, Gateway API policies replace each other instead of merging, and minor versions kept changing behavior we relied on.
4/ Our worst incident came from our own code. A default flipped, deployments failed for about two hours, and customers told us before our alerts did.
qovery.com/blog/nginx-to-env…
1/ On busy AWS clusters, every system image pulled from public.ecr.aws goes through your NAT Gateway, and those processing charges add up. ⚡
3/ You pay a little for private ECR and a VPC endpoint, and save on NAT Gateway charges, most on clusters with heavy node churn. Available on Qovery-managed EKS with Karpenter, a Qovery-managed VPC and NAT Gateways.
4/ When did you last look at the NAT Gateway line on your AWS bill?
qovery.com/changelog/2026-09…
1/ One team spent 12 weeks evaluating Qovery, decided it didn't need it, had two incidents, and came back. The migration from ECS to EKS then took 5 days. ⏱️
3/ The second team had three engineers and no infrastructure person. Staging was running on Qovery by the end of the kickoff afternoon, and the decision still took 7 weeks. Keeping the old stack running, so they could switch DNS back, made the decision easier.
4/ Guillaume's three ways to cut decision time: make the move reversible, make it cheap to prove, and migrate in waves with the database last.
What stalled your last migration decision?
Replay and chapters: qovery.com/blog/migration-ta…
1/ An AI agent with production access should run inside your cloud, next to the workloads it touches. 🔐
3/ The Qovery control plane only processes deployment metadata: configuration state, resource metrics and deploy events. We don't host your workloads, and on Enterprise you can self-host the control plane itself.
4/ Every run shows its trigger, duration and cost. Where do your agents run today, your cloud or a vendor's?
qovery.com/sovereignty
Qovery retweeted
the demo was the easy part 🫠
Second Opinion #2 is about everything after it: safety, prod access, and the token bill
Nov 12, Paris, with @avec_alan , @Qovery_ & @AWS
luma.com/4lhbur28
1/ During a Heroku exit, you pay Heroku and AWS at the same time for roughly 4 to 8 weeks. That overlap is what quietly drains the budget. 💸
3/ Activate is the startup path. MAP is the bigger pool for a migration, and an AWS Partner drives the request for you. Watch NAT Gateway and data transfer, the two charges that surprise people.
4/ Did your migration budget include the overlap?
qovery.com/blog/fund-heroku-…