Coolify orchestration shifts and the Dokploy versus CapRover choice

Resource overhead and the Heroku exit

Coolify commands the most attention in the self-hosted PaaS space with 60,955 GitHub stars as of June 2026. I find the resource overhead of Coolify problematic for teams running on tiny VPS instances. While it offers 280 plus one-click services, the platform consumes 5 to 6 percent of idle CPU and 500 to 700MB of RAM, which can strain small VPS setups. This leaves much less breathing room for applications compared to Dokploy, which uses only 0.8 percent idle CPU and 350MB of RAM. I would avoid Coolify if your budget forces you onto a 2GB RAM server where every megabyte counts for your database. The movement toward these tools follows Heroku’s February 6, 2026, decision to focus on stability instead of new features. A $3.79 Hetzner CX22 provides 4GB of RAM, making it a tempting alternative to a $25 Heroku Standard-1X dyno. Coolify provides git-push deploys and supports Dockerfile and Docker Compose. It offers one-click databases like PostgreSQL and Redis. It works with GitHub, GitLab, Bitbucket, and Gitea, and it provides per-branch preview environments per pull request.

Swarm orchestration and license limits

Dokploy and CapRover both use Docker Swarm for orchestration, but they serve different needs. I see Dokploy as the modern choice for teams that want native multi-node clustering and a clean interface. It has 36,826 GitHub stars as of August 2026, though its v0.30.2 status means interfaces might shift. You must check the license because everything in the /proprietary directory requires a commercial agreement, a condition that makes license purity a vital consideration for teams before they commit their production workloads to the platform. Dokploy supports Nixpacks, Heroku, and Paketo buildpacks, which provides flexibility for modern stacks. It uses a clean UI built with Next.js and TypeScript to provide a modern developer experience. Because it provides built-in container monitoring for CPU, RAM, and network usage, you can track resources in real time. CapRover, a veteran since 2017 with 15,139 stars, provides stability but lacks strong Docker Compose support. It relies on Nginx and has hit 100 million Docker Hub pulls, making it a highly proven tool for container deployment. I find the CapRover UI dated when compared to the Next.js builds of newer tools.

Scaling the engineering workload

The math of self-hosting heavily favors teams with fewer than 5 engineers. You already know that a $5 Hetzner VPS is a bargain compared to a $25 Heroku Standard-1X dyno. However, you must pay for the engineer hours required to manage updates and backups. If you use the 2024 Stack Overflow Developer Survey median US salary of $145,000, then one hour of weekly babysitting for your control plane costs $3,640 per year in engineer time. This cost compounds as your team grows. Once you reach 5 to 10 engineers or face your first compliance questionnaire, the manual scaling and thin RBAC of self-hosted tools like Coolify or CapRover will fail you. You must decide based on who patches the control plane, whose cloud account runs the workloads, how far the orchestration engine scales, and what governance you must prove to an auditor. The decision rule for growing teams remains simple: under 5 engineers with one environment, stay self-hosted. At 5 to 10 engineers with staging plus production, or the first compliance questionnaire, move to a managed or BYOC platform. Once you hit those points, the operational burden outweighs the savings of a $5 VPS. If your team grows beyond that, you should move to a managed or BYOC platform to avoid 2am incidents. Which tool survives the next major licensing shift?