Compute & Containers

Everything you’d reach for to answer “what’s running, and can you change it?” — across Compute Engine VMs and the two managed container platforms.

Compute Engine

• list_compute_instances(project_id) — VM instances across every zone, with name, zone, and status.
• start_compute_instance(project_id, zone, instance) — write. Starts a stopped VM.
• stop_compute_instance(project_id, zone, instance) — write. Stops a running VM.
• resize_compute_instance(instance, zone, machine_type, project_id) — write. Changes a VM’s machine type. The VM must be stopped first; resize then restart.
• create_compute_snapshot(disk, zone, snapshot_name, project_id) — write. Takes a snapshot of a disk — a quick way to checkpoint before a risky change.

Cloud Run

enoki does more than read Cloud Run — it can roll back and split traffic, both approval-gated. That has its own page: Cloud Run.

GKE

• list_gke_clusters(project_id) — Kubernetes Engine clusters in a project, with location and status.

GKE is read-only today — enoki tells you what’s deployed, but rollouts and scaling stay in your own pipelines.

Cloud Functions

• list_cloud_functions(project_id) — functions across every region, covering both 1st- and 2nd-gen, each with its state and environment.
• describe_cloud_function(function, location, project_id) — one function’s runtime, entrypoint, trigger (HTTP or event), and state.

Cloud Functions is read-only here — deploying a function needs a source upload and build, which is exactly the free-form surface the Guardrails rule out.

Next: Cloud Run.