No DevOps team? No problem.

Insy is a platform for creators who monetize communities and digital products: subscriptions, payments and Telegram and Discord integrations. The infrastructure was built alongside the product, starting in November 2025, by a small team with no dedicated ops department.

Six weeks later, production was running on Kubernetes. Here is the approach that made it possible.

One rule: everything as code

Nothing is clicked together in a console. Every layer of the platform is declared in a git repository, reviewed like application code and reproducible from scratch:

• Terraform provisions two Kubernetes clusters (Talos Linux) on Hetzner.
• ArgoCD keeps what runs inside them in sync with git: 22 platform components and 24 production apps.
• NixOS defines the support servers, down to the last package.
• devenv gives every developer the same stack locally, started with a single devenv up.

That principle drove most decisions. Local services use the same bucket names, regions and domain structure as production, so the code does not need a single if (dev) branch.

Releases without ceremony

Commits follow Conventional Commits, and every push to the main branch calculates the next semantic version automatically. A release generates the changelog, a GitHub Release and a team notification. The result so far: 256 releases with zero manual version bumps.

CI only rebuilds the services a change actually touches, and staging deploys merge themselves. Production stays behind one deliberate click in ArgoCD, so a human always signs off.

Observability from day one

Metrics, logs and traces flow into one place from the start. Beyond CPU and memory, the apps export product metrics: active users, funnel events and queue backlog. When something slows down, the team sees what it means for users, not just for servers.

Want to see the product side of this project? Read the Insy case study or get in touch if you are building something similar.