~/products/compute
// core-platform/compute
Containers, fully managed
Compute runs your app in managed containers — a container is your app packaged with everything it needs to run. Each one is kernel-isolated in its own lightweight VM, restarted automatically if it fails, and scaled with a single command. You're billed per second for the vCPU and memory allocated to each container. Containers keep running and billing while idle. HTTPS and traffic routing included. AI agents can run serverless instead — see Agents below.
$ setcloud scale api --count 3
✓ Scaled "api" to 3 task(s).
$ setcloud logs api --follow
Following logs (Ctrl+C to stop)...
[api] listening on :8080
[api] GET /healthz 200// 01 — services
Create a service, upload your container image using credentials the command-line tool fetches for you, and the platform runs it — networking, HTTPS, and restarts all handled. Or skip the image entirely and let Deployments build straight from your repo.
// 02 — scaling
Run setcloud scale api --count 3 and you're serving three copies of your app to handle more traffic. Containers that crash or stop responding are replaced automatically — no one has to wake up to restart anything.
// 03 — operations
Open Compute in the dashboard or use the CLI to operate standalone container services, follow logs, and roll back to the previous revision. Repository-backed deployments have their own project dashboard, logs, and terminal. Container operations are recorded in an audit log. Run bounded shell commands in the service terminal or with setcloud exec; your image must include /bin/sh. Sessions expire and commands have a five-minute maximum timeout.
// 04 — isolation
Containers run in isolated execution environments with scoped workload credentials. Database access uses project-specific credentials on shared private infrastructure. Your application still controls its own authentication and authorization.
// 05 — serverless agents
Create a dashboard agent with your instructions and tools, or import your own compatible repository under Agents → Custom code. Custom serverless execution scales to zero between requests unless always-warm is enabled. Your runtime must implement its own conversation persistence and recovery; startup latency varies.
USD rates from the current billing catalog. Regional references vary with attributed usage.
Loading current rates…
Allocated container CPU and memory bill while running. See Pricing for the reference estimate, exclusions, and optional custom-agent runtime rates.