What a Preview Environment per Agent PR Really Costs, and How to Cap It
Blog post from Qovery
A cost model for full-stack pull-request preview environments on an existing EKS cluster in AWS us-east-1 estimates that an API, server-rendered frontend, Postgres database, and Redis service using 1.1 vCPU, 2.25 GiB of memory, and 10 GiB of gp3 storage costs about $0.07 per hour on on-demand m6i.large capacity, or roughly $1.67 per day. The analysis identifies environment lifetime as the largest cost driver, with a weekly-lived preview costing about $11.72 and forgotten monthly previews exceeding $50 each. For 30 agent-generated PRs per working day, weekly manual cleanup could cost approximately $7,700 monthly, while automatic deletion on merge or closure reduces this to about $1,100; combining working-hours schedules, on-demand creation for only 40% of PRs, and Spot capacity for stateless application pods lowers the estimate to about $95. Stateful services are kept on on-demand nodes to avoid reliability issues, while Spot is proposed for API and frontend workloads. The model excludes shared EKS control-plane, ingress, networking, image-build, registry, and potentially significant data-transfer or database-seeding costs. It recommends dedicated preview node pools with maximum size limits as a hard spending cap, causing excess environments to remain pending rather than scaling cloud costs indefinitely. Qovery Free is described as supporting preview workflows on customer-managed clusters with a 1,000 deployment-minute monthly limit, although AWS infrastructure costs continue independently of that limit.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| Kubernetes | 6 | 1,368 | 107 | 49 | -73% |
| AI Agents | 3 | 2,096 | 365 | 132 | -70% |
| Developer Experience | 1 | 64 | 47 | 16 | -87% |
| MCP | 1 | 3,402 | 292 | 104 | -71% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.